Bulletin 35

Part 5
(of 7 Parts)

THE
q; SHOCK AND VIBRATION
BULLETIN

FEBRUARY 1966

A Publication of
THE SHOCK AND VIBRATION
INFORMATION CENTER
U.S. Naval Research Laboratory, Washington, D.C.

C.L.C-A.R I G140D
F..0
ID C
TECLUNICARLNGr,"o; ;
'PC~r~' J'.7 f)A APR 221966 1

v rr
rrg:tlnon

('~,v-..e / Office of
The Director of Defense
Research and Engineering

I)ISTRIBUI ON OF TiHIS D(:UMENT IS UNI.IMITED

 Bulletin 35
Part 5
(of 7 Parts)

THE
SHOCK AND VIBRATION
BULLETIN

FEBRUARY 1966

A Publication of
THE SHOCK AND VIBRATION
INFORMATION CENTER
U.S. Naval Research Laboratory, Washington, D.C.

The 35th Symposium on Shock, Vibration

and Associated Environments was held in
New Orleans, Louisiana on 25-28 October
1965. NASA Marshall Space Flight Center
was host.

Office of
The Director of Defense
Research and Engineering
 CONTENTS

PART 5

Transportation

A SURVEY OF SHOCK AND VIBRATION ENVIRONMENTS IN THE FOUR MAJOR

MODES OF TRANSPORTATION ........... ......................................
R. W. Schock, NASA, Marshall Space Flight Center and W. E. Paulson,
Brown Engineering Company, Kuntsville, Alabama

THE DYNAMIC ENVIRONMENT OF SPACECRAFT SURFACE TRANSPORTATION ....... 2I

J. W. Schlue, Jet Propulsion Laboratory, Pasadena, California

A REALISTIC DERIVATION OF A LABORATORY VIBRATION TEST TO SIMULATE

THE OVERLAND TRANSPORTATION ENVIRONMENT ...... ................... .... 37
John A. Kasuba, Aberdeen Proving Ground, Maryland

A METHOD FOR ESTIMATING ACCELERATIONS 0 SHIPPING CONTAINERS

MOUNTED ON AN IMPACTING RAILROAD CAR .\. . ................... 49
John J. Scialdone, NASA, Goddard Space Flight Centt r, Greenbelt, Maryland

PRELIMINARY ANALYSIS OF DATA OBTAINED IN THE, JOINT ARMY/AEC/SANDIA

TEST OF TRUCK TRANSPORT ENVIRONMENT .... \ .................. 57
J. T. Foley, Sandia Corporation, Albuquerque, New Mexico

ROUGHNESS MEASUREMENT AND SYSTEM RESPONSE EVALUATION FOR

HIGHWAY ENVIRONMENT ........................ ................ 71
J. R. Harvey and R. A. Wursche, Goodyear Aerospace Corl0ration,
Litchfield Park, Arizona

PROGRAM 624A-TITAN III-C TRANSPORTATION TESTS. . ..... ................. 87

Louis A. Molinari, United Technology Center, Sunnyvale, Calif6rnia and
Jack R. Reynolds, Space System Division, AFSC, Los Angeles, California

DYNAMIC ENVIRONMENT OF M-113 ARMORED PERSONNEL CARRIER ..... ........ 115

Q. M. Pomonik and N. G. Tinling, Hughes Aircraft Company, Culver City, California

INTERPRETATION AND APPLICATION OF SPECIFICATION REQUIRENyENTS

THAT SIMULATE VIBRATION RESPONSES OF EQUIPMENT BEING SHIPPED
BY COMMON CARRIER ................ .................................... 129
James E. Rice, Goodyear Aerospace Corporation, Akron, Ohio

Shock and Vibration Isolation

NEAR-OPTIMUM SHOCK MOUNTS FOR PIROTECTING EQUIPMENT FROM

ACCELERATION PULSES ............................................ 133
Ralph E. Blake, Lockhced Missiles & Space Company, Sunnyvale, California

SHOCK ATTENUATION USING PASSIVE ELEMENTS.. .......

. . .
. . .. . . 147
V. H. Neubert and D. L. Pyke, Pennsylvania State University, University Park, Pa.,
and D. F. Pocth, HRB-Singer Corporation, State College, Pennsylvania

INVESTIGATION OF THE VIBRATION DAMPING PROPERTIES OF VISCOELASTIC

MATERIAL USING A DELAY ANGLE TECHNIQUE ...... . ................ . . 159
Saul A. Eller and Levi Cohen, U.S. Naval Applied Science Laboratory,
Brooklyn, New York

ENERGY ABSORPTION BY DYNAMIC CRUSHING ............ ............... . 169

C. V. David, General Atomic Division of General Dynamics Corporation,
San Diego, California
i
 ______________

DRY tPRICTION DAMPING WITH FORCE PROPORTIONAL TO DISPLACEMENT ........ 179

Leon Wallerstein, Jr., Lord Manufacturing Company, Erie, Pennsylvania

RESPONSE OF RESILIENT MOUNTS UNDER SHOCK LOADING .................. ..... 187

E. A. Thornton and R. D. Short, Underwater Explosions Research Division,
x- David Taylor Model Basin, Portsmouth, Virginia
} "-,j
" A NEW APPROACH TO PACKAGE CUSHIONING DESIGN ...... .................. .... 193
Gordon S. Mustin, Special Projects Consultants, Inc., Washington, D.C.

OPTIMAL SHOCK ISOLATION SYNTHESIS ........ ........................... . . .. 203

T. Liber and E. Sevin, lIT Research Institute, Chicago, Illinois

VIBRATION ISOLATION THROUGH PNEUMATIC SPRING AND DAMPING .......... .... 217
Russel L. Hall, Edgerton, Germeshausen & Grier, Inc., Las Vegas, Nevada

DESIGN AND ADVANTAGES OF A TWO STAGE MOUNTING SYSTEM FOR THE

MAJOR MACHINES IN A SHIP'S ENGINE ROOM ......... ...................... 227
f
Robert M. Gorman, General Dynamics/Electric Boat Division, Groton, Connecticut

MODELING OF SPACECRAFT FOR LOW FREQUENCY NOISE REDUCTION ............ 235

Richard H. Lyon, Robert E. Apfel and Charles W. Dietrich, Bolt Beranek and
Newman, Inc., Cambridge, Massachusetts

FLEXIBLE TOROIDAL SPRING CHARACTERISTICS .......... ..................... 243

C. V. David, General Atomic Division of General Dynamics Corporation,
San Diego, California

VIBRATION QUALIFICATION SPECIFICATION FOR INERTIAL MEASURING UNITS .... 261

S. L. Burgwin, Honeywell, Inc., Aeronautical Division,-St. Petersburg, Florida

EVALUATION OF A DESIGN FACTOR APPROACH TO SPACE VEHICLE DESIGN

FOR RANDOM VIBRATION ENVIRONMENTS . . . . ......... ................... 271
D. E. Hines and D. A. Stewart, Douglas Aircraft Company, Santa Monica, California

AN APPLICATION OF DECISION THEORY TO A VIBRATION-FATIGUE PROBLEM .... 307

D. M. Aspinwall and R. E. Blake, Lockheed Missiles and Space Company, "
Sunnyvale, California
A GENERAL SOILS MODEL FOR SHOCK PROPAGATION STUDIES ................ .... 321
C. X. C. F. Miranda, University of Detroit, Detroit, Michigan and
David C. Kraft, University of Dayton, Dayton, Ohio

DISTRIBUTION ................... .......................................... 329

PAPERS APPEARING IN PART 1

Part 1 - Confidential
(Titles Unclassified)

BLAST LOADING OF MODEL ANTENNA STRUCTURES

R. Kirk Gregory, Southwest Research Institute, San Antonio, Texas

ANALYSIS OF RESPONSE OF EQUIPMENT ON A DROPTEST SHOCK MACHINE

Robert L. Bort, David Taylor Model Basin, Washington, D.C.

BACKGROUND AND CURRENT STATUS OF UNDERWATER EXPLOSION SHOCK

SPECIFICATIONS AND TESTING
Gerald M. Mayer, U.S. Navy Underwater Sound Laboratory, New London, Connecticut

A SIMULATED Z5-30 CPS DECK FOR SHOCK TESTING

Culver J. Floyd, Raytheon Submarine Signal Division, Portsmouth, Rhode Island

MAGNETIC TAPE RECORDING IN A SEVERE MISSILE ENVIRONMENT-A CASE HISTORY

J. P. White and J. Montsma, Bell Telephone Laboratories, Whippany, New J.ersey

LONGITUDINAL VIBRATIONS PRODUCED BY INTERACTION OF THE STRUCTURE

PROPELLANT FLOW, A COMBUSTION PROCESS IN THE LANCE PROPULSION SYSTEM
Ernest King Bramblett, II, Rocketdyne, Canoga Park, California

iv
 SIMULATION OF IMPULSIVE ENVIRONMENTS BY USE OF PYROTECHNIC DEVICES
F. A. Ottati, Avco Corporation, Research and Advanced Development Division,
Wilmington, Massachusetts

PAPERS APPEARING IN PART 2

Vibration Testing

THEORY OF DYNAMIC TESTS OF STRUCTURES

N. Norby Nielsen, Department of Civil Engineering, University of Illinois, Urbana, Illinois

FATIGUE LIFE UNDER VARIOUS RANDOM LOADING SPECTRA

Sherman A. Clevenson and Roy Steiner, NASA Langley Research Center,
Langley Station, Hampton, Virginia

COMBINED BROADBAND AND'STEPPED NARROWBAIND RANDOM VIBRATION

A. J. Curtis, J. G. Herrera, and R. F. Witters, Hughes Aircraft Company

ON THE USE OF MULTIPLE (MULTI-POINT) RANDOM EXCITATION WITH

APPLICATION TO SURVEYOR SPACECRAFT TESTS
A. J. Curtis, H. T. Abstein, and R. J. Varga, Hughes Aircraft Company,
Culver City, California

GROUND TEST SIMULATION OF LIFT-OFF AND TRANSONIC VIBRATION EXCITATION

MECHANISMS ON THE RANGER SPACECRAFT
Marc C. Trummel, Jet Propulsion Laboratory, California)Lnstitute of Technology,
Pasadena, California

METHODS OF CONTROL OF MULTIPLE SHAKER TESTING SYSTEM

Jack D. Newton, MB Electronics, New Haven, Connecticut

ON SYNERGETIC EFFECTS OF COMBINATIONS OF VIBRATION AND OTHER ENVIRONMENTS

W. P. Rader, J. D. Donahue, H. N. McGregor, and E. R. Wilson, Martin Company,
Denver, Colorado

A MULTIPLE-FREQUENCY, SHAKE-TESTING TECHNIQUE FOR STRUCTURES WITH

RAPIDLY-CHANGING DYNAMIC CHARACTERISTICS
1 '7" F. J. Hawkins, C. W. Skingle, and G. A. Taylor, Royal Aircraft Establishment,
Farnborough, England

A PROGRESS REPORT ON FORCE CONTROLLED VIBRATION TESTING

C. E. Nuckolls and J. V. Otts, Sandia Corporation, Albuquerque, New Mexico

VIBRATION ON SANDWICH PANELS IN A VACUUM

Clemans A. Powell, Jr., and David G. Stephens, NASA Langley Research Center,
Langley Station, Hampton, Virginia

THE VIBRATION DESIGN APPROVAL AND ACCEPTANCE TEST PROGRAM FOR THE
GEMINI SPACECRAFT - COMPONENT, MODULE AND WHOLE VEHICLE TESTING
James R. Daiber and Vincent S. Noonan, McDonnell Aircraft Corporation,
St. Louis, Missouri

TECHNIQUES FOR EXTENDING THE CAPABILITY OF RANDOM EQUALIZATION EQUIPMENT

C. D. Robbins and D. G. Clack, LTVElectrosystems, Inc., Garland Division, Dallas, Texas

A BROADBAND HYDRAULIC VIBRATION EXCITER

H. T. Strandrud, The Boeing Company, Seattle, Washington

ADVANTAGES OF MULTIPOINT CONTROL FOR VIBRATION TESTING OF COMPLETE

RANGER FLIGHT SPACECRAFT
Jack L., Cooper, Jet Propulsion Laboratory, Pasadena, California

VIBRATION INPUT CONTROL INVESTIGATION

Roger L. Carlson and Fred A. Chinquist, Honeywell, Inc., Minneapolis, Minnesota

A UNIQUE SUSPENSION SYSTEM FOR LONGITUDINAL VIBRATION TESTING OF LARGE

LAUNCH VEHICLES
James A. Schoenster, Jerome Pearson, and Grayson V. Dixon, NASA Langley Research
Center, Langley Station, Hampton, Virginia

V
 COMEINED RANDOM VIBRATION AND EXTREME TEMPERATURE TESTING OF
INTEGRATED CIRCUITS
James C. Burrus, Texas Instruments, Inc., Dallas, Texas

SI,
-.
IMPLICATIONS OF SPACECRAFT VIBRATION QUALIFICATION TESTING REQUIREMENTS
ON STRUCTURAL DESIGN
M. J. Baruch and S. Davis, Fairchild Hiller, Republic Aviation Div., Farmingdale, L. I., New York

USE OF IMPULSIVE LOADING TECHNIQUES IN THE STUDY OF SHIP VIBRATION

J. G. Viner, A. F. Kilcullen, and D. L. Ludwig, DTMB Acoustics and Vibration Laboratory,
Washington, D.C.
THE RELATIONSHIP BETWEEN A LOGARITHMICALLY SWEPT EXCITATION AND THE
BUILD-UP OF STEADY-STATE RESONANT RESPONSE
R. E. rMorse, TRW Systems, Redondo Beach, California

ANALYSIS OF THE ROTOR BLADE VIBRATORY STRESSES OF THE PROPULSION WIND

TUNNEL COMPRESSORS
R. A. Robinson, J. C. Ghilders, and D. I. Yando, ARO, Inc., Arnold Engineering Development
Center, Arnold Air Force Station, Tennessee

PAPERS APPEARING IN PART 3

Structural Analysis

ON THE RESPONSE OF ROCKET VEHICLE STRUCTURE TO CERTAIN ENVIRONMENTAL LOADS

John C. Yao, Northrop Corporation, Norair Division, Hawthorne, California

EFFECTS OF ORTHOTROPIC CORES ON THE FREE VIBRATIONS OF SANDWICH PLATES

M. J. Jacobson, Northrop Corporation, Norair Division, Hawthorne, California

THE EFFECT OF INERTIA VARIATION AND CERTAIN ASYMMETRIES ON THE

DYNAMIC RESPONSE OF AN ELASTICALLY-COUPLED SYSTEM
Darrell A. Frohrib, University of Minnesota and UNIVAC Division of
Sperry Rand Corporation, St. Paul, Minnesota

STEADY-STATE RESPONSE OF A MULTI-DEGREE OF FREEDOM SYSTEM SUBJECTED

TO RANDOM EXCITATION
J. J. Vaccaro, A Division of North American Aviation, Inc., Canoga Park, California

VIBRATIONS OF MULTILAYER SHELLS OF REVOLUTION UNDER DYNAMIC AND

IMPULSIVE LOADING
Stanley Klein, Aerospace Corporation, San Bernardino, California

THE RANDOM VIBRATIONS OF A MULTI-SUPPORTED HEAVILY-DAMPED BEAM

Denys J. Mead and Emily F. Wilby, Institute of Sound and Vibration Research,
University of, Southampton, England

CUMULATIVE DAMAGE CAUSED BY SHOCK EXCITATION

Sami F. Masri, Research Fellow, California Institute of Technology, Pasadena, California

SOME LATTICE VIBRATION PROBLEMS

George W. Morgenthaler, Martin Company, Denver, Colorado

THE SOLUTION OF DYNA;MIC BEAM PROBLEMS BY MEANS OF FINITE

Cis-HYPERBOLIC TRANSFORMSi
Gabriel, Cinelli, Argonne National Laboratory, Argonne, Illinois

RANDOM VIBRATIONS AND RANDOM FUNCTIONS

Clifford S. O'Hearne, 'Martin Company, Orlando, Florida
ANALYTICAL AND EXPERIMENTAL TECHNIQUES USED TO ESTABLISH STRUCTURAL
DESIGN LOADS FOR THE SURVEYOR SPACECRAFT DURING LUNAR LANDING
R. J. Harter and R. J. Switz, Hughes Aircraft Company, El Segundo, California

DYNAMIC ASPECTS OF METAL BELLOWS

Vincent R. Daniels, Bell Aerosystems Company, Buffalo, New York

vi
 LIFETIME EVALUATION PROCEDURES FOR RANDOM SHOCK AND VIBRATION
M. Zaid and P. Marnell, Technik Incorporated, Jericho, New York

APPLICATION OF THE DIRECT STIFFNESS METHOD TO THE ELASTIC MATHEMATIC

MODELING OF THE SATURN S-IB
M. E. McCoy, J. D. Sowers, W. J. Leaumont, and R. G. Higgins, Chrysler Corporation
Space Division, New Orleans, Louisiana

ACOUSTIC WAVES GENERATED BY THE MOTION OF PIPING CONTAINING A FLUID

Norman Lipner and Francis B. Fay, TRW Systems, Redondo Beach, California

ANALYSIS OF THE EFFECTS OF DYNAMIC COUPLING BETWEEN STRUCTURAL

RESONANCES AND A TIME SAMPLING DATA CONTROL SYSTEM t
L. R. Beuder and J. M. Brust, Nortronics, A Division of Northrop _,orporati~n,
Hawthorne, California

TRANSIENT RESPONSE OF LINEAR DAMPED, LUMPED, SPRING-MASS SYSTEMS BY

EXPERIMENTALLY DERIVED TRANSFER FUNCTIONS
R. W. Kelley, Sandia Laboratory, Albuquerque, New Mexico

INVESTIGATION OF DYNAMIC CHARACTERISTICS OF A 1/20th SCALE MODEL OF

THE LAUNCH PHASE SIMULATOR
Charles J. Arcilesi and Lloyd R. Bruck, NASA, 'Goddard'Space Flight Center, Greenbelt, ,Md.

THE ELIMINATION OF SUBSTANDARD PARTS BY ENVIRONMENTAL TESTING

Lyle E. Matthews, US. Naval Missile Center, Point Mugu, California, and
Raymond C. Binder, University of Southern California,, Los Angeles, California

A PRACTICAL APPROACH TO THE DETERMINATION OF ELECTRICAL SUPPORT

EQUIPMENT TEST REQUIREMENTS WHICH ASSURE PROPER OPERATION IN HIGH
STRESS SERVICE ENVIRONMENT
James M. Ray and Edwin' B. Blanford, General Electric Company,
Apollo Support Department, Daytona Beach, Florida

HIGH IMPACT SPACECRAFT EQUIPMENT

J. L. Adams and M. G. Comuntzis, jet Propulsion: Laboratory, Pasadena, California

EVALUATION OF ACOUSTIC ENVIRONMENTAL EFFECTS ON FLIGHT ELECTRONIC

EQUIPMENT
J. Sugamele, The Boeing Company-Airplane Group, Seattle, Washington

PAPERS APPEARING IN PART 4

Instrumentation

INTEGRATION AND DOUBLE INTEGRATION-A PRACTICAL TECHNIQUE

Merval W. Oleson, US. Naval Research Laboratory, Washington, D.C.
THE MEASUREMENT OF INTERNAL DYNAMICS OF EQUIPMENT
Charles T. Morrow, The Aerospace C6rporation, Los Angeles, California

PIEZORESISTIVE STRAIN GAGE ACCELEROMETERS INCREASE SPECTRUM OF SHOCK

AND VIBRATION MEASUREMENT CAPABILITY
W. E. Wall, Endevco Corporation, Pasadena,GCalifornia

VIBRATION MEASUREMENTS USING A MICROWAVE INTERFEROMETER

C. F. Augustine and J. E. Ebert, Weinschel -Engineering, Gaithersburg, Maryland

A WIDEBAND ABSOLUTE ACCELEROMETER CALIBRATOR UTILIZING A LASER FOR

MEASURING VIBRATORY DISPLACEMENTS
Robert B. Davis, Naval Air Test Center, Patuxent River, Maryland

UNPUBLISHED ACCELEROMETER CHARACTERISTICS

Boris Mangolds, Astro-Electronics Division, Radio Corporation of America,
Princeton, New Jersey

NEW PRECISION CALIBRATION TECHNIQUES FOR VIBRATION TRANSDUCERS

Walter P. Kistler, Kistler Instrument Corporation, Clarence; New York

vii-

i'*
 S - * -

OMNIDIRECTIONAL ACCELERATION SENSOR

A. J. Buschman, Jr., Harry Diamond Laboratories, Washington, D.C.

EFFECTIVE USE OF ACCELEROMETERS AS CALIBRATION STANDARDS

D. R. Workman, Lockheea Missiles and Space Company, Sunnyvale, California

MEASUREMENT OF ACCELEROMETER TRANSVERSE SENSITIVITY

D. W. Rockwell and J. D. Ramboz, Metrology Engineering Center, Bureau of Naval Weapons
Representative, Pomona, California
FIRST OCCURRENCE PROBABILITIES FOR EXTREME RANDOM VIBRATION AMPLITUDES
Cory L. Gray, Measurement Analysis Corporation, Los Angeles, California
SIGNAL DETECTION USING IMPULSE CROSSCORRELATION
S. W. Marshall, Texas Instruments, Inc., and A. C. Keller, White Sands Missile Range

A MEANS TO REDUCE RANDOM VIBRATION ANALYSIS TIME

N. Bahringer and R. W. Lochner, Jr., Honeywell, Inc., St. Petersburg, Florida
A CONTINUOUS FREQUENCY CONSTANT Q SHOCK SPECTRUM ANALYZER
G. W. Painter and H. J. Parry, Lockheed-California Company, Burbank, California

SLOPE ERROR OF POWER SPECTRAL DENSITY MEASUREMENTS

Robert L. Gordon, Pratt and Whitney Aircraft, East Hartford, Connecticut

THE EFFECTS OF PHASE ERRORS UPON THE MEASUREMENT OF RANDOM PROCESSES

Ronald D. Kelly, Measurement Analysis Corporation, Los Angeles, California
UTILIZATION OF A DIGITAL COMPUTER FOR ON-LINE ACQUISITION AND ANALYSIS
OF ACOUSTIC AND VIBRATION DATA
Daniel J. Bozich, Wyle Laboratories ResearchStaff, Huntsville, Alabama

A DIGITAL DATA RECORDING SYSTEM FOR STRUCTURAL DYNAMICS RESPONSE TESTING

M. H. Hieken, McDonnell Aircraft Corporation, St. Louis, Missodii

COMPARISON OF ANALOG AND DIGITAL METHODS FOR VIBRATION ANALYSIS

William K. Shilling, III, AF Flight Dynamics Laboratory, Wright-Patterson
Air Force Base, Ohio
AUTOMATIC REAL-TIME VIBRATION SPECTRUM ANALYZER SYSTEM USING DELAY
LINE TIME COMPRESSION TECHNIQUES
John L. Fryling, Gulton Industries, Inc., Trenton, New Jersey

THE ANALOG CROSS SPECTRAL DENSITY ANALYZER SYSTEM

R. L. Randall, Atomics International, 'Canoga Park, California

TRANSIENT DATA DISTORTION COMPENSATION

John D. Favour, The Boeing Company, Seattle, Washington

PAPERS APPEARING IN PART 6

Shock Testing

SHOCK TESTING TO SIMULATE RANDOM VIBRATION PEAKS \

1
J..A. Bailie, Lockheed Missiles and Space Company, Sunnyvale, California

10,000 G SLINGSHOT SHOCK TESTS ON A MODIFIED SAND-DROP MACHINE

Sam Marshall and LaVerne Root, Collins Radio Company, Cedar Rapids, Iowa and
Leonard Sackett, University of Michigan, Dearborn, Michigan

SHOCK SPRINGS AND PULSE SHAPING ON IMPACT SHOCK MACHINES

Richard 0. Brooks, Sandia Corporation, Albuquerque, New Mexico
SHOCK TESTING A SPACECRAFT TO SHOCK RESPONSE SPECTRUM BY MEANS OF
AN ELECTRODYNAMIC EXCITER
G. A. Gallagher, M.I.T. Lincoln Laboratory, Lexington, Massachusetts and
A. W. Adkins, Littleton Research and Engineering Corporation, Littleton, Massachusetts

viii
 -. 4

DESIGN AND PERFORMANCE CHARACTERISTICS OF A WATER JET ACTUATOR

Robert L. Henderson, Sandia Corporation, Albuquerque, New Mexico

RE-ENTRY OVERPRESSURE SHOCK SIMULATION TEST

William R. Kampfe, Sandia Corporation, Albuquerque, New Mexico

SHOCK TESTING WITH HIGH EXPLOSIVE INITIATED GAS DETONATIONS

W. M. Sigmon, Jr., Sandia Corporation, Albuquerque, New Mexico

A METHOD, OF PRODUCING LONG-DURATION AIR-INDUCED GROUND SHOCK

USING HIGH EXPLOSIVES
Eugene Zwoyer, George Triandafilidis and James Stras, Eric H. Wang Civil Engineering
Research Facility, University of New Mexico, Albuquerque, New Mexico

STRUCTURAL AND FUNCTIONAL TESTS OF A FULL-SCALE GEMINI RENDEZVOUS AND

RECOVERY SECTION AND AN AGENA TARGET 'DOCKING ADAPTER AS SUBJECTED
TO AN ORBITAL MOORING SHOCK ENVIRONMENT
N. E. Stamm and J. F. Siller, McDonnell Aircraft Corporation, St. Louis, Missouri

EXTENSION OF SHAKER SHOCK CAPABILITIES

James M. McClanahan and John Fagan, Astro-Electronics Division, RCA, Princeton, New Jersey

A NEW AIR GUN FOR SHOCK TESTING

Evan W. Gray and Tom B. Linton, U.S. Naval Weapons Station, Seal Beach, California

INVESTIGATION OF PARAMETERS AFFECTING DYNAMIC PRESSURES IN SUPER

PRESSURE GENERATOR USED FOR CANNON BREECH FATIGUE STUDIES'
R. R. Lasselle, J. E. weig and M. A. Scavullo, Watervliet Arsenal, Watervliet, New York
THE DOUBLE FORCE PROGRAMMER SHOCK TESTING METHOD - A NEW TECHNIQUE
FOR CONTROLLING SHOCK PULSE WAVEFORMS
F. H. Mathews, Sandia Corporation, Albuquerque, New Mexico

Specification of Shock

AEROSPACE SHOCK TEST SPECIFIED AND MONITORED BY THE RESPONSE SPECTRUM

K. Kuoppamaki, Consultant, Riverside, California and R. A. Rouchon, Lockheed Missiles
and Space Company, Sunnyvale, California

SPECIFICATION
Trwin Vigness,OFU.S.
ACCELERATION PULSES FOR SHOCK TESTS
Naval Research Laboratory, Washington, D.C

SHOCK TESTING TO SHOCK SPECTRA SPECIFICATIONS

S. M. Ostergren, General Electric Company, Philadelphia, Pennsylvania

ANALOG METHOD FOR STUDY OF SHOCK SPECTRA IN NONLINEAR SYSTEMS

Theodore F. Bogart, Jr., LTV Ling-Altec, Inc., Anaheim, California

A MECHANICAL SHOCK PULSE SURVEY

Francesco Palmisano, U.S. Army Electronics Command, Ft. Monmouth, New Jersey

PROXIMITY SPECTRUM-A NEW MEANS OF EVALUATING SHOCK MOTIONS

Edward H. Schell, Air Force Flight Dynamics Laboratory, Wright-Patterson
Air Force Base, Ohio

DEFINITION OF SHOCK DESIGN AND TEST CRITERIA USING SHOCK AND FOURIER
SPECTRA OF TRANSIT ENVIRONMENTS
M. C-ertel and R. Holland, Allied Research Associates, Inc., Concord, Massachusetts

STRUCTURAL RESPONSE TO IMPULSIVE LOADING (PYROTECHNIC DEVICES)

Vincent S. Noonan and William E. Noonan, McDonnell Aircraft Corporation,
St. Louis, Missouri

ANALYTIC DYNAMIC MODELING FOR IMPULSIVE ENVIRONMENTS

William C. Broding and John R. Henry, Avco Corporation, Research and Advanced
Development Division, Wilmington, Massachusetts

ix

...
.. . . .. .... ... . .. ..
 PYROTECHNIC SHOCK ANALYSIS AND TESTING METHODS
Allan R. Hoffman and James E. Randolph, Jet Propulsion Laboratory, Pasadena, California

SHAPED CHARGE SHOCK ENVIRONMENT FOR CENTAUR VEHICLE COMPONENTS

E. C. Noble, Jr., and fL. L. Batten, General Dynamics/Convair, San Diego, California
L VIBRATION AND SHOCK DATA FROM THE ATHENA BOOSTER
C. J. Moening and F. J. Benedetti, Aerospace Corporation, San Bernardino, California

PAPERS APPEARING IN PART 7

Prediction of Vibration Environment

TITAN I1-C AERODYNAMIC VIBRATION ANALYSIS

F. M. Condos, Martin Company, Denver, Colorado

PREDICTION OF RE-ENTRY VIBRATION

F. A. Smith and F. J. Benedetti, Aerospace Corporation, San Bernardino, California

PREDICTION OF THE BOUNDARY LAYER ACOUSTIC PRESSURE LEVELS OF A BLUNT

NOSE RE-ENTRY VEHICLE AT HIGH MACH NUMBERS
H. Saunders and D. E. Nestler, General Electric Company, Philadelphia, Pennsylvania

HIBEX MISSILE VIBRATION ENVIRONMENT CONSIDERATIONS

J. C. Furlong and H. M. Voss, The Boeing Company, Seattle, Washington

MARINER MARS 1964 ACOUSTICALLY INDUCED VIBRATION ENVIRONMENT

R. A. Schiffer and J. R. Hyde, Jet Piopulsion Laboratory, Pasadena, California

VIBRATION STUDIES CONDUCTED ON THE GEMINI INERTIAL PLATFORM

Dean Seefeld, Honeywell, Inc., Aeronautical Division, St. Petersburg, Florida

GEMINI SPACECRAFT FLIGHT VIBRATION DATA AND COMPARISON WITH PREDICTIONS

James A. Callahan, McDonnell Aircraft Corporation, St. Louis, Missouri

DETERMINATION OF THE WATER IMPACT SHOCK ENVIRONMENT

F. H. Collopy, ITEK Corporation, Lexington, Massachusetts

ORBITAL VIBRATION EFFECTS ON SCIENTIFIC INSTRUMENTS

R. J. Savage, Douglas Aircraft Company, Santa Monica, California

MOBILITY OF A RANDOMLY EXCITED STRUCTURE BY CROSS-CORRELATION AND

CROSS-SPECTRAI COMPUTATION TECHNIQUES
Will Gersch and David Sharpe, Purdue University, Lafayette, Indiana
COMPUTATION TECHNIQUE FOR MECHANICAL IMPEDANCE METHODS
James E. Smith, U.S. Navy Marine Engineering Laboratory, Annapolis, Maryland

ANALYTICAL AND EXPERIMENTAL EVALUATION OF DD 69Z CLASS HULL VIBRATION

J. J. Francis, Boston Naval Shipyard, Boston, Massachusetts and A. R. Paladino,
Bureau of Ships, Washington, D.C.

A PRACTICAL APPROACH TO THE PREDICTION OF THE NATURAL FREQUENCY OF

PRINTED CIRCUIT BOARDS
James T. Arnold and Fred P. Strother, Jr., Honeywell, ,Inc., Aeronautical Division,
St. Petersburg, Florida

MODAL CHARACTERISTICS OF ARBITRARILY-STIFFENED RINGS

John D. Sowers and Nolan P. Hymel, Chrysler Corporation Space Division,
New Orleans, Louisiana

MEASUREMENT OF CAPTIVE-FLIGHT VIBRATION RESPONSE FOR AGM-45A (SHRIKE)

AND FIREYE BOMB AIRBORNE WEAPONS
A. Weathersbee and W. Parmenter, U.S. Naval Ordnance Test Station, China Lake, California

Damping and Nonlinear Analysis

RESPONSE OF A NONLINEAR DAMPED OSCILLATOR TO RANDOM EXCITATION

A. W. Bell and A. E. Galef, National Engineering Science Company, Pasadena, California
X
 4
-

MATHEMATICAL MODELS OF NONLINEAR MULTIDEGREE-OF-FREEDOM STRUCTURES

Walter B. Murfin, Sandia Corporation, Albuquerque, New Mexico

FORCE TRANSMISSIBILITIES IN SPACECRAFT STRUCTURES

' Carl C. Osgood, Astro-Electronics Division, RCA, Princeton, New Jersey

ENERGY DISSIPATION IN A VIBRATION DAMPER UTILIZING A VISCOELASTIC SUSPENSION

Joh, P. Henderson, Air Force Materials Laboratory, Wright-Patterson Air Force Base, Ohio

THE EFFECT OF STRUCTURAL DAMPING ON THE MODAL RESPONSE OF

A RESONANT BEAM
Frederick C. Nelson, Dynatech Corporation, Cambridge, Massachusetts

A SIMPLIFIED APPROACH TO STRUCTURAL DAMPING DESIGN

G. E. Warnaka, A. J. Harris and B. W. Campbell, Lord Manufacturing Company,
Erie, Pennsylvania

USE OF THE SHOCK SPECTRUM TECHNIQUE FOR NONLINEAR SYSTEMS

Walter B. Murfin, Sandia Corporation, Albuquerque, New Mexico

NOISE AND VIBRATION CONTROL WITH FIBER METALLURGY

H. B. Karplus, V. J. Raelson, and H. Schwartzbart, TIT Research Institute, Chicago, Ilinois

MEASUREMENT OF COMPLEX SHEAR MODULUS OF VISCOELASTIC MATERIALS

BY MECHANICAL IMPEDANCE METHODS
Andrew J. Roscoe, III, E. V. Thomas, and W. Blasingame, U.S. Navy Marine Engineering
Laboratory, Annapolis, Maryland

iX
xt I:
 TRANSPORTATION

A SURVEY OF SHOCK AND VIBRATION ENVIRONMENTS

IN THE FOUR MAJOR MODES OF TRANSPORTATION*
R. W. Schock
Marshall Space Flight Center, NASA
and
W. E. Paulson
Brown Engineering Company
Huntsville, Alabama

An investigation has been conducted, the purpose of which was a review

of all available information and test data describing shock and vibration
environments in the four major modes of transportation. This infor-
mation has been collated, analyzed, and combined, to provide a unified
reference source of transportation environments for use by the packag-
ing and/or design engineer. A portion of the results of this investiga-
tion are presented in this paper.

INTRODUCTION 3. Many studies have been conducted on

military vehicles operating under the most se-
Over three hundred reports were reviewed vere conditions, which, although setting an up-
and contacts were made with fifty-five agencies per limit, are not representative of normal
and organizations active in the transportation commercial carrier environments.
field. Based on the data collected, acceleration
versus frequency envelopes were constructed 4. Many programs, although providing
for each of the four major transportation modes. meaningful information for the purpose for
These envelopes facilitate the comparison of which they were conducted, have no general,
transportation environments in the different application.

modes. Where applicable, the information and

data are also presented to show the effect of 5. Measurements were made on vehicles
various operating parameters on the transpor- that are obsolete or of limited interest.
tation environment. 6. Measurements were made over unknown
Although many programs have been con- or restricted frequency ranges, or other factors
ducted to collect transportation shock and vi- of importance to general criteria were not re-
bration data, only a portion of the collected data corded.
is applicable for use in a general transportation
criteria. The lack of applicability is summa- 7. The original data existed in a variety of
rized as follows: forms.

1. Few programs dealt solely with cargo In order to correlate the data, a require-
environments, ment of the review was to edit all reports and
include only response measurements of the
2. The data are not segregated and include cargo floor (i.e., input to the cargo). No at-
measurements from locations such as vehicle tempt has been made to define transfer functions
appendages, aircraft wing tips, etc. or cargo response relationships. In addition to

*Conducted by the MRD Division of General American Transportation Corporation under MSFC Con-
tract NAS8-11451.

1I
the constraint that data be included which are lack of significant levels may occur in certain
descriptive only of the cargo floor, unless frequency bands.
otherwise noted, the data have been further re-
stricted to standard commercial vehicles trav-
eling normal routes. Data obtained from (a) Aircraft
military vehicles, (b) cross-country terrain
operations, (c) special road course operations, Extensive shock and vibration measure-
and (d) special transporters, have been omitted ments on aircraft have been performed by the
from the study. Exceptions to the above are Wright Air Development Division (WADD).
(a) data for ships which describe the environ- Their most recent test programs cover the
ment at the aft perpendicular (the area of se- following aircraft:
verest vibrational environment), and (b) results
of tests with the 377PG (Pregnant Guppy Air- 1. (C-123), Medium assault cargo airplane,
craft), a special transporter. high wing, twin engine (reciprocating), three-
bladed propellers.
In the subsequent figures defining the en-
vironments, instrumentation and/or interest 2. (C-130), Medium range cargo airplane,
limits define the frequency ranges from which high wing, four engine (turboprop), three-
the data were taken. A lack of definition in bladed propellers.
some frequency ranges should not be construed
as a lack of environment, but merely a lack of 3. (C-133), Long range cargo airplane,
available data to define that environment, high wing, four engine (turboprop), three-
bladed propellers.

DATA PRESENTATION 4. (H-37), Cargo helicopter, single main

rotor plus torque compensating tail rotor, twin
The data used to develop the graphs sum- engine (reciprocating), five main rotor blades
marizing the shock and vibration environments and four tail rotor blades.
were extracted from numerous reports. The
original data existed in a variety of forms. The measurements were taken during all
Some data were in the form of peak accelera- of the normal service conditions, such as taxi,
tion versus frequency, other data were given in ground run-up, take-off, straight and level
peak acceleration versus duration. For the flight (at various altitudes and speeds, and
latter, unless otherwise specified, a half sinus- power settings), turns, descents, landing, and
oidal shape was assumed and a frequency com- landing roll. The effects of cargo load, speed
puted. brakes, and other control surfaces were also
investigated on some of the aircraft.
An acceleration versus frequency format
was chosen for the graph to utilize as much of These aircraft were instrumented with
the available data as possible. The selection of velocity pickups, and the vibration data were
the format was based on two principal factors: recorded on magnetic tape. Data reduction was
(a) it is the only one which will accept all use- performed with a Davies Automatic Analyzer
ful data, and (b) it permits data originally pre- employing a variable-frequency, narrow-band
sented in terms of power spectral density to be. (10 cps) filter.
transformed to its original form by merely
noting the batdwidth used in the original analy- Vibration data from the Pregnant Guppy
sis. G (rms) versus frequency is used for the (377PG), a low wing, four-engine (reciprocat-
format when the original data were reduced by ing), four-bladed propeller cargo airplane,
a Spectrum Analyzer. G zero-to-peak versus were also reviewed. This aircraft is used in
frequency is used for the format when original transporting space rockets and their allied
data were recorded with oscillographs and vis- equipment. A considerable number of vibration
ually analyzed. measurements have been taken on the cargo
floor during transportation of this special
Vibrations which occur in the cargo area cargo.
have dominant frequencies for particular vehi-
cles, loads, locations, and speeds. Even though The environment in these tests was moni-
conditions and circumstances vary, the graphs tored with accelerometers, and the data were
show that the vibration levels are fairly con- recorded on magnetic tape. A variable fre-
stant over wide frequency ranges. In some in- quency, one-half cycle, bandwidth filter was
stances, particularly with aircraft, a noticeable used in the analysis of the data. Frequency

2
 bands having relatively high vibration levels Railroad
were analyzed further. A distribution of the
accelerations was determined at these frequen- Data descriptive of the ratilroad shock and
cies. Data are available for a number of loca- vibration environment have been categorized
tions, for a number of flight conditions, and for into two major classifications:. (a) over-the-
various loads. These data represent one of the road operation, and (b) coupling. The'over-
1. most complete descriptions of the shock and
vibration environment for any transport vehicle,
the-road environment includes all data except
the shock motions associated with coupling or
humping operations.
Vibration data for the KC-135, a military
version of the Boeing 707 jet aircraft, has been Acceleration versus frequency envelopes
obtained from tests by Boeing. The data cover of the shock and vibration environment of rail-
ground run-up, taxi, take-off and cruise condi- road cars have been compiled from many
tions. The original report presents the vibra-
tion data in power spectral density (g 2/cps
sources. Because of the high amplitude tran-
sient vibrations which occur during starts, stops,
I
versus frequency). The data were conirerted to slack run-outs, and run-ins, these data have
i| g (rms) versus frequency for presentation in been segregated, when specified, from the vi-
this paper. The vibrations were monitored with bration data describing normal running condi-
accelerometers and recorded on magnetic tape. tions. These high amplitude vibrations in Fig. 3
Analyses were performed with a Davies ana- are labeled "transient," whereas the normal
lyzer with the following filters being used for running conditions are labeled "continuous."
different frequency ranges: Both of these plots have been constructed by
enveloping all reliable data for all types of
trucks, rail conditions, directions, and speeds.
Frequency Range Filter Bandwidth Therefore, the transient and continuous plots
(cps) (cps) in Fig. 3 represent the highest vibration levels
which would be encountered during over-the-
0-30 0.80 road operations.
30-50 1.33 The two plots appearing in the lower half
50-100 2.64 of Fig. 3 show the effect that soft ride equip-
ment has on the over-the-road vibration envi-
100-200 5.41 ronment. These plots were formed by envelop-
200-400 10.1 ing
floorthe maximum accelerations recorded on the
of a Minuteman transporter railroad
car
400-800 18.7 during cross-country operation. The data were
reduced by averaging the 4-6 maximum accel-
eration values within each frequency bandwidth.
1000-2000 43.5 The truck suspension system for this missile
car consisted of a combination air and coil
spring system in the vertical direction, and a
Figure 1 is composed of acceleration enve- pendulum system with snubbers in the lateral
lopes depicting the environment for propeller, direction. Damping is provided in both direc-
helicopter, and jet aircraft. These envelopes tions of motion. In the longitudinal direction
were obtained by encompassing the maximum isolation is provided by a s.iding center sill
K vibration levels for the respective classification, and a hydraulic cushioning device.
Data for the C-123, C-130, C-133, and 377PG
were used in developing the plot for propeller The effect of train speed on vibration lev-
aircraft, while the H-37 and KC-135 were used els is shown in Fig. 4 for a number of train
to describe the environment for the helicopter speeds (20, 40, 73, 80 mph). The curves de-
and jet, respectively. The plots show that the picting the environment at 40, 73, and 80 mph
vibration levels are highest for the helicopter, are a result of tests conducted by the U.S. Army
and lowest for the jet. Signal Corps. The data were taken with differ-
ent types of trucks. The vibrations on the cargo
Figure 2 shows the maximum acceleration floor were monitored with barium-titanate ac-
envelopes for several individual aircraft. celerometers and recorded on magnetic tape.
These plots were constructed by encompassing The frequency range of interest in these tests
the highest recorded vibration levels for all was 20-10,000 cps. The recorded data were
flight conditions. The plots are presented in analyzed by passing them through a series of
terms of g (rms) versus frequency. octal band pass filters. The results of the

3
 10.0 g.____*

1.0

-4

0.1 I
'I -Propellerj
(turbt and
.01 - reciprocatng)
-.- Helicopter

-RQUENCY(C PS) p'

0.10 0i 100 1000 100

Fig. 1. Aircraft acceleration envelopes (overall composites)

1.0

IV"
37 PG
.0__-------
VA .......
C-10
C-123
f
H-37
-* KC-135

FREQUENCY (CPS)
.001 1 11l I IIL11 1 11
0.1 1.0 10 100 1000 10000

Fig. 2. Aircraft acceleration envelopes (individual composites)

 10.0

1.0

E-4 _ _ _

0.1

.01

FREQUENCY (CPS)

0.1 1.0 10 100 1000 10000

Fig. 3. Railroad acceleration envelopes (over-the-road)

10.0

80mp

1.0
73ti

FA4
p
0.1

.01

0.1 0 10 _ _ too __ _000 _ 1000

Fig. 4. Railroad acceleration envelopes (over-the-road)

 analysis are presented in terms of average excitation. This method of data presentation
acceleration versus frequency (cps). has been chosen because it provides informa-
tion on all frequency components.
The data presented for 20 mph speed may
be misleading because of the different frequency Data are presented for both standard and '
I range. These data are a result of a separate shock-absorbing draft gears. These shock
program in which frequencies greater than 25 spectra were constructed by enveloping the
cps were not of interest. The data werea ana-
lyzed with a frequency analyzer having con-
peaks of shock 6-14
port. Figures
spectra given in the original re-
show shock spectra for
F
tinuous band-scan device, with a frequency standard draft gear at coupling speeds of 3.4,
resolution less than one-half cycle. These data 6.0, 8.0, and 10 mph. The data used in Figs.
were also presented in terms of average accel- 6-14 were obtained from tests conducted by
eration versus frequency. Sandia Corporation. The tests were conducted
by impacting a lightly loaded test car with a
Figure 5 consists of plots that show the hammer car.
effect of orientation (direction). These plots
are restricted tc include only the data descrip- Figures 15-18 present additional shock
tive of the continuous vibration environment spectra for a Miner draft gear at various im-
(i.e., data for transient vibration have been pact speeds. The Miner draft gear is a high
omitted). The plot shows that the vertical is capacity shock absorbing draft gear.
the most severe environment over wide fre-
quency ranges.
Railroad Distribution of Coupling
Speeds
Railroad Coupling
Figure 19 shows the distribution of cou-
Curves describing the shock and vibration pling speeds based on 3369 measured impacts.
environment for railroad coupling and/or hump- This plot has been constructed by averaging
ing operations are presented as shock spectra. the results of a number of independent investi-
A shock spectrum is defined as the maximum gations. Data are presented in terms of per-
acceleration of a series of single degree-of- cent of total impacts versus coupling speed
freedom systems resulting from a specific shock (mph).

10.0

II

..
\ .......-
~ ~~~~................o .
0 .-

I -
.01

Vertical
Lateral
........ Fore/Aft.
FREQUENCY (CPS)
.001 1 111.IIIlia I I 1111111 I I I11 ll I III l I I I11l1
0. 1 1.0 10 iO0 1000 10000
Fig. 5. Railroad acceleration envelopes
(directional composites)

- ..-

,r
 ! " 1000-

100

F.,0
T I 100

1.0

-....... 0.= .
%%Mec
o0 C/C

- 5 % C/C e ,
" F R EQ E NC Y %CPS )

0 .1 t.0 to 100 1000 10000

I Fig. 6. Railroad coupling shock spectrum (3.4 mph,

, fore/aft) standard draft gear
.0*

s 1000
ss 'It

100 .

i' z 00/0 '-"o=

-....... 0.,5% C/Co

5%C/Cc

0.1 - .
"
. IIIII
.I
FREQUENCY (CPS)
II 'l I a 11111( I I 1 11111111
I
0. 1.0 10 100 1000 " 100

Fig. 7. Railroad coupling shock spectrum (3.4 mph,

vertical) standard draft gear

i7
 1000 1
100~

100

1.01 10.0

0//C

..... 0. 5% c/c
- C
FREQUENCY (CPS)
0. 1 1.1 1111U1 11 11 1 1
0.1 1.0 10 100 1000 10000

Fig. 8. Railroad coupling shock spectrum (3.4 mph,

lateral) standard draft gear

1000 1 v

100.4 -
z . -"

E 10.

QU

FREQUENCY (CPS)
..........05 c/c
[i
0.1 1. to 10 100 I l l 10000IIIII

Fig. 9. Railroad coupling shock spectrum (6.0 mph,P

fore/aft) standard draft gear

8t
 1000
1

.a 100 . y"
/ ,

- - II

- J-

10

1.01
/c 0
"......0 % CC
5 0C/Cc

FREQUENCY (CPS)
0.1 - .i 1JI JJ
1IIIII I I mlill I I 111111 1 1I I III
0.1 1.0 10 000
1oo 10000

Fig. 10. Railroad coupling shock spectrum (6.0 mph,

vertical) standard draft gear

1000

100
. _..

"O -0 -
z
- o

.,~ 1.0
_-. 0%C/CC
. o% C/c c

FREQUENCY:(CPS)
I IIIIIIl I I 111ii1l I 1 l111ll 1 111 i I
0.1 II 1 I
0.1 1.0 10 1o0 1000 10000

Fig. 11. Railroad coupling shock spectrum (8.0 mph,

vertical) standard draft gear

- .-.-
1000

i~o . . 0.5. C/Cc

o/ C/Ce

-- 5 0/ C/Cc
-J
-FREQUENCY MCPS)
-.Ii IiI III
g Iul I I 111111L I
0.1 i....L.LJ
0.1 1. 0 10 100 1000 10000

Fig. 12. Railroad coupling shock spectrum (10.0 mph,

vertical) standard draft gear

- __ 0 %C/Cc
1000- ..-.

00

S10 __

.4 1.0

FREQUENCY (CPS)
.E$w
...... 0.5%c/cc

------- 5 % C/Ce

O.1 .0 10 1O0 1000 10000

Fig. 13. Railroad coupling shock spectrum (10.0 mph,

fore/aft) standa-rd draft gear
rL1 -

10
II
1000

Vr
I/,
100
Lz
I to

1.0
- 0C/Cc
...... 0.5% C/Cc
5 %C/Cc
FREQUENCY (CPS)
0.1 l i
0.1 1.0 10 100 1000 10000

Fig. 14. Railroad coupling shock spectrum (10.0 mph,

lateral) standard draft gear

1 "1000

--

100

4z,. --
'4 /

1.c
0 C/%
...... 0.5 % C/c c
15...:5 % c/c c
FREQUENCY (CPS)
I I Illll
too I
I I II III
111111to 1
0. 0 .1 1 1 1 II MII
1. 0 IN1 000 10000

Fig. 15. Railroad coupling shock spectrum (3.7 mph,

fore/aft) cushioned draft gear

11
 ItI

100

1000
0.!
II
4z

lo
10

10

0. CC
.... o5 %c/cC
- -- 5 % C/CC
FREQUENCY (CPS)
0. 1 I I l -ll I- - L L LI I a l l l l l aII 2 t l
0.1 1.0 10 100 1000 10000

Fig. 16. Railroad coupling shock spectrum (6.8 mph,

fore/aft) cushioned draft gear

1000 "

(p U
100

E-C 10__ _ _

1.0
00 C/C

FREQUENCY (CPS)

0.1 1.0 10 100 1000 10000

Fig. 17. Railroad coupling shock spectrum (9.8 mph,

fore/aft) cushioned draft gear i
= -.--- o C/-

- - 5 c/c(/

12
 4 _____________________

' 0

Ii 1000

100

J9

0
.1.0
0 % C/Cc
...... 0. 5 % c/cc
5 % c/ce
4 0.11 1I1 FREQUENCY
1
ll (CPS)-1
I1lll I1 I11ll 11111 1 Illlll 1 1 I 111

0.1 1.0 10 100 1000 10000

Fig. 18. Railroad coupling shock spectrum (12.0 mph,

fore/aft) cushioned draft gear

100

80

0 60

z
40 _ _ _ _ _ __ _ _

H 6

20

0 _

2 4 6 8 10 12
MPH
COUPLING SPEED

Fig. 19. Railroad distribution of coupling speeds

13
Ship envelope includes data recorded under extremely
rough sea conditions.
The data used in plotting the overall corn-
posite curves
intoconinuosfor
adships have been
tansint separated
ibrtion. Tan-The third and lowest plot in Fig. 20 is con-
into continuous and transient vibrations. Tran- structed from data taken on a 572-ft, single
sient vibrations curare defined
urin as thosendwhich
mneuvrs
emrgeny oc-
cur during emergency maneuvers and sam-screw
slam- velocityship. Measurements
pick-ups mounted atwere takenthrust
the main with
.
ruing (the impacting of the ship with water after veoi one tth antrs -ik
mig(the
owmhas l tigthe shp
water).
T hp aft o bearing foundation and to an angle welded to the
the bow has left the water). This separation of deck over the main transverse member at the
data is made because transient vibrations can aft perpendicular. The data were recorded on
usually be eliminated from the environment. straight runs and maneuvers during operations
Slamming, for example, can be avoided if the in calm seas, and at various propeller speeds.
ship avoids storm areas. Continuous vibrations
are defined as those which occur during normal
operations, including operations in rough seas. Figure 21 shows the effect of orientation
(i.e., the effect of direction). The data used to
Most of the data for the shipboard shock construct this figure were obtained from the
and vibration environment were collected by tests conducted on the 572-ft single screw ship
the David Taylor Model Basin. The major por- mentioned above. The curves show that the
tion of their data, however, have been recorded vertical vibration environment is the highest,
near the aft perpendicular (a line perpendicular followed by the lateral and fore-and-aft direc-
to the water line at the stern). This area ex- tions.
periences the highest vibrational levels on a
ship. The levels of vibration for the cargo Figures 22 and 23 show the effect of sea
area will always be lower. However, since state for two different ship lengths (L = 820 ft
meager information is available for the cargo and L = 380 ft). The curves show that the ac-
area, data for the stern location can serve as celeration levels increase with increasing fre-
an upper bound of the environment. quency from 4 to 10 cps, and are constant at
higher frequencies. The accelerations for the
The upper two curves in Fig.. 20 have been small ship (L = 380 ft) are almost twice as
constructed by enveloping all data applying to large as those for the larger ship (L = 820 ft).
continuous vibration on one diagram, and all For each ship class, the acceleration increases
data referring to slamming and emergency by a factor of two, when the ship operates in a
on another. The continuous vibration, composite rough rather than in a smooth sea.

10.0 Sla& &

Emergency

1.0

z Vibration - Rough
0. Seas
]J
.1
NormalManeuers
Calm Seas

.01

FREQUENCY (CPS)
.00f | I 11i ll
11 ItII I i ii|I I a li 1 111 I I I 1il111

0.1 1.0 to 100 1000 10000

Fig. 20. Ship acceleration envelopes

14

Ag
 .....
.
...................

10.0

1.0

z
2 0.1

.01 I
Vertical
I.! --- Lateral
... Fore/Aft
FREQUENCY (CPS)
.001 LULJ
0.1 1. 0 10 100 1000 10000

Fig. 21. Ship acceleration envelopes (straight runs)

10.0 I.
Definition of Sea States

i.L: -L 3
w 3 s w 1.5 s
1. 0 H : 720Ls H yoL
Slmi Sea State w

.00.1 H s ~L H s34 Ft.

14w 80 s w

0
L
w - Length of Wave
.01 H - Height of Wave
Aw
L s- Length of Shiip

FREQUENCY (CPS)

01 100toto 1000 10000

Fig. 22. Ship acceleration envelopes (L. 820 ft)

15
'4 10.0

Definition of Sea States

Slau
iate i. L -LL 3. L -L
-
Sea Sw-3 s w 1.5 s
4 H 4
S1.0
0 w
3 w 120 s 30 s

z 2. L 2 L 4. Lw:s L

0.1
_H
w
ToL s H
w
s34 Ft.

e L - Length of Wave

.01 H - Height of Wave

Ls Length of Ship

FREQUENCY (C PS) 11
0.1 1. 0 10 100 1000 10000

Fig. 23. Ship acceleration envelopes (L s = 380 ft)

The higher vibration frequencies (> 10 cps) vibration environment occurring during opera-
are due to machinery vibration, and are less a tion on paved roads.
function of sea state than the lower frequency
rigid body motions. Figure 25 shows the effects of cargo load I
on the vibration environment. These curves
are the result of a single measurement pro-
Truck gram. The tests were conducted with three
standard commercial semitrailers, each having
Figure 24 shows the effect of road condition one of three basic types of suspension (air-
on the vibration environment, ride tandem suspension, stable-ride single
suspension and single-axle spring suspension).
The upper curve in Fig. 24 has been ob- Tests were run at two load conditions, empty
tained by enveloping data from a number of in- and full, over a first-class asphalt road. Ver-
dividual test programs. This curve includes tical accelerations were monitored at three
peak values representing the environment ex- locations on the cargo floor (over the fifth
perienced in traversing rough roads, ditches, wheel, the center of the van floor, and over the
potholes, railroad crossings, and bridges. Data rear axle). These curves show that the vibra-
reduction procedures vary from one report to tion levels are practically unaffected by load in
another, but in most cases the method used was the lower frequency ranges. Higher frequency
to record the data oscillographically, and visu- components, however, are reduced on loaded
ally determine the peak (zero-to-peak) acceler- trucks.
ation and predominant frequency. This method
has been used extensively in transportation
studies, since it requires little auxiliary equip- CONCLUSIONS AND
ment, and since the magnitude of the significant RECOMMENDATIONS
predominant frequencies can be conveniently
and immediately determined. The most severe transient environment
associated with transport vehicles occurs dur-
The lower curve in Fig. 24 has been ob- ing railroad' car coupling or humping operations,
tained by enveloping paved road data. The whereas the most severe steady-state environ-
combination of these two curves show the dif- ment occurs during phases of aircraft trans-
ferences in vibration levels between vibrations portation. Railroad, truck, and ship follow
which occur while traversing potholes, ditches, aircraft in order of decreasing levels of
railroad crossings, etc., and the maximum environment.

16
 10.0

1.0

zA 0%-
0.1 ..

.01
Rougb Road
.
---- Paved Roads

FREQUENCY (CPS)
.001 I I I II miI a IIIIii I u ai
I I I I III
0.1 1.0 10 100 1000 10000

Fig. 24. Truck acceleration envelopes

10.0_____

1.0
4 C,,

z 1
Z -
-4 0.1

.01

....... Empty

FREQUENCY (CPS)
. 001L 'lI II IIIII I 1 1111l1l 1 1 I1IIIIII I I IIIIII I I I II
10
0.1 .0 T i00 1000 10000
Fig. 25. Truck acceleration envelopes

17
 The levels of vibration presented in this Future studies of interest that would greatly
paper are considered reliable, even though they enhance and supplement this investigation could
are derived from data analyzed by diverse re- include (a) correlation of fantail to cargo hold
duction methods. This conclusion is based on shipboard environments, (b) a more extensive
the relatively close correlation between data breakdown of transportation modes, (c) a sta-
from various reports (for similar test vehicles tistical processing of accumulated data, (d) a
and conditions). continuing effort of environmental definition,
particularly in the truck mode, and (e) an in-
Data are sufficient at the present time to vestigation of terminal handling environments.
define adequately the transportation environment
in terms of maximum acceleration versus fre- Due to the increased n~tionwide interest in
quency; however, the term maximum should be transportation environments, many programs
emphasized. For example, because of the pau- are currently being conducted in the previously
city of data describing the environment in the outlined areas, and should go far to supplement
cargo area of ships, data from the extensively the basic criteria contained herein.
monitored fantail area were utilized. Since
this is the highest encountered shipboard envi-
ronment. it does not represent the cargo hold, ACKNOWLEDGMENT
but does set an upper bound on the expected
cargo hold environment. The philosophy of en- The assistance and contributions of the
veloping all data lends itself to a conservative survey's principal investigator, Mr. Fred E.
environmental definition with a very high (99 Ostrem of General American Transportation
percent) statistical confidence level. Corporation, are acknowledged and appreciated.

DISCUSSION

Mr. Markson (E.R.A. Inc.): The set of Mr. Schock: Yes, over 300 reports were
ship acceleration curves appears to be a corn- reviewed.
bination of vibration accelerations and shock
accelerations. Is that correct? Mr. Fitzgibbon (Mechanics Research):
Why did you not present your vibration data in
Mr. Schock: Yes, the. impact vibrations terms of the power spectral densities rather
are transient in nature although they are of than g's rms ? In this manner you could show
rather long duration. It could be termed a an envelope of all of the power spectral densi-
shock vibration. It is a response, but over a ties from the various environments which would
longer duration than one would normally con- be a criterion for all environments.
sider a shock response.
Mr. Schock: We had spectral analyses in
Mr. Markson: What kind of instruments some cases during the course of the study,
did you use to pick up the two different types of while others presented only data from peak
accelerations ? detection meters or oscillograph records.
Most of the data on aircraft were in spectral
Mr. Schock: We used accelerometers, analyses and given in psd. We reduced these
mounted in the fantail area. to g-rms to get them in a form comparable
with the other data.
Mr. Markson: Did you use the same ac-
celerometers for reading vibration at 100 cps Mr. Fitzgibbon: Are you suggesting that
on the ship as you did for reading the low fre- the rms g's be taken from your curves and
quency shock impact? converted to psd to arrive at specifications for
shock and design?
Mr. Schock: No, there were two different
studies. One was for vibration, and the other Mr. Schock: No, I would not use these
was specifically to pick up the slam modes. curves in specifications. The paper itself
Different accelerometers were used in each breaks these curves down in much more detail
study. with more detailed parameters, such as the
direction of measurements, and individual air-
Mr. Markson: This is a summary of sev- craft measurements. This will greatly facili-
eral different reports ? tate the definition of a better specification.

18
 i-1 !

I would not try to envelope curves such as Mr. Krachman (TRW Systems): That truck
these for use in a specification. It gives the data seems to be a little low. Was this on any
designer a feel for what the maximum levels special type of truck?
will be and what to design for.
Mr. Schock: The maximum environment N
Voice: Could you classify the helicopter was on trucks that had standard truck gear.
and the propeller excitation as being one of the Some of them did have air-ride and some had
most severely characterized as steady state? only coil springs and single axles. The data in
the lower environment were taken over very
good first class asphalt roads, whereas the data
Mr. Schock: Yes, the helicopter data was in the upper environments were taken on a typ-
characterized as steady stte. It is caused by ical rough road which had pot holes, railway
the response of the structure to the steady state crossings, and bumps. It was not as rough as
prop wash. the Munson course.

, "
' 191 i1
i 19
 THE DYNAMIC ENVIRONMENT OF SPACECRAFT
"SURFACE TRANSPORTATION*
J. W. Schlue
Jet Propulsion Laboratory
Pasadena, California

The data presented in this report include shock and vibration measure-
ments over rough irregular roads, plus vibration measurements over
smooth highways. Comparisons are made between different vans,
smooth and rough roads, and three different locations within the same
van. Data plots envelop the maximum, 95 and 50 percentile levels of
shock spectra and power spectral densities.
A discussion of instrumentation applicable to measurements of trans-
portation dynamic environments is contained in the appendix. Electri-
cal and mechanical instrumentation are included.

INTRODUCTION spacecraft shipments to Cape Kennedy. The

evaluation test is referred to as the Ranger
The shock and vibration environment im- Block III test, since it was performed in sup-
posed on spacecraft during surface transporta- port of Block III configuration Ranger ship-
tion was determined by the Jet Propulsion Lab- ments. The trips to the Cape involved the
oratory in evaluation tests and instrumented Ranger 8 spacecraft that recently ph6tographed
spacecraft shipments. All flight spacecraft the moon, and a structural model of the Sur-
built by JPL have been transported by air sus- veyor spacecraft.
pension vans to the launch site at Cape Kennedy. The load distribution of the evaluation test
The decision to instrument spacecraft ship- included the Ranger TV tower at the aft-posi-
ments was made primarily for the following tion, the Ranger bus structure and electronics
reasons: at the middle, and a solar panel trailer at the
front. Steel plates were located in the middle,
1. Limited information was available on the also, as ballast. The ballast was intended to
dynamic environment of surface transportation. load the van suspension system to its nominal

2. It is desirable to have shock and vibra- loading for improved dynamic characteristics.
tion
On measurements in the event of an accident.
esnormal environment The Ranger 8 van was loaded in approximately
is determined, the same manner except that no ballast was in-
Once the omal enviroyint is frminu- cluded. A structural model constituted the pay-
this becomes the underlying reason for continu- load of the Surveyor shipment.
ance of instrumentation.

It is hoped the statistical study presented The Ranger equipment was installed in
in this paper will aid in predicting transporta- transportation structures secured to the van
tion environments. floor with no shock or vibration isolators.
Structural integrity between the van floor and
the spacecraft equipment was assumed, so that
DESCRIPTION OF TRANS- measurements at the van floor were considered
PORTATION STUDIES the input to the equipment itself. The Surveyor
model, however, was installed on an isolated
The data measurements for this study were trailer that did not linearly transmit the vibra-
made during a van evaluation test and two tion and shock of the van floor to the spacecraft.

*This paper presents the results of one phase of research carried out at the Jet Propulsion Labora-
tory, California Institute of Technology, under Contract No. NAS7-100, sponsored by the National
Aeronautics and Space Administration.

21

IRT 7 I
Consequently, the spacecraft structure and produced. The reference channel recorded only
transport trailer were instrumented in depth to the standard FM carrier frequency. The de-
determine the transfer functions and resonant modulated signal from the channel was recorded
frequencies of the system. Only one lata chan- on the oscillograph records. (This channel is
nel was determined applicable to this report. discussed in the instrumentation section.)
The Ranger vertical axis measurements used
piezoelectric accelerometers located midway 3. Data were selected from the "0" graphs
between the sides of the van, in the locations for power spectral density (PSD) and shock
depicted in Fig. 1. Two accelerometers were 4 spectrum analysis. These data selections were
also installed in the longitudinal and horizontal identified on the tapes with the time code. The
axes. quality of data selected was determined by the
reference channels.
The Surveyor measurements, were made
only at the aft position. The accelerometers 4. Preliminary PSD and shock spectra
were mounted on aluminum plates secured to were evaluated for required resolution and fre-
the van floor with lag bolts. quency content.

Mechanical accelerometers were also em- 5. The final PSD and shock spectrum anal-
ployed. These units were also mounted on ysis plots were programmed and computed.
metal blocks secured to the floor.
6. Selected PSD's were analyzed for am-
plitude statistics. (The method of analysis was
DISCUSSION OF DATA determined by evidence of noise content in the
data.) The existence of apparent noise dictated
The data to be discussed were measured a statistical analysis to estimate the energy
with piezoelectric accelerometers and recorded distribution. 'A
with a magnetic tape recorder. The analyses
were performed with digital computer programs.
Shock Analysis
During the trips to Cape Kennedy from
Pasadena, California, about 85 percent of travel The analyses of all shocks were performed
time involved smooth highways in good repair. in the sequence listed above. Figure 2 contains
However,the greatest percentage of vibration an amplitude-time history of a transient meas-
and shock analyses was with rough road meas- ured at the aft position of the van during the
urements. The data analyses were performed RA-8 shipment. The transient was located on
sequentially as follows: the tape via the time code, digitized, and run
through the shock analysis program, producing
1. The magnetic tapes were coded with the Figs. 2 and 3. Figure 3 contains the shock
NASA 36 bit time code. spectra calculated with an assumed damping
ratio of 0.050 or a Q of 10. The spectrum rep-
2. Oscillograph records of the data chan- resents the response of a single degree of free-
nels, reference channel, and time code were dom oscillator to the shock measured. The

FORWARD MIDDLE REAR

VAN FLOOR

Fig. 1. Accelerometer locations

22

.. (
 3- --- -4--+-

- 3.- _ Tip :.,

_-
0 .

-
&0 .4

IW

-0.8

1 .00 0.14 0.28 0.42 0.56 0.70

K. TIME sec
H Fig. 2. Observed transient RA8 aft position

1)assumed Q of 10 represents the Q of the hypo- Three shocks measured by each vertical
thetical oscillator. The theoretical- steady state axis accelerometer in the Ranger 8 van were
sine equivalent of the shock may be calculated selected as data to compare the response of the
simply by dividing the spectrum by the value of three channels to a given excitation. The max-
Q. Thb data were low pass filtered at 200 cps. imum envelope of the three shock spectra is
Although not well defined, the spacecraft fragil- plotted for each of the three locations in Fig. 4.
ity at low frequencies was of greatest concern. The aft position response contains a peak at 18
cps, a notch at 25 cps, and a second major peak
Although low pass filtered at 200 cps, the at 75 cps. The center and forward position re-
analyzed data are plotted to 100 cps only. It sponses peak at about 14 cps and decrease in
was determined that little energy existed above level to a minimum at 25 cps, similar to the aft
100 cps. The reader should bear in mind that response. Below 40 cps the van center indi-
at frequencies below 5 cps the frequency re- cates the lowest level response. Above 40 cps,
sponse rolls off somewhat. The general trend however, the center position becomes more se-
of the data below 5 cps is probably valid, vere than the forward section and approaches
however. the amplitude of the aft response at 60 and 88
cps. It is apparent from the figure that the en-
The observed transient plot is useful as a vironment at all van locations produces a simi-
cursory check of the shock amplitude envelope, lar response below 40 cps, but is dominated by
but does not readily associate energy levels local dynamic characteristics above this fre-
with frequencies. The shock spectrum plots quency. The g response levels are quite low,
must be referred to for frequency information, with a maximum of 3.8 g pk. The equivalent
and will, therefore, be the basis of discussion sine input at this frequency would be 0.38 g pk.
in this report. A Q of 10 will be assumed to
limit the quantity of data plots. Damping ratios Figure 5 compares the maximum shock
of 0 and 0.025 were also used to compute shock spectra of 9 Ranger shocks, measured at the aft
spectra, but are not discussed in this report. section, and 12 Surveyor shocks. The spread in

23
I .-
 -.. ,... ut . .......
It

-H- ----

044

_,Ci. ,! i

[AII
Ciii
U)
z

, --

o AAS PE

I
.1 ,-51

Fig. 3. Shock spectra RA8 aft position

response levels varies with frequency, although The maximum peak amplitude of response
some consistency in the responses is apparent. is 6 g pk at 95 cps. The maximum amplitude at
An important similarity is the peak response at 18 cps is 4 g pk. The 95 percentile and maxi-
18 cps in spite of the diffeiences in vans, load mum spectra have generally the same shape,
weights, and distribution. This peak may result with a maximum dispersion of about 1.5 g pk.
from phenomena such as wheel bounce. The The 50 percentile or average spectrum plot is
Surveyor response is generally between 2 and 4 also similar in shape, and often 2 g less than
g pk, whereas the RA-8 response is primarily the maximum curve at various frequencies.
between 2 and 3 g pk. In both cases the energy This indicates that the spread or variation in
levels of the response as a function of frequency shock levels is rather small. It is apparent
generally increase above 25 cps. from the spectra that shock-mounted equipment
should have a system resonant frequency well
below 18 cps.
Perhaps the most significant data yet pre-
sented arecontained in Fig. 6. The shock spec- The spectra presented in Fig. 6 could be
tra plotted in this figure present the maximum, used to establish transportation tests for equip-
95 and 50 percentile levels of a data sample of ment to be transported by air suspension van.
21 shock spectra, including 9 from Ranger 8, The relatively large data sample provides con-
and 12 from Surveyor. The spectra were com- fidence that the spectra envelop typical road
puted with an assumed normal distribution. The shocks. Equipment degradation resulting from
21 spectra were all van aft position data. As in shock loading should be insignificant. The max-
previous plots, peak amplitudes are very prom- imum equivalent sinusqidal input would be 0.6 g
inent at 18 cps, indicating a common van char- pk at 95 cps, if derived from maximum spectra
acteristic based on a significant sample of data. of Fig. 6, or 0.4 g pk at 18 cps.

24
 0

4e" :::-==CENTER:-

II 0

to

"II

0 20 40 100

FREQUENCY cps

Fig. 4. Maximum s~iock spectra RAB aft, center, forward positions

Vibration Data Analysis The maximum, 95 and 50 -percentile level

The esuts ata nalsisare
f viraton spectral densities of Surveyor shipment vibra-
The esuts
f viraton ata nalsisare tion data are shown in Fig. 71for a normal dis-
contained in power spectral density (PSD) plots. tribution. The data measurements were made
[The data selected for PSD analyses were low during rough road travel. The maximum and 95
pass filtered at 180 cps. The accuracy of data percentile curves are very consistent and show
below 5 cps is somewhat impared by frequency little difference in amplitude. Below 9 cps, the
response roll off. Each plot contains the spec- average energy levels deviate from the maxi-
tral density in db with a reference of 1 g2/cps. mum by 3 to 4 db. This difference decreases to
The frequency scales start at 1 cps, and often
.. . . .
about 2 db above 9 cps. An energy peak at 18
include frequencies beyond 180 cps. Data be- cps appears in the PSD's as it did in the shock
yond 100 cps are usually of little significance. spectra. Howev~r, a lower frequency peak at
12 to 13 cps is also contained in the PSD. All
PSD's were computed from data measured data samples used in this statistical anilysis
in the Ranger Block III test, Ranger 8 shipment, display the two peaks prominently, indicating a
and Surveyor model shipment. Both rough and characteristic response of the van to road vibra-

'Iurements
smooth road data are presented. Data meas-
during the Ranger Block III test are
from rough or detour road surfaces only. The
tion. The spectrum shapes between 1 and 5 cps
may be evidence of a van resonance in contrast
to assumed wheel bounce excitation between 10
vibration data represent the same accelerome- and 20 cps. The explanation for data charac-
ter locations as the shock spectra. teristics cannot be made definitely at this time.

25

MWE
 .1

"I I "

o I,
044
0*

0 20 40 60 80 100

FREQUENCY cps
Fig. 5. Maximum shock spectra RA8 and Surveyor aft positions i.

Figure 8 compares the maximum PSD of locations in the Ranger Block III van. Each en-
rough road Surveyor data with smooth road Sur- velope was derived from a total of seven vibra-
veyor data. Both plots contain the characteris- tion data amples. As previously noted in the
tic peak at 18 cps and are very similar in spec- shock spectra, the measurements indicate both
tral shape, with the exception of the frequency
band 6 to 15 cps. In this region more peaks are
consistencies and inconsistencies between chan-
nels. Measurements at the van rear and middle
I4
displayed by the smooth road data. Amplitude positions contain a peak at 2 cps. The corre- (
differences between the plots substantiate a sig- sponding peak at the forward position occurs at
nificant difference in road conditions.,*Energy 4 cps with an amplitude of -21 db. These peaks
differences of 4 to 5 db are common and often are probably van resonance frequencies. The ~
far exceed this. It is interesting to note that second peak occurs at approximately 12 cps at
!-I 4 -,
the greatest margin occurs at 12 cps. Low all three locations, and may result from wheel
level excitation does not seem to excite this bounce FREQUECY cp spectral den-
excitation. At 12 cps the
frequency, perhaps indicating a nonlinearity in sity amplitudes are close to -20 db at both the
the van. rear and middle positions. The level is greater
at the front, however, and reaches a maximum
Although not shown in the figures, the of -15 db. Beyond 12 cps the rear and middle
spread between the maximum density plot and measurements roll off in amplitude, whereas
the 50 percentile remained quite consistent be- the forward measurements indicate a peak at
tween rough and smooth road data. 35 cps.
Figure 9 contains a composite of the maxi- It is apparent tnat the environment of the
mum PSD envelopes at the three accelerometer forward position was most severe during the

26
0 0460
 -:10
.A
0
40

5 95

0=5

0
110 20 40 60I 0

4J

-60 ZELM679 34689

Frequency cps-30

Fig. 7. Suvyrsetrldniyaximum,
lvlsokpetaR
95%, 50%,b ndureog rf odton

227

->40
 -20.*

o :Rough Roaid:.

m Z Fig. 8. Surveyor
-40. maximum PSD en-
to velope -rough vs

z1 . Smooth Road smooth road

2 3 4 567891 2 3 4 5 67891 Z 3

FREQUENCY cps (1-300)j

-10

FORWARD#
N I If
Fig. 9. Test -RA
aft, center, forward CEN-40
position - maximum z
envelope PSD i

J 1

-70 1
2 3 4 5 6 7891 2 3 4 56 7891 2 3

FREOUENCY cps (1-300)

28

"'Pi
 RA Block I tests. At the 3-db down points of The spread in amplitudes is of little signif-
the forward position 12 cps peak, the spectral icance, since the road conditions were different.
density value is -18 db, and is about 3 cps in
bandwidth, producing a level of 0.22 g rms. A plot of the maximum, 95 and 50 percent
PSD levels computed from all PSD analyses is
Figure 10 indicates a different comparison contained in Fig. 12. Normal distribution was
between the locations for RA-8. The aft spec- assumed. Thirty-seven data samples are in-
tral density has the greatest envelope with a cluded in this statistical analysis. It is appar-
few exceptional frequencies. The forward posi- ent in the figure that a considerable difference
tion is less severe below 2 cps, but exceeds the in amplitude exists between the curves because
center measurements at higher frequencies. of the varied surface conditions represented.
The comparison of amplitudes for the three lo- The RA Block III measurements were made
cations varies with frequency, so that a maxi- over a very rough detour road, whereas RA-8
mum level position may become the minimum and Surveyor measurements are from highway
level position at a different frequency. driving. These energy levels, therefore, cover
severe, rough highway and normal surface con-
Figure 11 compares the maximum spectral ditions and should envelop the vibration envi-
densities of the aft position of the three van ronment of any eventuality in a cross-country
measurements. The important comparison to trip (barring accidents).' Again the peak fre-
be made from this figure is the relative shape quencies of 12 and 18 cps are prominent. An
of plots. RA-8 and Surveyor vans indicate an energy amplification is also indicated at 2 cps.
energy peak at 18 cps. Ranger Block III data do
not show a peak in amplitudes at 18 cps, but do In general, the vibration data presented in-
peak at 12 cps, similar to the first peak of the dicate low energy levels at all frequencies
Surveyor data. Both these peak frequencies within the 1 to 180-cps band of analysis. Ap-
should be considered if shock mounts are used. proximately 56 PSD's were included in the data

P. -30. ___ T
....... ..

I H..... "O" V:: 7,

. ...

-40.1

U -3o. A

04ENE
Z ] .. ... .............. "-"

FREQUENCY cps (1-300)

Fig. 10. Test - RA8 power spectral density - smooth road - maximum
envelope aft, center, forward positions

29

-"7
 -I Ij
.10 -

- L.
.. ~
:i
" 'r

I HH I iii
RA -BLOCK-oIIIt
-ti'

.b~~
* :
l"
,t

-20' a , I
U ....

30

.0 SURVEYOR 1
.' -
S-40--
-4
44
-A-8....... .........

-50

-60~=" . ... .. ...

1 2 3 4 56 7991 Z 3 4 5 6 7891 2 4
Frequency cps (1 -300) r

Fig. 11. Maximum PSD levels - RA Block HI, RA8, Surveyor aft position

presented. The maximum amplitude of vibra- following conclusions can be drawn from the
tion
The contained
attenuationinofthe plots
this is -19 db
amplitude at 12 cps.
depends on the data presented,
discussion in theand from the
appendix (theinstrumentation
appendix on in-
frequency, but is usually significant. It should strumentation was warranted by significant
be noted here that Fig. 9 contains a -15-db peak problems in this area):
at 12 cps. This peak energy was measured at
the forward position of the van, however, and is 1. The shock and vibration environment of
not contained in the plots of Fig. 12, which cover an air suspension van should present no degra-
only the center and aft positions. Figure 12 dation problems if safe driving regulations are
does envelop, therefore, the vibration energy at imposed on the truck operator and no accidents
the recommended van locations for installing occur. Speed limits and periodic truck inspec-
payload equipment. tions should be imposed regulations.

No discussion of the longitudinal and hori- 2. Equipment should be hard-mounted to

zontal axis data is included in the report. The the floor of the van if the dynamic characteris-
measurements in these axes were as much as 40 tics of a support structure or shock mount have
percent less than the vertical axis measurements. not been accurately determined. The mount or
structure dynamics may be coincidentally
"tuned" to the input and amplify the response.
CONCLUSIONS The amplitudes measured at the van floor are
rarely greater than .1g pk. At low frequencies
Based on JPL's experience with the meas- the center of the van floor, lengthwise and width,
urement and analysis of transportation data, the, has the less severe environment.

30

7,
 -10

-2 0
q-ii'H'"'
i-
j .i'
"L. .. ; ... ......:
:......
.: " ":p .....' ...
... -. -- - 1.

ri U
-zo ]
. !:ff
|
. . I MA.::H-'
..- . +.,
'
.. -. r-1

95% ---
---
----. I : ..
3 NN
\
.-.. ,I
40"
.-.
-30

--
Ik
-' LL..
-- I
506..4 -

40)
IL.
4b

-50

-6o 1 - -
2 3 4 56 7891 2 3 4 5 678 91 2 3
Frequency c's (1-300)
Fig. 12. Maximum, 95%, 50% PSD levels - RA Block III, RA8, Surveyor
aft and center positions

3. The position of least vibration or shock tape transport definitely should be a closed loop
severity in a van is dependent on the frequency, design (as discussed in the appendix).
the individual van characteristics, and to some
extent the amplitude of excitation. The aft and 6. A method should be developed to code
center positions seem most consistent, although the magnetic tapes if a tape recorder is used.
their amplitudes often exceed those in the for- Time coding is required for identification
ward position. For this reason, it may be de- computer control for shock or PSD analysis,
sirable to design shock mounts for the aft or Cop te nt ro onitis,
center positions if mounts are required. and correlates events with road conditions.
center and aft positions are most predictable.
7. A reference channel should be included
4. When selecting the mode of instrumen- with data channels as a means for cursory eval-
tation for a van shipment, forethought should be uation of data quality and for identification of
given to the purpose of the instrumentation, microphonics.
for data acquisition
The underlying reason
should influence the sophistication of the in-
strumentation,
l " the location of transducers, the measurements airshock
'aboard an for
*8. The requirement and vibration
suspension van is
mechanical recorders, the Sampling of data re- suremrnsoran supoducenlit
quired, and the final data analysisr important primarily to record the loading that
may -occur in the event of an accident. Normal
5.iMagnetic tape rcdesholbese- sufc.transportation should produce little
Slected with caution for this application. The concern.

31
7
1%
 Appendix

APPRAISAL OF INSTRUMENTATION TECHNIQUES

The effort involved in acquiring transpor- used. Wow and flutter were problems in spite
tatlon vibration data has resuled in consider- of an attempt to shock isolate the recorder. In
able revision of instrumentation techniques. vibration tests conducted prior to shipment, the "1
Since the final data quality is dependent on the shock mount system demonstrated a resonant
instrumentation, a brief appraisal of instru- frequency of 4 cps, which was acceptable, as-
mentation techniques is warranted. suming the dominant van excitation would be
from wheel bounce at approximately 12 cps.
The attenuation of the isolation mount was ap-
PIEZOELECTRIC ACCELEROMETERS proximately 50 percent at 12 cps. The maxi-
mum gain or amplification was 2.5 at 4 cps.
The data discussed previously were meas- The recorder was operational during the shake
ured with piezoelectric or crystal accelerome- test and recorded a 10-cps sine wave as a ref-
ters. These accelerometers, or other electri- erence input. The point to be made here' is that
cal accelerometers with the same capabilities, the recorder was the weak link in the instru-
must be used to record data signals adaptable mentation chain and caused some loss of data.
to detailed analysis. Recorder deficiencies are avoided, however, if
the recommendations below can be followed:
The accelerometers used in the JPL stud-
ies were 1-oz instruments with a frequency re- 1. A recorder with a closed loop tape
sponse of 2 to 7000 cps. An advantage realized, transport should be used. The JPL recorder
with these accelerometers is their minute size, was an open loop machine and was not designed
which facilitates installation. Charge ampli- for the application. Figure A-1 depicts a closed
fiers were used to amplify the input to the tape loop transport as compared to the open loop in
recorder. They were set at 3 g pk full scale. Fig. A-2. The closed loop transport is designed
The full-scale voltage output of the amplifier to maintain a constant tape velocity over the
was attenuated with a series resistor, making record head, regardless of transient changes in
the amplifier output compatible with the re- the tape reel speeds. It also maintains contact[
corder full-scale deviation, but yet maintaining between the tape and record head. - V
low frequency response.
2. As high a record speed as possible
Associated with the use of electrical ac- should be used if the recorder does exhibit sus-
celerometers is the tape recorder. Two differ- ceptibility to shock and vibration. The record
ent tape 'recorders were used in the JPL data speed seemed to influence the recorder's sus-
acquisition. Both recorders were made by the ceptibility to its dynamic environment. Data
same manufacturer but differed in the number recorded at 7-1/2 and 30 ips showed much less
of channels available. The recorder used in the evidence of distortion than the 1-7/8-ips data.
evaluation tests was a 14-channel recorder, Data recorded at 30 ips was "cleaner" than
whereas the other used for the S/C shipment those at 7-1/2 ips. Since the data channels
was a 7-channel machine. A motor generator were FM recorded, the record speed did not
set powered the recorders, since van power was affect the low frequency response, which would
not available. indicate that a steadier tape velocity across the
record head must exist at higher speeds be-
In each application, the data were acquired cause of tape reel inertia.
in the FM mode. The standard FM carrier fre-
quency related to the record speed was direct- 3. Arrangements far in advance of the
recorded on the reference channel. No trans- shipment should be made to perform an evalua-
ducer outputs were recorded on this channel. tion of the recorder's ability to operate in the
service environment. If necessary, corrective
The primary function of the reference measures can be taken to improve the proba-
channel was to identify data distorted by re- bility of quality data acquisition.
corder wow and flutter. This provided ftrther
assurance that data selected for analysis were "Off the shelf" recorders are available that
valid and not recorder microphonics. As was have been designed for operation in a dynamic
expected, the channel did indicate periods of environment. If the application requires so-
spurious noise, most commonly during the RA-8 phisticated instrumentation, these recorders
shipment when a record speed of 1-7/8 ips was should certainly be considered.

32
 PINCH
ROLLER

IDLER
TURN-AOUgND CAPSTAN

HEAD

PINCH
ROLLER

Fig. A-1. Closed-loop tape drive

PINCH
ROLLER

\ hINERTIA
IDLER

HEAD

CAPSTAN

Fig. A-Z. Open-loop tape drive

MECHANICAL SHOCK RECORDERS obtain the 1.5- to 5-cps narrow band noise.
Center frequencies were 15, 20, 25, 30, 40 and
An immediate appraisal of the amplitudes 50 cps and the noise duration in each test was
of shock and vibration occurring during trans- 60 sec.
portation of delicate equipment is often re-
quested by cognizant personnel. An instrument The rms amplitude of the noise was limited
applicable to this requirement is the mechani- by the shaker displacement at the lower fre-
cal shock recorder. These devices simply count quencies, but at 25 cps and above, all 4 chan-
4 the occurrence of amplitudes of shock and vi- nels including the 5-g channel were actuated.
bration that exceed calibrated g levels. No fre- The curves in Fig. A-3 were derived from am-
quency information or time correlation can be ptitude density plots of the input vibration.
obtained from the shock recorder. Typical g Figure A-4 contains a density plot of 1.5-cycle
level calibrations on a 4-channel recorder are bandwidth noise. With the noise duration and
1, 2, 3, 4 g. the center frequency known, the number of
peaks with amplitudes corresponding to the re-
A number of these recorders have been corder channels were calculated. The percent
used by JPL in parallel with the accelerometer deviation in the calibration curves is the per-
instrumentation for checks of amplitudes en- cent difference between the recorder channels
route. Most of'the devices were 4-channel ac- and the actual number of input peaks at each
celerometers calibrated at 1, 2, 3 and 5 g. center frequency. Negative percentages indi-
cate the recorder count was below the actual
Prior to use, these recorders were tested number of peaks. A sample calculation is
on a shaker for calibration data. A set of sam- shown below for clarification. The plot of Fig.
ple calibration curves is shown in Fig. A-3. A-4 contains the distribution of peaks at the
The vibration input consisted of narrow band center frequency of 25 cps, with'a noise band-
noise at selected'center frequencies. A noise width of 1.5 cps. The calculations are per-
source was routed through a tracking filter to formed for the 1 g channel:

33

S---
 ACCEL 9412 X AXIS i. g 5 B.
. .. g 5 B.W.
1..
+100
. ........ .2 g 1.5 -B.W.

+80 +80 -............. .. .... .... .2 g 5 -B.W.

.... .. ------ 3 g 1.5 -B.W.
-p60
X HE 3 g 5 -B.W.

+40
Z
0
+20O "*
0

0ENTE .
-20-......... .......... ... T

lZ .-40>

~VTT.
-100~ i-i I .... .. ...

-60%

1.idica
ot Fsig Aecn
3f eancalepo accelerometer
o w calibrations pr

tive channels. ..

2. Total number of input peaks = center

frequency x duration 2 =25 60 2 = 3000. One probable cause for recorder "readings" ,
being below the input values is the tendency ,'
3. Total number of peaks exceeding 1 g = noted of some channels to stick when the coun- -
percent from step 1 above x total peaks = 0.35 ter contains an arrangement of digits such as
3000 = 1050. 999. This seems to be a deficiency in the
4. Mechanical recorder 1 g counts = 1166. mechanism. ,-'i
The use of mechanical recorders should }:

cen:.
5. Percent deviation = 116/1050 = 11 per- not be discounted, however, in spite of the
......
problems .. above. The JPL data
noted ........
curves
indicate that the devices are useful tools for
i
The plots of Figs. A-3 and A-5 indicate the convenient shock and vibration amplitude meas- C

folkowing mechanical recorder characteristics: urerdents at the most predominant levels of 2.g
or less. In general, the suitability of instru-
1. The percent deviation between actual mentation techniques is dependent on the data
number of peaks and the recorder counts is parameters required, the sensitivity of the pay-
quite high for most channels. These percent- load to the environment, and the funding and
ages are, of course, affected by the tolerances manpower available to support the project. In
of the density plot data. chanical accelerometers cr recorders are most
in tme
ofthe
dratin
oiseand y th reaing ranycpicatinaeutl.clbae e
2. The recorders demonstrate a lack of
consistency in side by side tests. This is dem- Calibration of the mechanical recorder can
onstrated by Figure A-5, which is a composite require considerable effort and planning.

34

U-: .. .
... ...
....
 1.0

0.9

0.8

0.7

! 0~~~.6... --

0.5
P(X)

0.3
0.4

O., Jz

0.0
-9.4 -1.0 -4.6 -2.2 0.2 2.6 5.0 7.4 9.8
ACCELERATION X, (M)
TEST RA STAT ACCEL EVAL STANDARD DEVIATION = 2.3418 NO. OF DATA PTS = 0. 1600E05
LOCATION REF BLOCK XAXIS TmE. "25 CY CF*1. 5 C.Y BWW SC = -0.5652E-01
MEAN z 0.2304E-00 SAMPS/SEC = 800.0000 . KC= 0.3235E-01

Fig. A-4. Probability distribution function peak acceleration

Figure A-3 indicates that the recorder response data proves that the accuracy inczeases as the
varies with noise bandwidth and freqtiency. Fig- frequency decreases. However, the calibra-
ure A-5 gives evidence the recorders are in- ti.3ns are adequate in that the' minimum center
consistent from one unit to another. These frequency of 15 cps does put the prominent peak
limitations require each channel of each unit to at 18 cps within the calibration frequency range.
be calibrated, and to be calibrated for the pre- 1These peaks are generally 3 to 4 cps at the 3-db
dicted service environments, if reasonable ac- down points, which is compatible with the cali-
curacy of measurements is required. bration bandwidths of 1.5 and 5 cps. The JPL
calibration should be adequate for an "inthe
In the case of the recorders evaluated by ball park" idea of the vibration and shock max-
JPL, narrow band noise of 1.5- and 5-cps band- imum amplitudes. If transients or significant
widths was used to simulate the service envi- vibration amplitudes are noted in a sampling of
ronment excitation. Shaker displacement limi- data from a trip, the shock recorders provide
tations eliminated center frequencies less than useful information in estimating the number of
15 cps, which is unfortpnate since the calibration these amplitudes that occurred.

35
 I B.W.

M.. U. .. +i --.
'
. ....................... .-
.-....
....
60i:i ::iMi ; : T:: 1::.

n0
0Z"T 4.:LN
Re-:s:eac
M. levenso:(Langle:: r
.. . ........

.
Q og 105-
with :C:e:n
.... :0 .0
... .

an..
+

n i
-
we a u
-a.m.i
40- -40 .. ,.. . .. . ......
. .. .+ . . ,+, ....

0+,+ ... .... '+i ++.+

.... ... . .. . .. . .
+..
80 20 25-- 30 35 40
.... ............. 45 50
....-.

_____________._____,____________;_--1 ... ....

... .. i
15-40 03 4 55

ofQyo+r+asmig+ndgveyu ra
valu uiga f 0i that this is a .c.nervaie

We
Mr.~~~~~~~~~~~
sn o 0 e .cle...rti fteQs

Fig A5.Copaiso o tospeancal a plitudees alireaopotion alth

u sokrg.a Fi A5. Comparsoe t wo mepnechnclaclrmtrclbain

Mr. SCleeso WeanareyusingaacQCoft10. :Weiratio of e Q's. 0 nddmig aiso

E*1
us teourst shock
o a sn pte theiesonse2,00 nd0 nti ae eaeuiga

-606

DICSSO
 A REALISTIC DERIVATION OF A LABORATORY

I VIBRATION TESTTO SIMULATE THE OVERLAND

TRANSPORTATION ENVIRONMENT
John A. Kasuba
,
Aberdeen Proving Ground, Maryland
I"I
This paper treats the problem of deriving a vibration schedule for se-
curely tied down vehicle cargo. The derivation is based on the princi-
ple of equivalent damage which relates the severity of the field and
laboratory environments. This principle was applied in deriving an
automotive test schedule to simulate the overland cross-country and
paved road transportation environments of truck, tracklayer, and
trailer-type vehicles. The automotive test schedule was combined with
the existing aircraft test schedule to present a more complete trans-
portation environment.

INTRODUCTION THEORY OF SCHEDULE

DERIVATION
The primary objective of this paper is to
present the techniques used to develop a labo- Description of Cargo
ratory vibration test schedule having damage
potential equivalent to that encountered in actual
overland transportation environments. It will be assumed for tile present that ve-
hicle cargo can be represented by a one-degree
The initial test schedule was developed at
AedePrvnGrudi193(-)Itws of freedom mechanical system with known
,Aberdeen Proving Ground in 1963 (1-3). It was characteristic constants. Consequently, the
derived from data gathered on several repre- natural frequency, the Q and the bandwidth of
sentative vehicles operating on fixed profile the system can be calculated. The first two
courses. In the opinion of Development and quantities indicate where, in the frequency
Proof Services personnel, the Belgian block spectrum, resonance will occur and how much
and paved road courses for wheeled and tracked damping is present to prevent violent vibration.
vehicles, respectively, were adequately severe The bandwidth spans the, frequency region be-
to simulate cross-country behavior. However, tween the half-power points of the amplification
it was felt that testing on cross-country courses factor and indicates the region of high amplitude
would have been desirable. This was not pos- response.
sible because course selection was dictated by
a landline data acquisition system.
The inputs to the mechanical system are
The acquisition of a radio telemetry sys- obtained by recording cargo-bed accelera-
ter removed therestriction on course selec- tions on magnetic tape in analog form. The
tion; thus, the derivation of a new standard was signals from the tapes are then filtered with
undertaken to incorporate cross-country oper- a pass-band filter whose center frequency
ation, to gather-data on vehicles previously not and bandwidth correspond to the natural fre-
tested, and generally to make a more extensive quency and bandwidth of the mechanical sys-
study of vehicle behavior. The automotive test tern. The filtered data represent that portion
schedule so derived was combined with the ex- of the entire frequency spectrum which could
isting aircraft test schedule to present a more cause the mechanical system to fail by fatigue
complete transwportation environment, failure.

37
4I Fatigue Failure and Damage Potential z
E ni(Si) = ne(Se)/A(4

The degree to which a material is able to i( A(

withstand repeated stress reversals is indi-
cated by S-N curves. The general equation for Furthermore, since the stress level is dire6tly
these curves, which are straight lines on log- proportional to acceleration, Eq. (4) becomes:
log coordinates, is given by z

SNA R (1) E ni(g 1 ) 1/A ne(ge)i/A (5)

where S is the stress level, N the number of K

reversals at s to cause failure, A the slope of The combination of n, and ge thus de-
the line, and R a constant, scribes the time and amplitude of a sinusoid
which has the same damage potential as several

When a material is subjected to several sinusoids of different amplitudes. Furthermore,

levels of alternating stresses, failure cannot be if the unknown quantities, that is, ni, gi, and
predicted by S-N curves alone. In this case, a /A of Eq. (5) could be determined, then a
theory of cumulative damage such as Miner's simulation schedule for the mechanical system
criterion must be used. The mathematical ex- could be established.
pression for this theory is The results of the one-degree
of freedom
mechanical system example can now be sum-
z C! c (2) marized. It was assumed that both the descrip-
=i tion of the cargo and the excitation to which
this cargo was subjected in the field were
where available. A simulation schedule had to be ob-
tained which would expose the cargo to the
Z = number of different stress levels, same damage or damage potential to which it
would be exposed in the field. This could have
ni = number of stress reversals experi- been done by obtaining the number of times, ne,
enced at a stress level si, a single level of sinusoidal acceleration, g ,
would be applied to meet the necessary require-
N1 = life at stress level si as indicated by ments. It is important to note that this proce-
the S-N curve, and dure did not exclude any levels of acceleration;
it considered the frequencies which are ex-
C = indicator of damage accumulation. pected to cause damage, and it applied the
principle of equivalent damage.
According to the theory, failure occurs
when C = 1; however, experiments indicate that
C varies to some extent. Assigning a value of Independence of Material Proof
C can be avoided by requiring that an item sub-
jected to a multiplicity of stress cycles at many The distribution of positive or negative
stress levels experience the same damage po- peaks within the bandwidth of filtered data was
tential at a single stress level. Under these nearly Gaussian. The distribution is expressed
requirements Eq. (2) becomes: by
g2
n
i ne (3) 2N 2a2 (6)

N1 Ne 'f ai 14

wher~e where
n = frequency of occurrence of peaks,
n= number of stress reversals at a
stress level of Se, and N = number of cycles in the data sample,
Ne = stress reversal life of the item at a = rms value of acceleration within the

stress level Se' passband, and

A combination of Eqs. (1) and (3) yields: g peak values.

38
 The substitution of this equation into Eq. where ne/N is the time-scaling factor and ge/a
(5) after the summation is taken to the limit is the amplitude-severity scaling factor. The
yields: solution is shown graphically in Fig. 1. A rel-
atively limited range of amplitude scaling fac-
CO Z 7)g tors is shown since it is undesirable to increase
n(g)/A 2N f /A 22 dg. (7) the level of simulation acceleration beyond that
V'- 0 ( actually present in the field environment.

In the absence of excessive shock, it is The figure indicates that when the severity
highly unlikely that a peak acceleration of either and time scaling factors are approximately
infinity or in excess of 3.5 times the standard equal to 3 and 0.02, respectively, the test
deviation is possible. Introduction of these schedule applies for any material having a I/A
limits and a new variable defined by value between 5 and 20. This range is so wide
that it includes most materials; thus, the limi-
2 tation'imposed by selection of A is removed
gU (8) and the test schedule may be applied to any
2a2 material.
yields
Generalization
(3.5)2
n(g) N/A f 2 (9) Initially, it was assumed that vehicle cargo
- u e'ud, could be represented by a one-degree of free-
dom mechanical system with known character-
The solution of this equation is given by an istic constants. Generally, this is not true;
incomplete gamma function which, after rear- such an accurate description of cargo is not
rangement of terms, becomes: available. A more realistic assumption is that
cargo can be represented by many one-degree
) of freedom subassemblies, all having different
ne i/A 2 1/2A (2 1 3.2
-
N
" -- + , 2- ), (10) characteristic constants. In this case, the
V57i amount of data filtering must be based on a

0.05
Amplitude Scaling
Factor

S2.8
0 0.02
4,

to

cc 0.03 2.9

Er,

0.023

0.0. ' 6'3.2

0

Fig. 1. Time scaling factors vs amplitude scaling factors

39

TTX -Pa
value of Q, which is representative of all the Acceleration data were gathered on wheeled
subassemblies. and tracked vehicles as they were operated
over various cross-country and fixed profile
If the selectior) of the pass-band filter is test courses. The vehicles were loaded at
based on a subassembly with a small Q, then their rated capacities with either dummy loads
the filter bandwidth is wide and the amount of or inert ammunition. Selection of vehicles was
energy in the filtered data is large, resulting in based on the frequeney of use and expected use
an overtest of subassemblies with large Q val- in the future. Course selection was dictated by
ues. Exactly the opposite is true if the selection the courses which are used for type-classifying
of the filter is based on a large Q. Since an particular vehicles. Testing was conducted
overtest produces an acceptable item and under- over courses in a frozen or hard condition. In
test does not, the filter must be selected on the each case, accelerometer outputs were re-
basis of a small Q. Furthermore, an overtest corded on magnetic tape after they were suit-
of high Q subassemblies is justifiable because ably transmitted and time indexed. A schematic
they do not represent the best design practice. diagram illustrating the data acquisition system
Therefore, a conservative value of 2.8 was se- is shown in Fig. 3.
lected to provide a margin of safety in the final
result. With this value, the entire frequency
spectrum can be subdivided into frequency in- Data Reduction
tervals by placing electrical filter bandwidths
next to each other as shown in Fig. 2. The fil- The initial step in data reduction was to
tered acceleration data within each bandwidth determine if data recorded on one portion of a
include acceleration components which cause course were similar to data recorded on another
most of the damage to the associated mechani- portion of the course. Mathematically, this
cal subassemblies. equivalence is referred to as a stationary proc-
ess, that is, a process that does not vary with
time. The existence of such a process would
SCHEDULE DERIVATION permit data gathering runs of shorter duration
Data Acquisition and equally short data reduction runs. The es-
tablishment of this process was based on the
The data acquisition system consisted of a root mean square value of acceleration. The
test vehicle and an instrumentation system. rms values were determined for various

I f 3
-f f2 f 3 j - frequency

Fig. 2. Successive electrical filter bandwidths

motions. Acceleromete ansystem

Amplifying Tape

i Electronics Recorders

\ ~ ~Time
orCode Oscilloscopes

Fig. 3. Data acquisition system

40
segments of an entire data record and compared. number of reversals by subtracting the number
Such an investigation indicated that a stationary of crossings between successive amplitude
process existed on the fixed profile courses, levels.
but not on the cross-country courses. Conse-
quently, short-duration data runs on the fixed In the calculation of the original schedule,
profile courses were adequate, while maximum it was necessary to select a 1/A value for the
possible time runs on the cross-country courses material which was most likely to fail in the
were necessary. system. In that case, aluminum was the pre-
dominant material with 1/A equal to 19.6. Sub-
The data were passed through a set of half- sequently, it was shown that the schedule was
octave filters having a Q of 2.8 into an axis- nearly independent of material. Since the data
crossing analyzer. This analyzer detects the for the new schedule fall within the previously
number of times the amplitude of a signal ex- mentioned restrictions, 1/A was picked as 20
ceeds each of ten preset, linearly separated, for convenience and for consistency with the
positive or negative levels of amplitude. The original derivation.
data from this analyzer can then be used to
evaluate the summation quantity of Eq. (5). The Data for this example came from vehicle
data reduction system is shown in Fig. 4. operation on a cross-country course for which
a stationary process does not exist. However,
it is reasonable to expect that the maximum
Data Evaluation data run of 1.96 miles, on that paiticular
course, is representative of any other distance 7
Data from the axis-crossing analyzer were which could be traveled on the same course.
scanned to determine which filter center fre- Such a data run is at least ten times greater
quency contained significant levels of acceler- than is necessary for a data run on a fixed pro-
ation. A sample axis-crossing data sheet is file course. For convenience, the results are
shown in Fig. 5. In this case, the center fre- extrapolated to 1000 miles of vehicle operation.
quency of interest is 5.6 cps. The mechanics
of calculating a test schedule from these data In the process of calculating the schedule,
are illustrated in Table 1. it is possible first to pick g, the equivalent
acceleration level, and then to calculate t(,, the
The axis-crossing analyzer is constructed time of test equivalent to 1000 ,miles of vehicle
to indicate only that an acceleration signal fell operation, or to do the reverse, that is, pick t e
between two successive acceleration levels, and calculate ge. The results of such an ap-
Therefore, it is assumed that the level fell proach are shown in Fig. 6 which illustrates the
midway between two successive levels of ac- original, and presently used, schedule and in
celeration. The number of crossovers in each Fig. 7 which illustrates the tentative new
amplitude level are then converted to the schedule.

Tape Playback
Equipment

SAmplifying Half-octave Amplitude

E lectr nic
l o s Filter Analyzer

L4Oscillographs ,

Fig. 4. Data reduction system

41
 a, -C

00

I>-c

V
01 I

in cc~m:, ---

1 at C N .c
-Av ,b

I" '

oco

a0

'n 0

0 -Wc v -4
.4 u UL %D V 0
m &A - - -
K4 4.4

s 10 $

i F 4
u 0

V42

- -- ~s'- -- ;- - ~ ~ 's w'--

 TABLE 1
Sample Calculation

Vehicle: Truck, cargo, 2-1/2 ton, 6x6, M35.

Course: Cross-country No. 1.
Transducer Location: 2 vertical, front of cargo bed.
Average Road Speed: 19.7 mph.
Data Sample Time: 360 sec.
Equivalent Length
of Data Sample: 1.96 miles.
Filter Center Frequency: 5.6 cps.

Amplitude setting of 0.2g 0.4 g 0.6 g 0.8 g 1.0 g 1.2 g 1.4 g 1.6 g
axis-crossing
analyzer

No. of crossovers 1501 702 332 151 68 28 8 0

No. of reversals n i 799 370 181 83 40 20 8

Average amplitude g. 0.3 g 0.5 g 0.7 g 0.9 g 1.1 g 1.3 g 1.5 g.

2ni(g.)l/A 799 (0.3)20 + 370 (0.5)20 + ... 30.680 = ne(ge)/A

where

1 = 20. The equivalent acceleration test level g, is picked as 1.5 g,

A

ne = 30,680
- = 9.23 cycles =4708 cycles and
(1,5) 20 1.96 miles 1000 miles

te -f ne 4708
= 6056 =
14.0 rain.

The validity of the approach used to derive If the item does not have a clearly defined res-
the original test schedule was established by onant condition, the frequency spectra would
conducting a test to compare the actual trans- have to be covered by using a sweeping rate
portation environment of artillery projectiles outlined in MIL-STD-810 (5) which specifies
with that of the laboratory vibration test. The that for cycling tests of less than 500 cps max-
amount of damage to the high-explosive fillers imum frequency, the frequency range shall be
resulting from road and laboratory tests was cycled logarithmically from minimum to maxi-
about the same (4). mum in 7.5 minutes for the total test time
specified. If the item has reasonably distinct
resonances, it would have to be vibrated at the
SCHEDULE APPLICATION principal resonant frequencies (not more than 4)
with a dwell time at each resonant frequency
Automotive Transportation Phase equal to 1/6 of the total test time. The rest of
the frequencies would have to be cycled loga-
The schedule is set up to represent 1000 rithmically as previously mentioned.
miles of vehicle operation. Therefore, the
user of the schedule must first determine what
type of vehicle will transport the item and what Combined Automotive and Aircraft
distance or equivalent damage-inducing distance Transportation Phase
this item will be transported. For example, if
it is determined that an item will be transported Although this paper deals with the automo-
1500 miles on a tracklayer vehicle, this item tive environment, cargo has a great-probability
would have to be tested for 20xl.5 or 30 minutes. of being subjected to other transportation

43
!w

4i
L
16g92MINUTES1000MLES

-TRACKLAYER

x 0.10
U
z 2.59 15IvNUTES/ 1000 MILES
- TWO-WHEELED TRAILERS
0I.J

0 QOI
ill

1,,0MILES .5g 1,5 MINUTES/IOO

1 0 44 5,6 70FREQUENCY (CPS) 010721 10

LoL
Fig. 6. Original automotive test schedule

I
I5IUESIO
oo lI I-

212 62 1000
314 10 1
.0 44 54.6 01o
.g 520MINUTES/IOOOMILES
1(CPS)10
FREOUENCY

I
z I~I I -- SMUTES/IOMILES
_2.5!
-I WrTR TRAILERS
08- - - - ITRC- i
AN EMITLES
III I
TRCSAN E TALR
-
I___I_ I i
_____ schedule
4 test
9 01292210
Fig. .' ORial 0automotive
3 S,
0
4457
Om iv 2qTESt /schedLE
Fig. 7. eI III
se au

I I ,

t44

- -
-
II
environments as well. It has been established SUMMARY
that the aircraft and automotive vibration envi-
ronments are more severe than the railroad A test schedule was derived for cargo
and ship vibration environments; thus, the lat- transported by truck, tracklayer, and trailer-
ter two environments can be neglected (2). The type vehicles. The schedule is restricted to
aircraft portion of the test schedule was ob- cargo which must be securely fastened to the
tained from MIL-STD-810 (5) which depends on cargo bed; otherwise, the validity of the test
the weight of the cargo. For test items weigh- schedule will be in doubt. This restriction is
ing more than 50 lb, the acceleration amplitude required because the schedule is representative
can be reduced by 1.0g for each 10-lb incre- of a vibratory environment and not of a shock
ment of weight over 50 lb. However, the level environment. The automotive test schedule was
will not be reduced by more than 50 percent of combined with the existing aircraft schedule to
the originally specified test level. This de- present a more complete transportation envi-
rating has been incorporated in the aircraft ronment.
portion of the combined automotive and aircraft
environment test curves shown in Figs. 8 The derivation of the test schedule was
through 12. Furthermore, the upper frequency based on the principle of equivalent damage.
levels for these curves are also weight- This principle requires that when vehicle cargo
dependent. The derating of the frequency is is exposed to a cert.in degree of possible dam-
shown in Fig. 13. age under field conditions, cargo must also be
exposed to the same degree of possible damage
The application of the combined test sched- in the laboratory. Such conditions relate the
ule requires that for the automotive portion, the severity of the test with distances that the cargo
test time be computed on the basis of 15 or 20 will be transported.
minutes of exposure per 1000 miles of vehicle In the process of deriving the schedule, the
operation. For the aircraft portion of the test following safety factors were incorporated:
schedule, the test time is equal to either the
actual flying time oi to 3 hours, whichever is 1. Maximum summation quantities for each
less. The rate of cycling is as outlined in the vehicle irrespective of acceleration direction
previous section. were used;

Lo

\
iI I
IZ 2.g 15 MINUTES/I000 MILES
I I I- TWO-WHEELED TRAILERS
I - L5 LSMINUTES/IOOOMILES
TRUCKS AND SEMITRAILERS
I v- .-- 2OMt'uTES/IO00 MILES
TRACK-LAYES
INTERSECTION OF AUTOMOTIVEvI O.Og

ANDAIRCRAFT SCHEDjJLES I I
o-o06------------ -r -

.0 -ll II I iii
I! IIl I a
II I '

i I' i I
________I___II'____
I II I I I a

III Ili I I a
I I I II II I
0.001- I -I I
II ,I ,
1.0 4.4 54ZO 10 41720 3337 52 100 313 1000
FREOUENCY (CPS)

Fig. 8. Combined automotive and aircraft

test schedule, item weight, less than 60 lb

45
 LCL

!.0 '

J -- I1.5q 15MINUTES/IOOOMLES
- TRUCKS AND SEMITRAILERS

I1 OOMILES
.0E20 MINUTES/
TRACK-LAYERS
I I
Uj

01-

~00 29 - ------

U3 INTERSECTION OF AUTOMOTIVE I I
"j AND AIRCRAFT SCHEDULES I
II II I
II II I
il II I
II II I
II II I

II II I
II II I
II II I
II II I

000!1 I '.000__
I 44 5.4 10 2632 52 100 280 ;000
FREQUENCY 'CPS)

Fig. 9. Combined automotive and aircraft

test schedule, item weight, 60 to 70 lb

I 25g 15MINUTES/IO00 MILES

TWO-WHEELED TRAILERS

,J0.10
U]
U,

S0025---------- -- -- ---
_j INTERSECTION OF AUTOMOTIVE i\
"
00 AND AIRCRAFT SCHEDULES
oo 001 I I
3.52

II

I _ _ _ _,'

00011
1.0 7 10 O 4452 100 262 1000
FREQUENCY (CPS)

Fig. 10. Combined automotive and aircraft

test schedule, item weight, above 60 lb

46
 II \1,

H
I I
\-
\
I.5g ISMINUTES/IOOOMILES
-TRUCKS AND SEITRAILERS
'"

- o~ II

Sl.Og 20 MINUTES/IOOOMILES
I I-TRACK- LAYERS
U] II \

0
0=
02Z -- -
: ,I -- T
~II
1-
_WINTERSECTION
(0 AND AIRCRAFT OF AUTOMOVE I I\
SCHEDULES II
II __ I \

3..
I Ii I A M I 'I I0 M
II II I
AN
I0 AIRCRA
II
445. SCDEDULERAISI
IQ II I
II II I
II II I

0.01I t I I I
10 44 54 10 3037 52 100 243 1000
FREQUENCY (CPS)
0 ~ ~ ~ ~
Fig. 12. Combined automotive and aircraft
O 20________
MINUTES/1000MILES

test schedule, item weight, 70 to 80 lb

IO

I I \\ -TRUCK AND SEMITRAILERS

I log 20 M hUTES/iOOOMILES
I, \ TRACK-LAYeRS
"' \

z it
- I
o 08 I- - - - - -

-J INTERSECTION OF A TOMOTIVE' ,
o
01 AND AIRCRAF T SCHED)ULE-S
... I
oII
I I IiII I
II II I5
ii II
I I Ii II
II II I
I I I I I
0 01
,I I I I I
1.0 4 5.4 I0 33 40 52 100 221 1000
FREQUENCY (CPS)

,Fig. 12. Combined automotive and aircraft

test schedule, item weight, above 80 lb

47
 Ui

"1 ~ __
~~600 l

500b4 , i

400

2oo.

Fig. 13. Weight vs frequencyU for aircraft cargo1-I

2. The "filterbandwidths were based on a permits the test item to achieve maximum re-
low Q; sponse.

3. High acceleration values were extrapo- CONCLUSIONS

ated to 1000 miles of vehicle operation; andTh te ni u of d r v g t et st c e ul
is based on sound principles and results in a
4. A sinusoidal forcing function was spec- realistic automotive test schedule for securely
ified which, unlike a random forcing function, tied-down vehicle cargo. "

REFERENCES.!

1. J. A. Tolen, "The Development of an Engi- an Engineering Test Standard Covering the

neering Test Standard Covering the Auto- Automotive Transportation Environment,"
motive Transportation Environment,'" Aberdeen Proving Ground Report No. DPS-
Aberdeen Proving Ground'Report No. DPS- 1633, April 1965
1127, November 1963
2. "Te
JA. olen Dvelomen of n Egi- 4. H. Graves, "Final Report of Research Test -
2.J Te
A oeneelpet fa Eg-of Cross-Country Transportation- Vibration
neering Test Standard Covering the Trans-TetoAmuion"brdnPovg".
porttionEnvronmnt o Maeria," Aer-Ground Report No. DPS-1519, January 1965 ,
deen Proving Ground Report No. DPS-II90,
December 1963
5. MIL-STD-810 (USAF), Military Standard "
3. J. A. Kasuba, "An Analysis of the Methods Environmental Test Methods for Aerospace
and Criteria Utilized in the Development of and Ground Equipment, June 1962

DISCUSSION

Mr. Forkois (U.S. Naval Research Labora- Mr. Forkois: This would indicate that you
or2_): What was the maximum g level that you did not have any hard spring bottoming.
attaified in these cross-country runs ? Was it L.Ksb:W re oeiiaeti
as hih as 0.gfrom our data by screening the readings through
Lt. Kasuba: No. It went up to perhaps 3 g, oscillographs and so on. We tried to avoid un-
certainly not more than 4 g. usual conditions of accelerometer ringing.

48
 A METHOD FOR ESTIMATING ACCELERATIONS
OF SHIPPING CONTAINERS MOUNTED ON AN
IMPACTING RAILROAD CAR
John J. Scialdone*
Goddard Space Flight Center
Greenbelt, Maryland

A method is presented for estimating the response of a container and

its contents, when a railroad car on which they are mounted impacts a
group of braked cars standing. in a classification railroad yard. Equa-
tions are derived for the container car impact equivalent velocity, con-
tainer car acceleration, car body transmissibility, and container trans-
missibility. By applying characteristic data of the car and its coupler,
the natural frequency of the car is calculated, and the equivalent impact
velocity to the container car is obtained from the actual impact speed
and the car weights. Then, input accelerations of the car and shipping
container are calculated over a 1- to 14-mph range of input velocities,
for bottoming and non. bottoming of the coupler during impact. From
these results, the container transmissibility is obtained, since we know
the inner container design natural frequency, its damping characteris-
tics, and the forcing frequencies resulting from the impact at the car
coupler. The resultant accelerations at the inner container are obtained
for 0 and 10 percent coupler viscous damping ratios. The calculated
values are compared with the upper bound values of the accelerations
obtained from impact tests. The results indicate that if a 10 percent
value for equivalent viscous damping of the car coupler is used, for car
impact speeds above 7.5 mph, accelerations at the inner container and
at the car bed adjacent to the outer container will exceed 1.8 g and 20 g,
respectively.

INTRODUCTION speeds between 5 and 7.5 mph are shown super-

imposed on Fig. 1 of this report. Because
Railroad impact tests were conducted to these impact speeds may be exceeded (1), the
confirm the suitability of a shipping container analysis which follows was performed to:
rigidly attached to the floor of a railroad car.
The container consisted of an outer steel shell 1. Correlate the rests results and deter-
supporting internally a spring- and damper- mine the responses of the components of the
mounted inner container to which was firmly system during impact;
attached a dummy weight. These tests con-
sisted of impacting the container transport car 2. Obtain the accelerations at the outer
traveling at a speed of 7 mph against three container and inner container for impact speeds
standing braked freight cars. The tests were other than those of the tests; and
intended to simulate the majority of impacts
experienced during normal railroad switching 3. Establish the limiting impact speed fo the
operations in railroad classification yards. acceptable specified inner container acceleration.

The tests indicated that for the 8-mph im- METHOD OF ANALYSIS
pact, the shipping container attenuated the lat-
eral accelerations to the inner container to The method used here for estimating car
1.8 g. The results of these tests at impact impact accelerations is intended to avoid the

*Part of this work was done while the author was associated with Westinghouse Astronuclear Lab.,
Pittsburgh, Pennsylvania.
49

. ...
_.......
= ....
r-
. .-.-- = -- :
' . '. ._.. .. ... . .. . .. .... __
 ,J-

-,,- - -

3 #j

- TI TEST RESULTS '

z1I -

0 2 4 6 8 12 14 16 10
IMPACT SPEED (mph)

Fig. 1. Inner container acceleration p

vs car impact speed

solution of nonlinear differential equations and Also, a mass momentum transfer equation
the use of analog and digital computers. It em- can be written expressing the balance of mo-
ploys the momentum and energy equations to mentum between the impacting car and standing
establish a velocity change applied to a single cars:
degree of freedom system. The response of
this system including the excited frequencies is
used to calculate the response of the container.
WA VA
g -
WA + WB
WA VWA WI
g
(2) 1
Since the momentum and energy equations are
not applicable for impacts during which sliding By -solving this equation for VAB, we get
of wheels may occur, i.e., for nonconservative
systems, the results obtained by this analysis WA
for the car accelerations can be modified to VA W VA" (3)
+ WB
include, if necessar'y, an empirical coefficient. WA
The coefficient is obtained by comparing the By substituting Eq. (3) into Eq. (1), replacing
calculated results with experimental data re-
ported in the literature for humping tests. ple , an dividin
in cconstant, the of
pler spring and dividing both sides of
the equation by WA, we obtain
Container Car Input Velocity Change F 2 KW
F ,K:( (4).
1 W;A9
A 4"WE/.:
WA ~ !::
Multiplying the numerator and denominator of

the right side of Eq. (4) by g and extracting the

square root yields the acceleration, G, in unit,,;
- - ,- - ,- j WW of acceleration of gravity, experienced by WA,
as follows:

The kinetic energy of the impacting car F A Kg

will be used to impart a kinetic energy to the G - F - VA WA Kg (5
standing cars and compress the impacting cou- WA 9AB A
plers. The following balance of energy equa-

tion for the strain energy, the initial kinetic In the above expression, (Kg/WA) = 2 ; i.e.,
energy, and the kinetic energy of the entire the natural frequency of the impacting car con-
system after impact can be written: sists of a lumped mass wA and the spring con- A
stant k, provided by the coupler "springattached
2 to the mass. The other square root term can
Fd WAVA (WA + "E) VAB () be considered as the fraction representing the
T- 2g 2g amount of the initial velocity returned to the

50
impacting car. The velocity change applied to experienced if bottoming had not occurred. The
the container transportation car can then be equation is:
given by __...

G = GO B
VE VA 1-V "
"A(6)
A G1 "o L "
VA Wd B L

This equivalent velocity will be used in esti- where G8 and d are given by either Eq. (7),
mating the acceleration provided by the-impact or Eqs. (8) and (9).
to the container car. In this type of impact, the excitation felt
by the car will consist of a low and a high fre-
Container Car Acceleration quency acceleration. These two frequencies
will be related to the spring constants
of the
Impact With No Bottoming of Coupler - coupler (K.) and that of car bed frame- (KB),
Equation (5)provides the acceleration felt by and the bottoming deflection of the coupler (,i,)
the car on impact if coupler damping is as- and the deflection (d.), which would exist if
sumed not to be present and coupler bottoming bottoming had not occurred. The relations
has not occurred. Equation (5) can be expressed providing these frequencies are given in terms -

simply in terms of velocity and natural fre- of the frequency which would exist if bottoming
quency: had not occurred, as follows (2):

Go=VE - (7) wo 2 sin-, d s

9 a) 7 do

If an amount 8 of the critical damping is

present at the coupler, the acceleration felt by K
1
the car will be (2): + . - " "tan' '0
KBKB(d2 / - )

E f N tp (8)
e 6 G0
GVE--'& Go e
and
where the time, tP, corresponding to maximum
acceleration is: -
1
(14C2)(1 2) /2 O - 2-tan-1 (2

a"
N ( 12 Lf_3

The coupler deflection cprresponding to where &j is the high frequency, co is the low
these accelerations will be given by frequency, and w the frequency which would
exist without bottoming.
d GW (9)
Car Body Transmissibility

The frequency of oscillation under this

type of impact will be obtained by the lumped It will be assumed, conservatively, that the
car weight and the coupler spring constant. impact load experienced by the car frame at
the coupler will be transmitted undiminished to
Impact With Bottoming of Coupler - When the container mounting. This assumption is
the coupler bottoms, the behavior of the car based on a conclusion that no significant differ-
coupler system will be the same as that of a ence in g loading was noted at different loca-
system possessing a bilinear elasticity. The tions of the test car during impacts (3).
spring will deflect up to a closure deflection d,
with a spring rate K,, and beyond this deflec- Container Transmissibility and
tion with a much higher spring rate KB. It is
shown (2) that for the case of bottoming impact, Ir
the acceleration can be obtained in terms of the Impact With No Bottoming of Coupler The
acceleration and deflection which would be inner container accelerations will be calculated

51
by modifying the accelerations to the outer relations previously indicated in the section
container, Eq. (7) or (8), by the transmissibility entitled "Method of Analysis," it is possible
for a Viscous dnimped system. The transmissi- now to calculate -the.accelerationat the system.
bility is given by However, the parameter to be used must first
be established. These are the car weights,
2 spring constants, natural frequencies, and
I + (2,8 (13) equivalent impact speeds.
T- (13) l
1 (21 2 ,2 The total weight of the cars will be WA+ WB =
- + [2,3 71,200 + 153,600 = 224,800 lb.

In this expression, o is the forcing fre-- Equivalent Impact Velocity

quency of the excitation, which in this case is
the natural frequency of the car considered as The equivalent impact velocity to the con-
a single degree of freedom lumped mass at- tainer car is, from Eq. (6), 4
tached to a spring. The other parameters, 8
and w,, are characteristics of the container .0 A
design. v VA 2 0.826 (16)

Impact With Bottoming - For this case, the

transmissibility will be calculated also by the Spring Constant and Natural
use of Eq. (13). This relation will be employed Frequencies
to find two transmissibilities, (a) one for the
low frequency acceleration, obtained from Eq. The spring constant on impact is provided
(7) or (8), and (b) the other for the high fre- by two couplers in series. The spring constant
quency acceleration, obtained from Eq. (10). for the coupler will be estimated from data
To obtain these transmissibilities, the forcing given for the energy absorption and travel of
frequency, ,-, to be used, will be: (a) L,Ioh- the coupler during impact, as specified by AAR
tained from Eq. (11)for the low frequency exci- (4). The draft gear must have a minimum ca-
tation, and (b) wH, obtained from Eq. (12) for pacity of 18,000 ft-lb, and the travel of the gear
high frequency excitation. The magnitude of must be not less than 2.5 in., From this data,
the resultant acceleration R produced by the the spring constant of a single coupler being
two excitations AL and AH will be obtained by deflected is
the use of the cosine law for the addition of two
harmonic functions. Thus,
Ko = 2E = 18,000 x12 x2 = 75 x I03 lb/in.
(17)
d2 (2.5)2
2
R = AL . AH
+ AL AH COS (14)
For two couplers in series the spring con-
where the angle 0 will be the difference between stant will be
8
the phase angles OH and L of the high and low
frequencies accelerations; i.e., Ko 2E 18,000 x 12 l (18)
K =52 35 .xl03 Ib/in., (18
2 2d 2 2.52
[2O 23 and their relative displacement before bottom-
1
0 tan 2 (15) ing is
"Nd ,= 2 x2.5. = S in. (19) !

CALCULATIONS The natural frequency for the car acting on

a single coupler can then be calculated as
Car Weights
x 10 3 X
KgA V/ (20)
The weight of the container car WA, includ- Y, K - 5 386 20.2 rad/sec,
ing the container, was 71,200 lb. The weights 0 WA 71,200 (
of the three stationary cars were 51,200, 52,000,
and 50,400 lb, respectively, providing a sta- corresponding to 3.2 cps. For the car acting
tionary total weight WB of 153,600 lb. With the on two couplers,

52

'
 For a 10 percent critical damping present
13.75 rad/sec, (21) in the system, Eq. (8), plus the relation for the
V/Kg = T,10 .36 tiite for peak magnitude, reduces-to
corresponding to 2.2 cps. G 0.82 G= 0.443 VA. (23)

The deflections corresponding to Eqs. (22)

Car and Outer Container Input and (23) will be from Eq. (9),
Acceleration
Input With No Bottoming of Coupler - The do = 0.518 VA 712 = 1.058 VA
3
magnitude of the acceleration experienced by K 0. 35 x 1O

the container and car, where no bottoming of and

the coupler occurs, will be obtained by the use
of Eq. (7) or (8), together with Eqs. (11) and GW7 .2 x 0 (25)
(21). For no damping present at the coupler, do -- = 0.443 V4 !-- = 0.902 VA (
K 35 103
Eq. (7) will reduce to
V 61N 0.826 (13.75) VA The results of-these equations for several
V- g 32.2
32N impact speeds have been shown in Fig. 2. "

Input With Bottoming of Coupler - To apply

0. 826 x 13. 75 x 5280 VA (22) Eq. (9), it is necessary to estimate the spring
= 32.2 = 0.518 VA, constant KB of the car body. From a graph (5)
of applied force versus body yield, it can be
where VA is expressed.in mph. estimated that the car body spring constant is

68- LEGEND:
64- Undesirable condition: I
64- Unbalanced load & coupler misalig., ref. 7,
-.- Ideal condition for sliding sills, hydr. i
damp. coupler, ref. 7
-,- Flat car w/trailer impacting.
56-- 3 braked cars, ref. 9 .
- Flat car w/missile impacting
-- 3 braked cars & crane, ret. 3
Z -.- Max. acce:er,car w/friction gear &

,- 48- rubber gear. ref. 4 KIT

I-A-P =59.6Kip

C, 40 ______ , I W. =68.2-70.4Ki

Z 32

'Cz 2-
24 , KI
W71.2
.....
A p
.V.
o0: Calculated =
Io Accelertion.'-

_______/ I _FA7OKI
IIle

06 2.76 8 0 12p

00. ' : IjI" .

C
'e

IMPACT SPEED (mph)

C
Fig. 2. Car and outer container acceleration
vs car impact speed with zero coupler vis-
cous critical damping
53
 .4t

KB 1310 x 103 lb/in. (26) not used for the range of impact speeds being
considered.
By substituting Eqs. (16), (18), (19), (22) or
(23),'(24), and;(26) into Eq. (10), the following
equations are obtained for acceleration with Container Transmissibility -
bottoming: Accelerations at Inner Cnntainer

1. For no damping at the coupler, Impact With No BoV.6ming of Coupler - The

transmissibility for impact speeds less than
v = 4.55 mph for no coupler damping, and V =
03 110 + 21 5.5 with 10 percent coupler damping will be ob-
A
A V 35 103 + 058 VA 351 tained from Eq. (13), where the following sub-
stitutions havebeen made:
0.51 8 x 6 .12VA/1 2.7
2
1 N 11.9, the design natural frequency V
VA of the inner container with its content.

21.7 2. 8B= 0.35, the fracticn of critical viscous

3
.1 6 VA 21 3.16 VA2 21. 7 ((27) damping included in the suspension. This is
VA2 also a design value.
2. For 10 percent critical damping at the 3. c = 20.2 rad/sec, the excitation fre-
coupler, quency resulting from the oscillation of the car
when a force is applied to the car coupler.
2.72 /VA2 - 21.7 (28) This value was obtained from Eq. (20).

Equations (27) and (28), applicable when With these values, the transmissibility for
bottoming has occurred, will be valid for ve- impact speeds lower than those indicated above
locities producing do > d,. From Eqs. (19) will be
and (24), the velocity corresponding to these
deflecions will be 0212
T =11.91
2
0.70. w
d s 5120,. 2 1] I 20.2 U..
YA > 1.05 >
- > 4.73 mph 1 22 +] 2 x0.35 x 2
- .1 11.9 11 (29) A,

and when 10 percent damping is included, The accelerations to the inner container re- 0

d sulting from this transmission are plotted in

[
VA 0.90- 5.5 mph. Fig. 1.
The values from Eq. (27) are shown in Impact With Bottoming of Coupler - SinceR
Fig. 2. during this impact, low and high frequency ac-
celerations as obtained from Eqs. (22) or (23)
and (27) or (28), respectively, are present at b,
Superimposed on Fig. 2 are the results of ''2 r4
several investigators (3,4,7,9), shown together the outer container and the transmissibility ist
with the envelope of maximum acceleration de- a function of the forcing frequency, the high and
data of separate low frequencies during this impact have to be T
tests (7). by combining.the
termined
found. The frequencies are indicated in Fig. 3. S]
The equations used are Eqs. (11) and (12),
= 3 fl
The wide range of results in these tests which, with
lb/in., Y103 lb/in,for
1310substitution
KB = the o = 35x10
andK 0&o 20.2, t,
can be explained if one considers the effects of B
car woights and design, the type of friction reduce to
gear, the status of maintenance, and the in-
strumentation and other factors on the response 20.2 1634-20. 1041 tan" 2. 67 x 10.2
X0. 0
of a car to impact. Comparison of the results oH d 2 (0
of this analysis with the test data indicates that 25
these results fall wiihin expected results and
represent an average value of possible results. and
In view of this, the use of an empirical coeffi- 20.2 0.637 sin5 2.2 (31)
cient to account for the braking of the cars was WL do (H3

54

- T~~T7'
 7'
P:; .7. Id. f.

S- - - -

8
2,

"' 1 -.-.
'.. "i

S 0 F4 28 C 32
ad o 36 40 44 48 5

0 24 28c '32car--36 'pcse 40 '44 48 2

fq
LOW FREQUENCY,W,,L (RAD/SEC) I , iL
128 " 136 144 152 160 168 176 18
HIGH FREQUENCY,.dH (RAD/SC) :

Fig. 3. Car and outer container !

frequencies vs car impact speed

Having obtained WL and OH as in Fig. 3, the damping on the magnitude of the resultant ac-
transmissibilities are obtained by using Eq. (13) celeration at the inner container is also shown.
with the design parameters of the container and Superimposed as data points on these curves
with the forcing frequency ()-- w. and &)L. Also are the results of impact tests performed by
the relative transmitted accderations- AH and the container designer. The comparison indi-
AL and the phase angles are obtained from Eq. cates that the calculated values agree with the
(15). The resultant accelerations at the inner upper bound values obtained from the tests.
container obtained from Eq. (14) are shown The comparison also suggests that a 10 percent
plotted in Fig. 1. viscous damping at the coupler would have rep-
resented the condition of the test container car
coupler.
RESULTS AND CONCLUSIONS
The following conclusions can be drawn
Figure 2 shows the acceleration loads felt from the analysis:
by the car bed and by the outer container 'at va-
rious impact speeds. The graph shows also the
test results reported in the pertinent literature. 1. The accelerations to be expected at the
The discontinuity in the curves indicates the inner container for various impact speeds can
speed at which the coupler bottoms. This be predicted by this type of analysis. A 10 per-
change in the slope of the acceleration is con- cent viscous damping should be taken as a rep-
firmed by test results reported in the litera- resentative value of the damping provided by a
ture (6,8). railroad car coupler.

Figure 1 shows the accelerations at the 2. Impact speeds higher than about 7.5 mph
inner container resulting from the attenuation will produce inner container accelerations
of the accelerations of Fig. 2 through the con- greater than 1.8 g and car bed acceleration
tainer suspeni'o. The effect of the coupler above 20 g.

55

- ,.----. -?
 I
REFERENCES

1. "Fundamentals of Guided Missile Packag- 6. J. M. Roehm, "Dynamic Testing of Freight

ing," Naval Research Laboratory, Depart- Cars," ASME Paper, June 15, 1962 t
ment of Defense, Ch. 6, p. 5
2. R. D. Mindlin, "Dynamics of Package Cush- 7. J. H. Farrow, "Shock and Vibration Trans-
ioning," Bell System Technical Journal, portation Environmental Criteria," Memo-
Vol. 24, Jul.-Oct. (1945) randum to NASA Environmental Data Sub-
c mt
3. R. Eustace, "Railroad Humping Testing on
Units of the Pershing Weapon System," In-
stitute of Environmental Science, Proceed- 8. R. E. Barbiere and W. Hall, "Electronic
ings, 1962 Designer's Shock and Vibration Guide for
Airborne Applications," WADC Tech. Rept.
4. C.M. Harris and C. E. Crede, Shock and 58-363-ASTIA AD 20-1095
Vibration Handbook (McGraw-Hill, New
York), Vol. 3, Ch. 3, 1961 9. E. J. Kirchman, "Rough Road and Railroad

5. L. K. Sillcox, Mastering Momentum Humping Spec. Testing," Environment

(Simmons-Boardman Publishing Corp., Quarterly, Vol. 4, No. 3, Third Quarter
New York), 1955 (1958)

56
 4--

PRELIMINARY ANALYSIS OF DATA OBTAINED IN

THE JOINT ARMY/AEC/SANDIA
TEST OF TRUCK TRANSPORT ENVIRONMENT*
J. T. Foley
Sandia Corporation
Albuquerque, New Mexico

This paper describes the manner in which the data obtained during the
joint Army/AEC/Sandia test of truck transport environment was evalu-
ated to arrive at the "best means" for reduction, presentation, and
analysis. Four systems of data reduction and presentation were used .
on experimental portions of the tiuck test data, and the resultant out-
puts of each reduction system were evaluated to arrive at the best
means of describing the environment.

INTRODUCTION Model A737C-4-350 and A46C-15-350 accelerom-

eters. A block diagram of the system is shown in
The truck transportation environment test Fig. 5; pickup locations are shown in Fig. 6.
referred to in this paper was a joint effort to
obtain measurements of the transportation dy-
namic environment. Participants in this effort AMOUNT OF DATA
were the US Army Transportation Engineering
Agency (USTEA), Ft. Eustis, Virginia; the AEC Data were recorded continuously throughout
Division of Reactor Development, Germantown, the trip with the exception of short intervals of
Maryland; and Sandia Corporation's Environ- time between Ft. Eustis and Wilmington and dur-
mental Operations Analysis group, Albuquerque, ing about 15 minutes of travel in Texas. This
New Mexico. yielded thirteen 2500-ft. rolls of recorded mag-
netic tape data covering 55 hours of moving
The test vehicle was a flatbed tractor- truck transportation environment. Voice com-
trailer on which was mounted a 15-ton radioac- mentary, recorded on the average of every 5
tive materials cask (Figs. 1 and 2). The truck minutes throughout the trip, describe road types
and cask were instrdmented to measure inputs encountered, weather conditions, speed, time of
to and responses of the 15-ton load. day, and correlation information concerning
parallel instrumentation being operated by
Data were obtained from the unloaded USTEA personnel.
truck during travel from Ft. Eustis, Virginia,
to Wilmington, Delaware, and from the loaded
truck during travel from Wilmiington, Delaware, PRELIMINARY ANALYSIS OF DATA
to Albuquerque, New Mexico (Fig. 3). Upon
arrival at Albuquerque, "apparent weight" The ultimate repository for the data obtained
measurements of the truck, with and without by Sandia in this joint test is the Sandia Corpora-
the load, were made (Fig. 4). tion Environmental Data Bank. The users of
this bank are primarily designers, dynamicists,
The Sandia instrumentation system used and laboratory test engineers. They use such
an Ampex Corporation Model CP-100, 14-chan- data to determine what to design against and what
nel tape recorder to record the outputs of to test to with respect to the expected environ-
Endevco Corporation Model 2235C, Statham ment. In addition, the Environmental Operations

*This work was supported by the United States Atomic Energy Commission.

57
 Fig. 1. Semitractor and trailer used for test (1)

_A_

"44"

Fig. 2. Loaded and instrumented truck (1)

Analysis group, which operates the data bank, method would provide data from these truck
will use the results of this and other similar tests that could best meet these requirements.
tests to derive standard transportation en- The four methods evaluated were:
vironmental test levels for Sandia Corporation
products. 1. Oscillograph records;

The results of a user survey of environ- 2. Power spectral density versus frequency
mental data requirements within Sandia Cor- plots;
poration about three years ago indicated that
environmental data ideally should meet the fol- 3. Cumulative instantaneous amplitude dis-
lowing four criteria: tribution plots; and

1. Data should be statistical; 4. Peak amplitude distribution versus fre-

quency plots.
2. Data should be graphical;

3. Data should present those parameters The data used for illustrating the evaluations
that can be measured in the test laboratory as in this paper were taken from the truck test
well as in the field; and under the following conditions:

4. Data should lend themselves to various 1. Concrete highway (high speed, unloaded);
means of "combination" by numerical or graphi-
cal means; or both. 2. Blacktop highway (low speed, unloaded);

In an attempt to fulfill these requirements,

four methods of reducing and presenting vibra-
3. Concrete highway (low speed, loaded); and L
tion data were evaluated to determine which 4. Blacktop highway (high speed, loaded).

58
0-.. "34 -

14 . -

41 - 4. 5.

0 "4
* 4 $4~ ., =,

0,0

4J4-.

4J 0
oco0 4)

-Z 4j 41 0

ccU) 0

4 $4 0
 1%

Fig. 4. Test setup for measuring system

"apparent weight" (1)

The pickup location illustrated in this amplitude magnitudes encountered. This pre-
paper is in all cases the "input to load" meas- sentation is illustrated in Fig. 8. I
ured on the truck bed to the rear of the cask lo-
cation. It was considered, therefore, that as a
"quick-look" method of reduction/presentation,
the oscillograph traces and bandpass analyses
OSCILLOGRAPH RECORDS were useful tools but would not be an adequate
means of final presentation. Any tabulations of
Figure 7 shows representative traces ob- amplitude values as hand-read from such
tained from the input location near the load. records would not be accurate enough or pro-
These representations are graphical and present vide a sufficient description of the environ-
parameters (amplitudes and times) which can ment.
be measured in the laboratory and in the field.
They are neither statistical in nature, nor do
they lend themselves to combination by numer- POWER SPECTRAL DENSITY VERSUS
ical or graphical means. By choosing a faster FREQUENCY PLOTS
playback rate (expanding.the record), some
estimates of frequency could be made as well Data samples were then reduced and pre-
as measuring amplitude; however, these would sented in power spectral density (PSD) form
not be only time-consuming and tedious but (Figs. 9 through 12). Since the data in the test
would be subject to human error. were obtained continuously, it was possible to
use a tape loop for PSD analysis that was of
An alternative method of data presentation long duration. The filter bandwidth selected for
by oscillograph trace was also considered. good amplitude resolution was 5 cps; the mini-
This method consists of playing back the data mum time length of record using this bandwidth
through fixed bandpass filters. This presenta- to approach stationarity was estimated to be 20
tion "separates" the amplitudes encountered seconds. The time length of record used
into regions or bands of frequency. This was initially for loop analysis was approximately 300
considered an improvement over use of simple seconds.
amplitude/time presentations, but still would
require a considerable amount of hand reduc- It immediately became apparent that a prob-
tion and interpretation with respect to the lem existed in using a PSD presentation format.

60

- - -.-
 Endevco Corp. Remote
Model 2646M1 Control
Endevco Accelerometer Charge Amplifiers
Model 2235C,-

. 8 Identical Channels ----

To Tape Recorder Channels T.R.
3, 6, 7, 8, 9, 1O, 11 Channel
Number

Statham, Inc.
Model CA9-10 12
Strain Gage
Statham Accelerometer Amplifier
Model A73TC-4-350O

r Statham, Inc.
I , Model CA9-10 1
Strain Gage
Statham Accelerometeri Amplifier /
Molel A46C-15-350
I Ft. Eustis Pro ert_
Statham, Inc.
Model CA9-0 5 Ampex Corp.
Strain Gage Model CP-100
Stathan Accelerometer Amplifier 14-Channel
Stathmdeleromet2Tape Recorder,
Model ASA-12-120 2 Speed: 1 7/8 ips,
Microphone Input Response: dc to
1o n625 cps, 2500
Ft. Reels

Shorted Reference 4

Standby Reserve 2 1 1
Voice Monitor .. ..

E_ CEC Amplifiers
"< Model 1-113B
Remote >CEC Oscillograph r

Control ! 1
Model"'5-l -PJ ~ CEC Strain
~~~Gage
Amplifiers
D
Strain Bridge
_tten()l
ttw (K
SSystem
Eustis Property__ _

Fig. 5. Block diagram of transportation test instrumentation (1)

The root-mean-square (rms) signal from the conditions indicated that not only amplitude mag-
record of Condition A varied so that it was im- nitudes changed, but that the shape of the power
possible to obtain a PSD plot over this time spectrum changed with respect to truck speed,
length of record. It was necessary to use one- load, and road condition. Although the PSD for-
third of this tape loop length (160 seconds) to mat gave a good graphical data description and
obtain a PSD versus frequency plot. It appeared the parameters could be reproduced in the test
that we would be involved in analysis of a non- laboratory, a disadvantage was that, statistically,
stationary process during some road/load con- any one of the PSD analyses might not be character-
ditions. istically a reasonable picture of the environment.
Subsequent comparisons of PSD plots ob- An alternative considered was to make up a
tained under varying truck speed, road, and load composite tape composed of samples from many

61
T.R. Order Transducer Input tplif ier (
Chnnel of Serial Sensitive Reference

1 7 1 A 2235C A-68 27.71 p-q/P-g 180 p-qp~p Vertical 4 Cok trunnion, driver's side (J)

2 - . - Voice logging channel

3 5 1 A 2235C PA-83 29.46 P-q/p-g 180 p-qp-p Vertical 5 Cask trunnion, right side (C) $
** 9 2 B A5A-12-120 1258 16.8 Tsrv/v/g 3.43 9 Longitudinal B-1 Truck bed over rear axle (N)
5 ~0 8 2 B A5A.12-120 1658 22.5 pv/v/g !2.88 g Vertical B-3 Inside T.R. box at right rear mounting bolt(S)
6 1 1 A 2235C FA-24 29.73 p-q/p-g 180 p-q,..p Longitudinal 6 Truck bed aver 5th wheel (3)
7 3 1 A 2235C FA-21 27.03 p-q/p-g 180 p-qp.p Vertical 7 Track bed cross-member 2 feet forvard of cask(R)
8 5 1 A 2235C FA-62 28.88 p-q/p-g 180 p-qp 1 p Longitudinal 8 Cask trunnion, left (driver's) aide (D)
9 1 A 2235C FA-18 29.29 p-q/p-g 180 p-qp.p Vertical 3 Truck bed cross-member 2 feet aft of cask (p)
10 7 1 A 2235C F3-28 29.49 p-q/p'g 180 p-gp.p Longitudinal 10 Cask trunnion, right aide (3)
11 1 1 A 2235 C EA-52 27.88 p-(dp-g 180 p-9p.p Vertical 2 -Truck bed over 5th wheel (J) 7

12 6 2 B A73TC--350 178 118.7 puv/v/g -0.918 g Transverse B-3 Top geoetrical center of the cask (A)
13 2 2 B A -15-35o 624 Unknown !2.97 g Vertical B-2 Truck bed over 5th vheel; transducer
supplied by Ft. Eustis WL
14 Shorted reference channel

- 1 - Endevco Corporation 5* - Removed at Wlmington, Delavwae due to excessive noise ond crosstalk to other strain
Pasadena, Calif. gage transducer systems; probably due to excessive input cable length.
2 - Statham Instruments, Icc. "4- Used to monitor the Ft. Dstla shock isolation mount from Ft. Eustis, Virginia to
Ics Angeles 64, Calif. Wilmington, Delavare; removed at Knapp .is Corporation, channel held in reserve
for balance of trip.
- A
A * arium TLtanate Crystal *4555 Se Figur. 3 for esmet transuacer locations by letter.
B - Strain Gage

Fig. 6. Magnetic tape directory (1)

CONDITION A encountered and then derive a new composite,

UNLOADED
TRUCK TRAVEL ON CONCRETE
HIGHWAY AT HIGH SPEED there would be no way to do this from the data
3G- as presented. He would have to obtain the
original data tapes and reprocess them to 6b-
0- tain the new result. The feeling, then, was that
users would find that the PSD format, although
offering many good advantages, would be un- '.

t-----satisfactory for alternative analyses. Further,

due to the indications of nonstationarity en-
countered, it was felt that PSD pzesentations of
CONDITION B some individual conditions could be misleading. -

UNLOADEDTRUCK TRAVEL ON BLACKTOP

HIGHWAY AT LOW SPEED

3G CUMIULATIVE INSTANTANEOUS AMPLI-

TUDE DISTRIBUTION PLOTS

Figure 13 illustrates the third means of

t analysis and presentation which was investigated.
This system presents the data on the ordinate in
Fig. 7. Examples of terms of its time-cumulative instantaneous am-
oscillograph traces plitude density in a selected bandwidth. Figure
13 presents "A" conditions using a 5-cps band-
width. The abscissa is the variation of the sig-
section3 of the continuous recording, i.e., to nal from a normalized reference (rms) (a/u)
analyze an ensemble of records. This could be value. The solid line superimposed on this plot
done, and a composite PSD versus frequency is the result of a Gaussian random (white noise)
plot obtained that contained contributions of all signal input to this reduction system, to illus-
road types, speeds, and/or the variation in trate the deviation of the measured data from
load to give a composite summary description the theoretical statistical normal distribution. !
of the environment which would be statistically V
appropriate. The major disadvantage to this This system of analysis/presentation ap- ,2
alternative approach was that the contribution
of individual characteristics such as concrete
road versus blacktop would be lost in such a
proached the criteria for "good" presentation,
in that it is both statistical and graphical in
nature. But it offers difficulties in that the
i
composite. If a future user of the data wished parameters are not normally those controlled
to exclude, for his purposes, some conditions in the laboratory or measured in the field.

62
 FILTER
BANDWIDTH
CPS

180- 240
(2Z-1/2 - 30) -',.--%-'-%-

120- 180
(15- 22-112) ' . v -' %

80-120
(lO"15) " ,
_ 0.1

COMPOSITE
SIGNAL

40-80
(5- 10)

20-40
(2-112 -5)

0-20 --------
(0- 2-1/2)

1. FIRST LISTING OFFILTER BANDWIDTH IS REAL TIME PLAYBACK

NOTES:
2. BRACKETED BANDWIDTH @8 x REAL
VALUE IS "EFFECTIVE'
TIME PLAYBACK
3. THIS RECORD IS A PLAYBACK @8 x REAL TIME OFTRUCK
TRAVEL OVER A ROUGH CONCRETEHIGHWAY @50 MPH

Fig. 8. Bandpass oscillograph playback

.10
SPEED: 50 MPH
LEVEL CONCRETE WITH EXPANSION JOINTS
.05 INPUT ALONG THEVERTICAL AXIS AT 50"F
.04 ACCELEROMETER'LOCATION: 24 INCHES TO
.03 REAR OF RAM
CASK LOCATION
.02 LOOP LENGTH: 16 SECONDS

L:. .01

00

;; .004
z
z .003
~.002
~ 001

.0003
.0002

2 34 5 10 20 30- 40 50 100 200 400

(CPS)
FREQUENCY

Fig. 9. Vibrationtestof tractor-trailer, unloaded

Further, the combination of individual incre- records to perform additional analyses. This
ments of data from the truck test would pose data presentation, however, gives further in-
similar problems as those which would be en- formation concerning the nature of the amplitude
countered in obtaining a composite PSD plot; distribution. Fi'om these data plots, non-Gaussian
i.e., one would have to obtain the original tape distributed phenomena were inferred. It was

63 I
~l
 .10

SPEED: 35 MPH
.05 LEVEL BLACKTOP
.04 INPUT ALONG VERTICAL AXIS AT 501
.03 ACCELEROMETER LOCATION: 24 INCHES
TOREAR OF
.02 RAM CASK
LOCATION
LOOP LENGTH: 300 SECONDS
C- .01

.00

.004 -
.003
-j

.001

.0004
.0003
.0002

. 1 I I I
2 345. 10 20 30 40 50 100 200 400
FREQUENCY
(CPS)
Fig. 10. Vibration test of tractor-
trailer, unloaded

1.00

SPEED: 35 MPH
.50 CONCRETE HIGHWAY
.40 INPUT ALONG THEVERTICAL AXIS AT 401
.30 ACCELEROMETER LOCATION: 24 INCHES TO
REAR OFRAM
.20 CASK LOCATION
LOOP LENGTH: 300 SECONDS
.10

.05
In.04
z 0

L .01

00

0.or
1 2 3 4 5 10 20 30 40 50 100 200
FREQUENCY
(CPS)

Fig. 11. Vibration test of tractor-

trailer, loaded

6i
 1.00
fre
.50 pri
.40 SPEED:50 MPH syl
LEVELBLACKTOP
INPUT ALONG THEVERTICAL AX!S
j
AT 40F
.20 ACCELEROMETER LOCATION: 24 INCHES TO W-
REAROF RAM as
VI CASK LOCATION ---
C .10 300 SECONDS
LOOP LENGTH: tio
es!
ac,
.05
S.03 tei
<~ .02

a- sa
C .01

.005

.002 A.
.001
1 2 3 4 5 10 20 304050 100 20
(CPS)
FREQUENCY

Fig. 1Z. Vibration test of tractor-

trailer, loaded

%0 - DATASIGNAL

80%

60% -

) SPEED 50 MPH, LEVELCONCRETE

VERTICAL AXIS "WHITE NOISE'
40% - LOCATION: 24 INCHES TO REAROF
a" RAM CASK LOCATION
FILTER BANDWIDTH: 5 CPS
TIME LENGTH OF SAMPLE: 160 SECONDS
20%

"3a "20 1la Do 1a 2a 3a

Fig. 13. Vibration test of tractor-

trailer, unloaded

deduced from the data characteristics exhibited PEAK AMPLITUDE PROBABILITY DIS-
that the environment over most roads consisted TRIBUTION VERSUS FREQUENCY
of a low-level complex vibration on which were
superimposed a great number of repetitive A fourth system of data presentation,
shock pulses. peak amplitude probability distribution versus

65
frequency, was then employed. The format of nature, yet graphical in format; the parameters
presentation is shown in Figs. 14 and 15. This
system is an automated extension of the fixed
bandpass filter oscillograph record system that
measured (g versus frequency) could be
measured in the test laboratory as well as in
the field, and the format of the data lends itself
L ;

was illustrated in Fig. 8. The data are presented to a number of means of comparison by numeri-
as the number and magnitude of the accelera- cal or graphical methods.
tion peaks encountered in each pass bard. In As one may see in the illustrated format of
essence, then, it is a probability distribution of Fig. 14, however, the bandwidth is rather wide
acceleration peaks, the total number of peaks for the purpose of describing the environment
encountered, and the number of peaks encoun- encountered in truck transportation. Our prior
tered in each bandpass so measured. evaluations, as well as this data presentation,
This analysis/presentation system best indicate that the area of interest lies primarily .4
satisfied requirements, for it was statistical in in the 0-20-cps band. Further, it would be

Vibration,File( 1),Cour.t( 352260) Tape 72430053

(ACCELERATION DISTRIBUTION IN PERCENT*)

TIM 9
ACCELERATION, O-PK G TRUCK BED CROSS-MBSER
140.0 2 FEET TO REAR OF CASK
CONCRETE HIGHWAY, HIGH
00.SPEED, UNLOADED
72.0

52.0
37.0 !
27.0

19.0

14.0

10.0 i
7.2

5.2
3.7
3. 1.32 0.00 0.01
2.7

1.9
0.43 0.02 0.01 h
0.22 0.01 0.02 0.01 0.00

0.22 0.03 0.01 0.01 0.01 0.00

1.0 84.82 99.94 99.95 99.98 99.99 100.00 99.99

FREQUENCY(CPS) ..... .

0- 20- 40- 80- 120- 18o- 240-

20 4o 80 120 180 24o 350
SAMPLES PER CHANNEL
461 12567 23493 39386 45522- 89546 141285
VALUES IN THE TABLE SHOW THE PERCENTAGE OF ACCELERATIONS AT THE CORRESPONDING MAGNITUDE

Fig. 14. Vibran data presentation

66
desirable to measure at least one more ampli- A third modification indicated on this format
tude increment between 0 and 1 g. is presently being promulgated. This is the in-
clusion of the rms value obtained in each pass
We are currently investigating two pro- band (and/or its range of variation). With this
cedures to improve this format. The first is value included, the amplitude probability distri-
to obtain peak counts at the 0.5-g level. The bution obtained in each pass band can be normal-
second is to drive the data tape into the fixed ized and then tested to determine how closely it
filter system at a multiple of real time to pro- approaches the Rayleigh distribution of acceler-
duce effectively a narrow band analysis. Our ation peaks predicted for narrow band random
goal, then, is to obtain the final format in the vibration. With the addition of this modification,
form as illustrated in Fig. 15. we feel that the environment so described will

Vibration, File( ), Count Tape

(ACCELERATION DISTRIBUTION IN PERCENT*)

Acceleration, O-Pk G
140.O
100.0
72.0
52.0
37.0
27.0
19.0
14.o
10.*0
7.2
5.2
3.7
2.7
1.9
1.4
1.0
.5

FREQUENCY(C S)

0 2 5 10 15 22-i
2- 5 10 15 22- 30

RAS G

High
Low

SAMPLES PER CHANNEL

*Values in the table show the percentage of accelerations

at the corresponding magnitude

Fig. 15. Modified vibran

67
Lble to measure at least one more ampli- A third modification indicated on this format
acrement between 0 and 1 g. is presently being promulgated. This is the in-
clusion of the rms value obtained in each pass
t
e are currently investigating two pro- band (and/or its range of variation). With this
es to improve this format. The first is value included, the amplitude probability distri-
ain peak counts at the 0.5-g level. The bution obtained in each pass band can be normal-
I is to drive the data tape into the fixed ized and then tested to determine how closely it
system at a multiple of real time to pro- approaches the Rayleigh distribution of acceler-
ffectively a narrow band analysis. Our ation peaks predicted for narrow band random
then, is to obtain the final format in the vibration. With the addition of this modification,
as illustrated in Fig. 15. we feel that the environment so described will

Vibration, File( ), Count ( ) Tape

(ACCELERATION DISTRIBUTION IN PERCENT*)

Acceleration, O-Pk G
140.0
100.0
72.0
52.0
37.0
27.0
19.0
14.0
10.0
7.2
5.2
3.7
2.7
1.9
1.4
1.0
.5

FREQUENCY(C PS)
0 2 5 10 15 221j
2f 5 10 15 22 30

RKS G
High
Low

SAMPLES PER CHANNEL

*Values in the table show the percentage of accelerations

at the corresponding magnitude

Fig. 15. Modified vibran

67
 t

offer a direct means of determining the ran- vibration tests or to estimate the probable force
domness of the amplitude distribution of the en- magnitudes being delivered to the cargo. Since
vironment sample - a feature not presently acceleration amplitudes will be in terms of
directly available from other data presentation probability of occurrence, it permits the use of
formats. analytical techniques to determine what ampli-
tude values should be selected as being ap-
propriate in each pass band for a vibration
CONSTRUCTION OF A COMPOSITE test, and what amplitude values may be more
DESCRIPTION OF ENVIRONMENT appropriately accounted for in shock tests.

As described to this point, we arrived at a 2. Within each of these two load condition
method of obtaining and describing a statistical composites, the number of samples taken will
sample of the dynamic environment in terms of be based on the frequency of occurrence of each
its peak acceleration probability density versus road type as encountered during the trip and the
frequency. The determination of the length of frequency of occurrence of road speeds achieved.
time during which to take this sample is now These frequencies of occurrence are presently
not necessarily the length of time estimated to being obtained from the voice transcript and
approach stationarity as with PSD-type analy- -activity logs recorded throughout the trip.
ses; it is more appropriately a sufficient period
of sampling time to obtain a statistically repre-
sentative number of peaks counted in a band- SUMMARY
pass of interest.
It has been determined that the environment
For a general guideline, we have adopted description that will be provided as a result of
the rule that at least 30 peaks must be counted this truck test will be composed of:
in a given pass band for the sample distribution
to be considered representative. Thirty peaks 1. Individual acceleration amplitude proba-
counted provides 99 percent confidence that the bility distribution versus frequency presentationm
maximum deviation of a measured g value from representative of the various road types and
the true population distribution is 29 percent. speeds encountered for the (a) unloaded and (b)
By way of illustration as to the degree of confi- loaded truck;
dence to be placed on measurements where
greater numbers of peaks are counted, a band- 2. A composite amplitude probability dis-
pass where 10,000 peaks were counted provides tribution for the two load conditions based on
99 percent confidence that the maximum devia- probability of occurrene of road type and ye-
tion of a measured g value from the true popula- hicle speed; and
tion distribution is 0.16 percent. This statisti-
cal probability description is nonparametric- 3. Plots of apparent weight of the loaded
i.e., it does not presuppose a normal (Gaussian) and unloaded truck versus frequency.
distribution, or any other. Preliminary sam-
ples taken over a 60-second real time period
appear to more than fulfill the 30-count mini- It is believed that the method of environ-
mum requirement for the truck transportation mental measurement and description outlined
environment data. here is one which, at the present state-of-the-
art of environmental analysis, will be most useful
Our next problem was how to relate, in a for customers of the Sandia Corporation En-
composite picture of the environment the ef- vironmenta Data Bank. We hope to extend this I
fects of road type, truck speed, and load. This approach of combining measurement and activity
is to be done in the following manner: analysis to other types of truck, train, aircraft,
and missile applications, and perhaps to apply
1. A composite peak amplitude probability this method of analysis to direct force measure-
density will be obtained for (a) the unloaded and ments obtained during operational activities.
(b) the loaded truck. These composites used in
conjunction with the appropriate "apparent It is further believed that replications of
weight" measurements which were obtained at this method in other tests provide a consistent
these two load conditions can be used to obtain basis for comparison and/or combination of
estimates of the input force versus frequency environmental measurements in a manner I
conditions. These estimates can be used for superior to that of just "enveloping" data
derivation of force-controlled laboratory taken from a number of tests.

68

-- - - - ~ - - - - - - 7- 7 -
 BIBLIOGRAPHY

Bendat, J. S., "Random Process Theory and Ap- Piersol, Allan, "The Measurement and Interpre- -

plication," Measurement Analysis Corpora- tation of Ordinary Power Spectra for Vi-
tion, Los Angeles, Cal., May 1964 bration Problems," Measurement Analysis
Harris and Crede, Shock and Vibration Hand- Corporation, Los Angeles, Cal., June 1953
book, Vol. 2 (McGraw-Hill, New York), Sandia Corporation Vibration Analysis Commit-
1961, pp. 22-1 to 22-26 tee, "Vibration Analysis for Weapon Devel-
Mortley, J. L., "Joint Army/AEC/Sandia Test opment Test Programs," Sandia Corpora-
of Truck Transport Environment, Decem- tion Report SC-DR-321-61, April 1962
ber 7-17, 1964 (Test No. T-10767)," Sandia Sandia Corporation Environmental Operations
Corporation Report SC-DR-65-278, August Analysis Group, "Weapon Development Test
1965 Data," Sandia Laboratory Engineering
Otts, J. V., "Impedance Measurement of a Flat- Manual Procedure S-49, August 6, 1965
bed Truck," Sandia Corporation Test Re- Thrall, G. P., "An Analysis of Amplitude Prob-
port No. T-10768, March 31, 1965 ability Measurements," Measurement
Otts, J. V., "Force-Controlled Vibration Test- Analysis Corporation, Technical Documen-
ing," Sandia Corporation Report SC-TM-31, tary Report No. FDL-7DR-64-116, March
February 1965 1965

DISCUSSION

Voice (TRW Systems): What kind of accel- at these various levels. The levels shown started
erometers did you use? at 1 g, then went to about 1.2, 1.4, 1.8 and so on.
Our modification Will be to get another count be-
Mr. Foley: There were quite a number of low 1 g.
them; primarily we used unbonded strain gage
accelerometers in parallel with piezoelectric Mr. Zell: Then they were not in terms of a
accelerometers. In.some spots strain gages percentage of occurrences?
were used as backup for cross-monitoring.
Mr. Foley: Yes, they were. They were the
Mr. Zell (Picatinny Arsenal): What is the actual counts of peaks, and in terms of percentage
procedure to get access to the data bank? at each of these levels.

Mr. Foley: Just write and ask us. We will Mr. Rhodes (Endevco Corp.): What was the
be glad to send anything in the way of unclassi- full-scale acceleration, generally, in the raw
fied data. data channels ?
Mr. Zell: Were the bandwidths variable Mr. Foley: We set up full scale as plus or
with the frequency in this vibran format which minu jgiTe data in the .pperindicate this
you showed? at the low frequencies, from 0 to 20 cps, in the
-Mr. Foley: In the system the bandwidth is one vibran format shown. Eighty-five percent
fixed in the sense that the bandpass filters are of the peaks counted in a data sample are below
fixed. We are, however, using playback speeds 1 g. Then there is a gap with no peaks up to 3 g.
faster than real time to prod67e effectively a It gives a pretty good indication of what might be
narrower analysis with the system. This was called steady state versus transient data super-
found to be necessary, particularly in this trans- imposed on a low level steady state.
H] portation situation, since the great majority of
the acceleration peaks encountered were below Mr. Edgington (White Sands Missile Range):
20 cycles. We felt we had to get narrow in the Did the data on both the loaded truck and the un-
low frequency region to take a better look at it. loaded truck show 3 g as a composite waveform?

Mr. Zell: What were the statistical levels Mr, Foley: Yes. The 3 g showed as the
used in the presentation in each frequency band- peaks and at about the same place. The inter-
width? esting thing is in the apparent weight measure-
ments of the vehicle itself between the loaded
Mr. Foley: The statistical levels are not and the unloaded situation. The apparent weight
preset. An actual counting was made of the peaks plots indicate that the truck is effectively

69
IiI

J
j decoupled from the load above 10 cps. There is Mr. Foley: Well that is difficult to answer.
not a great deal of difference between the lo~ided On the basis of the vibran, the duration would
and the unloaded truck in terms of forces de- be equivalent to something between 0 and 20 cps.
livered in the range of 10 to about 200 cps, Fifteen percent of the counts down in this area
where most of the vibration tests are operating. were at 3 g and the rest of them were about 1 g.
Mr. Levin (Bureau of Ships): What was the
duration of the -impulses at the 3-g level com-
pared with that at the lower g levels?

7i

70
 ROUGHNESS MEASUREMENT AND SYSTEM RESPONSE
EVALUATION FOR HIGHWAY ENVIRONMENT
j. R. Harvey and R. A. Wursche
Goodyear Aerospace Corporation
Litchfield Park, Arizona

The paper presents analytical procedures to evaluate the shock and vi-
bration environment of highway mobile transport vehicles. The road
roughness measurement technique was used to obtain the road profiles
and power spectral density estimates of a large number of road sam-
ples in West Germany and Central Arizona. The generated roughness
data were used as input to a representative highway mobile weapons
system to obtain the statistically average and peak system response.
The calculated response data are then summarized to predict the dy-
namic environment of the payload and the transport vehicle.

INTRODUCTION 1. The actual road profiles in the area of

planned deployment were obtained and using a
In October 1962, Goodyear Aerospace Cor- transient response mathematical model of the
poration began an intensive program to develop vehicle, maximum payload response was calcu-
a technique for analyzing the ride dynamics of lated.
a complex vehicle traveling over a road with a
known profile. The objective of this program 2. The displacement power spectral den-
was to provide a design tool for optimizing the sity (PSD), computed from the road profiles,
suspension system of the vehicle before its was used as a forcing function to a harmonic
final fabrication and test. response mathematical mod9l of the vehicle to
calculate the displacement and acceleration
The problem was to provide both an accept- rms response.
able maximum and root-mean-square (rms) ac-
celeration environment to the payload and its Figure 1 is a simplified block diagram of
associated equipments. the suspension system analytical techniques.
The problem was approached using two Road profiles measured in West Germany
techniques: and in Arizona have been used to calculate the

INPUT MATH MODEL OUTPUT

STABLE PROFILE TRANSIENT DISPILACEMENT
RESPONSE ACCELERATIONS

MATH MODEL O
PSD INPUT (HARMONIC RESPONSE) DISPLACEMENT DENSITY
DISPLACEMENT DENSITY TRANSFER FUNCTIONS ACCELERATION DENSITY
RMS RESPONSE

Fig. I. Simplified diagram of suspension system analytical techniques t

71

-- ,... . . -.. .. . . . . . . . - . .
 , A

I
.1
acceleration environment envelopes presented
in this paper.

This-paper presents, in
an abbreviated
generated by interrupting a relay located on the
odometer assembly. The distance between
voltage rises is 5.01 in. *'"
form, the total method for analyzing the ride A third channel was used to record the
dynamics from the road surface to the calcu- audio description of the road sample being
lated accelerations on the payload. measured. The recording equipment used in .

the survey is shown adjacent to the profilometer

PROFILE MEASUREMENT instrument panel in Fig. 3. Tf
TECHNIQUES Approximately 200 samples, covering all

The instrument used for measuring the types of paved roads, were recorded in West
profile was of the slope variance type (1). It Germany, and 55 samples on Arizona roads.
was selected because it is very rugged and All of the sample lengths were 1000 ft.
easily transported.

The instrument consists of a trailing arm DATA REDUCTION OF PROFILES

assembly that extends, in the deployed position, The data reduction of the profiles was ac-
outward and in the right-hand track of the tow-
ing vehicle. A horizontal reference line is es- complished by first amplifying the raw data on
tablished by a gyro mounted on the upper portion the magnetic tape in an analog computer. It
of the trailing arm (Fig. 2). A synchronous was then fed into an analog-to-digital converter,
motor, located at the pivot point of the wheel where the slope voltage was digitized and re-
frame assembly, provides a voltage that is pro- corded at each voltage pulse from the odometer.
portional to the slope of the wheel frame with The digitized data are then fed into an IBM 7094 -

respect to the gyro-established horizontal ref- digital computer for profile calculations.
erence line. The voltage is recorded in analog A - i o nm
form on a 7-channel, FM tape recorder. Analog-to-digital conversion was accom-
plished by using a special purpose converter
A square-wave voltage trace is recorded connected directly to the IBM 7094 digital com-
on a second channel. This square wave is puter, as shown in Fig. 4. To facilitate

/k

TO WING VEHICLE ~ ~ ~
PIVOT SUPPORT

I I IN. DIAMTER WHEELSi Z;

FRIICHAME;" "
-SYN HRO ,

I t

Fig. 2. Profilomneter in displayed position

72
 -
ii
L

DEMODULATOR

Fig. 3. Profilometer instrument panel,

Arizona road roughness study

RAW DATA PACE

ON TAPE ANALOG
MAGNETIC DRIVE COMPUTER
TAPE

PROFILES IBM
AND 7094 - A-D CONVERTER
PSDIS

Fig. 4. Block diagram of data reduction procedure

73.-
analog-to-digital conversion and to eliminate characteristics. It was necessary, therefore,
unwanted and superfluous signals, all data were to remove the general profile characteristics
filtered prior to conversion to eliminate wave- from the data. Approximately half of the pro-
lengths less than 1.5 ft. files are linear, ard most of the remainder can
be effectively approximated by a linear form.
The "ramp effect" of linear profile trends can
TERRAIN PROFILE CALCULATIONS be removed by subtracting from the data a line
fitted to the data.
The rectangular coordinates of points along
the profile were calculated by the relationships: It is also desirable to prewhiten the data
before calculating the PSD's. A standard tech-
nique of proved adequacy for this type of
Yi S (sin 0) j data is
1=1

and an
y>'y= _ (3
Y i -Yi- 1 (3)

xi AS (cosO) , (2) Because of the method of integrating the

j=1 original slope data to obtain the profile, Eq. (1),
and in view of the small angles encountered,
where Eq. (3) can be taken as the terrain slope times
(S. The linear approximation to the profile can
(sin 0) = sine of terrain slope angle at effectively be made and removed by subtracting
increment j, the average slope from the slope data, i.e.,

As = distance along the terrain be-

tween odometer pulses (5.1051), y 1, n
Y, (4)
]
iYi -n Y"
x i = horizontal distance in inches, =
yj = elevation at x i in inches, and The PSD function of Y" differs from the
desired PSD of Y as a result of the integration
= the number of an individual required to obtain the profile from the slope
increment, and the factor of AS contained in Y". The cor-
responding Fourier transform produces a func-
As the profile was calculated, a tape was tion, wMS (2), where co is frequency in radians
prepared
profiles. which ultimately
This master tapecontained all the
can be used for per inch. The computed spectrum must conse-
quently be divided by this function to convert it
listing and plotting any or all the profiles. The to the spectrum of Y.
profile coordinates were moved within the IBM
7094 computer core to make way for the next
set of data as it came from the analog-to-digital A surveyed strip of highway 1000 ft long
converter. The discrete sampling of the con- was used to check the profilometer accuracy.
tinuous data signal (sine of the slope angle) was In the length of the course, the maximum devi-
performed at equal increments in space (along ation between the calculated and surveyed pro-
the terrain), which were represented by the file was 2.8 in. (Fig. 5).
odometer pulses. When the slope data were
converted to profile data, the horizontal incre- The PSD function associated with each pro-
ments were considered equivalent. The slope file (slope) is computed by the techniques of
angles are small enough in all tests to make Refs. 2 and 3. Each PSD is calculated from the
this approximation acceptable. normalized profile slopes defined by Eq. (4).
The AS associated with Eq. (4) is used to de-
termine the PSD estimates as functions of spa-
POWER SPECTRAL DENSITY tial frequency. The digital computer program
COMPUTATIONS is a modified SHARE Subroutine (No. 13236),
"CS TUKS" or "Tukey Spectrum Estimation."
PreliminaryThe computational procedures are defined below.
files and some of the corresponding PSD's in-
dicated that the power produced by general
terrain slopes was several orders of magni- The autocorrelation function was calcu-
tude greater than that for the roughness lated by

74

__________________
 N .V .111:Z. ..... . . ..... . . .-

DISTANCE (FEET) L - -. * L-. ..

....
M. 3 7 7.c-.! -t-

-j2.
I V
-. .l-
...... .: 1-- -

-4I

n .... ....................... ~
A~-

M.i I

~~7 _~4~-

I ~afiS ioO to 44 ao S. ~o2K. d~ ~Un.

ff~ Vlele M a~as. . ~a4M 32 4e ~l 4e 0. 2.64.
DITNC INHS)5

-r 7
 ~
1......~ - ~~~~~~~~~~~....
~ .*:..
4 1 :~~2d
.. 1~ *:i: " ~_

I
mm'

+Z

~:~ 4b1S O-A

41 k

_ ~

-i ' _-.-F i l
~._

7 .
17 7
.
....

w1
711 I. J..I
-Ai
4T17
1 -17 7

... 7:7.
-.
..

.. ~~4 ......
} - rI : 1 r

T::t *~
~ a.. ia~ a~ 0g. ~ ~O 7?o'?90Ir 171_
b

~~as ~ 7g. . m.
.O~~o m
4~ '4r5.
4140432 14
'i.4a 54 52. 4.. I.

rips Ar

-T-
 V U
A 12jT
_j 2.Q

'*1* +

I :Ii7
N: I

4 ::~:
IL

J -1

J; 4
 Ndensity = sq in./cycles/in.) is plotted versus
A(L) Yi-LY the spatial frequency (S2= cycles/in.). All PSD
((-= ) i--L+l samples fall within the envelope established by
the upper and lower boundaries of the plotted
n envelope.
n L) i-L )7 ,(5
i=L+1 i=L+1 The variation of the power content is more
pronounced in the low-frequency range (long-
where wave formation) which indicates the long-term
trends of the profile but does not describe its
L = 0, 1, 2,... 99, and actual roughness content.
n = total number of data points. The same envelope was also determined
for all Arizona road samples, resulting in a
The second term in Eq. (4) removes the mean. slightly lower power content which was caused
by the inclusion of some recently completed
Next, the PSD estimates were found by Interstate highway sections (5). A comparison
of the two road PSD envelopes with previously
X(K) =determined aircraft runway data (6) is shown Ln
X(K) = - 2e(L) cos A(L) + A(O) , (6) Fig. 7.
L-1
The presented summary envelope of hard
where surfaced profiles (Fig. 7) establishes the pre-
liminary displacement density input envelope
K = 0,1,2,... 99, that can be expected for road mobile systems.

8K = 1/2 for K=0 or K=m,

MATHEMATICAL MODELS
8K = 1 for all other values of K,
The generated road profile and PSD data
m = number of spectral estimates, and were then used as input to 'a highway mobile
transport vehicle to perform an'extensive sus-
2e(L) = 1 + cos (L//m). pension system analysis. The actual transporter-
launcher configuration shown in Fig. 8 was rep-
The latter function is termed a lag window resented by the dynamic model in Fig. 9,
by Tukey (3) and serves to concentrate the simulating the vertical plane motion of the
principal maximum or main lobe of the PSD rigid and elastic body modes of the dynamic
near zero frequency. X(K) is an estimate of system.
energy per unit interval of F centered on F= K,
for values of K equal to 1, 2,... 99, where F The mathematical models developed to de-
equals frequency -(cycles per inch) times m scribe the time-dependent and frequency-
times ns. dependent system response were formulated by
usi;-" Lagrange's equations of motion:
The number of spectral estimates was
chosen to be 100. Using an approximate num- d ZT ZT + +
50 degrees
ber of data points equal to 2500 gavepractical dt qr +
qr +qr (7)
a
of freedom. This was considered
minimum and would have been further reduced For oscillations about the equilibrium positions
if more spectral estimates had been taken. of the dynamic system, the kinetic energy (T)
Each PSD was stored on a master tape which is independent of the system displacements, and
finally contained all PSD's. This tape was used Eq. (.7) can be reduced as shown by Eq. (8):
for listing and plotting purposes, and also can
be used, like the profile master tape, for fur- d T Fd ZU
ther processing. - + + =. (8)
dt , + =
qr 0

RESULTS OF PSD COMPUTATIONS Differentiation of Eq. (8) using matrix opera-

tions results in the standard (coupled) equations
The results of all PSD samples calculated of motion shown in matrix notation by Eq. (9):
for West German roads (4) are summarized in
Fig. 6. The power spectral density (displacement [m] { } + [c'{4} + (k]{q) (Q} . (9)

77
4.104

103~
104 0

ce 17 - _

1 OM2

10-

0 04

10-3 5 X 103 IOM25X102 01

FREQUENCY (CYCLES/IN.)
PSD 1-7 -CONCRETE GOOD 3/4' UPGRADE AT START, LEVEL END
PSD 1-72 -CONCRETE GOOD LEVEL SMOOTH NEW
P.SD 2-8 AR - BITUMINOUS LOW FAIR, SLIGHT DOWNGRADE
______ - PSD 2-13 - BITUMINOUS INTERMEDIATE POOR; 3/40 UPGRADE, QUITE ROUGH
-.. PSD2-23 -BITUMINOUS LOW FAIR; 1mI/4o DOWNGRADE, WAVY
- PSD 2-5 -BITUMINOUS HIGH FAIR; 10 DOWN FOR 3/4 WAY, THEN 40 UP.

Fig. 6. PSD envelope of West German roads

78
 0.0

0.0

C-4

I._

LL.

(RAD/P\
C
QE
FR
E
z 0.01 0111U
RE

LL(T WAEENT
Fi.7 u mrtnelp franS'

Vc LU9

zt 0.0079
 By use of the superposition principle and the
orthogonality relationship of the modes (eigen-
vectors), the equations can be uncoupled as
indicated by Eq. (10):
uTR
E](]}+(uTR~c~]i}f

+ ( TR[k] [u]){(} = uTR{Q). (10)

From performance of the indicated matrix

operations, the uncoupled equations of motion
are obtained as shown in matrix notation by ' j
Eq. (11) and expressed individually by Eq. (12):

Transporter-launcher configuration [M] b?} + C] {i} + [K] {7) {N}, (11)

Fig. 8.

cO3 c02 I.. CoIJ 01

'03 3 02 t"
(x2
Y2 a2 J 020l.1 ..

k l~. k1

-022-2
k2
k
c2 C
I
1

L-174

Fig. 9. Mathematical model -- 18 degrees of freedom

80
 4_

Mrrir + Crrr + Krr7r Nr(t). (12) harmonic model solves for the harmonic re-
sponse caused by a harmonic input and uses
The complete discussion of the mathemati- the calculated transfer function s to determine
cal models is given in Ref. 7. Figure 10 shows the PSD response for a given road PSD input.
only the sequence of operations that were per- The program also determines the mean square
formed to analyze the dynamic system by both (ms) value of the response by the summation
digital and analog computers. The coupled process of Eq. (13):
equations of motion were used directly in the
analog program to obtain the harmonic sysiem
response. In the digital computer programs, =
J
PSD(c) dc. (13)
they were first uncoupled before proceeding to
0
solve for either the time- or frequency-
dependent response. The transient model
solves for the instantaneous system response, The rms value represents the statistical aver-
giving the physical motiou of the system (accel- age system response resulting with a given road
erations, velocities, and displacements). The PSD input.

MODEL

_4.
COUPLED EQUATIONS' COMPUTER

UNCOUPLED
EQUATIONS OF MOTION
I
TRANSIENT INPUT I HARMONIC INPUT PSD
ROAD PROFILE, BUMPS SINUSOIDAL ROAD INPUT

TRANSIENT HARMONIC
MATHEMATICAL MATHEMATICAL
MODEL MODEL

GENERALIZED GENERALIZED
COORDINATES COORDINATES
4(t), q(t) c(W), q(W)

,'y (t)t i ) 1*)/

TF --
ii
I PSD RESPONSE 2
PSD = PSD xTF I
I STATISTICAL VALUES
M.S. = t-.2(t)dt
TATISTICAL VALUES
M.S. =fOOpsD

Fig. 10. Outline of mathematical model solution

81
STATISTICAL SYSTEM envelopes for vehicle velocities of 30 and 50
RESPONSE mph are shown in Figs. 11 and 12. The rms
accelerations along the length of the payload
Acceleration density calculations were are.presented in Fig. 11. The corresponding
made for the complete transporter-launcher rms envelope of the trailer is shown in Fig. 12.
using the PSD data of all the road samples of The influence of the vehicle velocity on the
West Germany. Sample rms acceleration system response is indicated in Fig. 13.

0.5-

ENVELOPE OF ACCELERATION RMS VALUES

ALONG LENGTH OF MISSILE
0.4 - FOR 50 MPH
FOR 30 MPH

1
S---

- ROUG
HROAD
<j0.3

7
~~zzz

0.1 - MSSANI

30 60 90 120 150 180 210 240

MISSILE STATION (IN.)

-Fig. 11. Transporter-launcher system response envelope -payload

0.4
ENVELOPE OF ACCELERATION RMS VALUES
ALONG LENGTH OF TRAILER
- FOR 50 MPH
FOR 30 MPH
- -
p
- /1

" "'"
'-" ,.... . ........ D ''

100

TRAILER LENGTH (IN.)

Fig. 12. Transporter-launcher system response envelope- trailer

82

- . - "-
 O PAYLOAD FORWARD
0 PAYLOAD AFT
0.6 - ROUGH ROADS
- -SMOOTH ROADS

0.5

- 0.4
/
0

U 0.3
U
U

b 0.2

0.1

0 10 20 30 40 50

VEHICLE VELOCITY (MPH)

Fig. 13. System response vs vehicle velocity

The results of the statistical solution (PSD hard-surfaced roads. They can be used as a
respdnse) were then compared with the results preliminary input envelope in the fatigue analy-
obtained by the deterministic solution (transient sis of road mobile vehicles.
response), and good correlation was obtained.
The peak response determined from the tran- The generated system response for a high-
sient solution was, in general, from two to four way mobile weapons system shows a sample
times the rms value, determined from the ms solution of the analytical procedure formulated
value of the instantaneous response. This is in b Goodyear Aerospace Corporation. Both the
good agreement with the statistically expected
peakinput by adye
and systemrespne aorpor an
response analyses canbe
be use
used
peaks,vvery effectively to determine the expected dy-
namic environment of any road mobile vehicle.
Since the mathematical models (computer pro-
grams) are written in the most general form,
CONCLUSIONS employing mostly matrix operations, they can
A roughness measurement technique and be used with any dynamic system.
its subsequent application to the statistical
evaluation of the system response of a highway Optimization of the raw data processing
mobile transport vehicle has been presented. technique makes it possible to record large
numbers of test strips and convert them to ac-
The plotted displacement density envelopes tual profiles and PSD input data. These can be
of hard-surfaced profiles provide a good sta- used with either a statistical or deterministic
tistical description of the roughness content of solution.

83
 REFERENCES

1. J. Sattinger and S. Sternick, "An Instru- (Goodyear Aerospace Corporation, Akron,

mentation System for the Measurement of Ohio), July 1963
Terrain Profile," U. of Michigan, Tech.
Rept., December 1961 5. GERA-868, "Road Roughness Study of
Arizona Highways Transporter- Launcher
2. R. B. Blackman and J. W. Tukey, "The Subsystem," Revision A (Goodyear Aero-
Measurement of Power Spectra from the space Corporation, Litchfield Park, Ari-
Point of View of Communications Engineer- zona), August 1964 f
ing," The Bell System Technical Journal, .

(January and March 1958) 6. J. H. Walls, J. C. Houbolt, and H. Press,

"Some Measurements and Power Spectra of
3. J. W. Tukey, "The Sampling Theory of Runway Roughness," NACA T.N. No. 3305
Power Spectrum Estimates," Symposium (Langley Aeronautical Laboratory, Langley
on Applications of Auto Correlation Analy- Field, Virginia), 1954
sis to Physical Problems (Woods Hole, 7. GERA-693, "Phase I - Final Technical t
Massachusetts, ONR, 1949), pp. 47-67 Report, Transporter-Launcher Subsystem
(WS 325A), Part B - Volume 10 - Suspen-
4. GER-11092, "Reduction of Highway Profile
Data for MMRBM Transporter-Launcher
Structural Analysis," Supplement 1
sion System" (Goodyear Aerospace Corpo-
ration, Litchfield Park, Arizona), March
1963
I
DISCUSSION

Mr. Fitz ibbon (Mechanics Research): It matrix and the damping matrix is not observed, 4
would appear that this type of mathematical the modes zre coupled, even for very low damp-
model would be most useful for design purposes ing.
when one is studying various suspension sys-
tems, namely, various shock absorbers. Ap- Mr. Wursche: This is true. In some
parently you have used a modal analysis tech- cases when the off-diagonal elements are
nique which leads to some very serious neglected, adjustments for them are made
deficiencies. One must have particular when the equations are coupled.
constraints on the damping properties that go
along with the spring constants, primarily Mr. Fitzgibbon: There is an approach
the c matrix. The damping matrix has to be using a complex eigenvalue solution where one
proportional to the k matrix. How do you in- can, in fact, handle arbitrary damping.
tend to handle this problem ? i
Mr. Wursche: Yes, we also have used this

Mr. Wursche: It is true that the analysis approach, but it usually cuts down the number
is limited to damping less than 30 to 40 percent of degrees of freedom to about 12. If one tries
critical, which is slightly above most of the to run 18 degrees of freedom, some simplifying
damping factors used in commercial vehicles, assumptions must be made, because more than
They usually run about 15 to 20 percent critical, twice the computer storage is needed to get a
Most of the missile system damping coefficients complex eigenvalue solution. In a PSD analysis
are around 40 percent critical (Minuteman), and the effects are smoothed out and one gets just
for some of the others it is slightly less. Be- statistical values.
yond 40 percent critical damping there is a
problem in the coupling of your matrices. So Mr. Scott (Sandia Corp.): The profilometer
it is true that this analysis is limited in that seems to depend a great deal on the wheel di-
sense. There is no suitable means, however, ameter, on the tire pressures, and on the fact
of going to a very highly nonlinear system un- that the vehicle is rolling along in a straight
less a nonlinear or a transient analysis is per- line and there are four wheels on the truck.
formed. There is no harmonic solution possible. How did you decide what tire pressure and what
wheel diameter to use?
Mr. Fitzgibbon: I was not concerned about
the magnitude of the damping but the formation Mr. Harvey: We did not decide this be-
of it. If consistency between the stiffness cause the profilometer was already in existence.

84
We tried different pressures in the tires but on the missile itself? What was the upper limit
the tires are so large and the pressure is so of the frequencies recorded?
low that there was no sensitivity due to that.
Because of the diameter of the tires we could Mr. Wursche: The upper frequency limit
not measure wavelengths accurately under was about 40 cps. Most of the response was
1-1/2 feet, but the response of our vehicle was below 10 cps, either due to the natural frequen-
not sensitive to these so all the wavelengths cies of the missile's suspension system or the
under 1-1/2 feet were filtered out. coupling and the intereffect of the main trailer
suspension and the wheel hop frequency. Above
Mr. Forkois (U.S. Naval Research Labo- 10 or 12 cps the response dropped off very fast.
ratory: Did you consider using a machine to
simulate the road profile so that you would 40-cps response above 1 g or 1/2 g? What
have a means for actual testing of the vehicle kind of levels did you get on the missile?
and the transportable item?
Mr. Wursche: The peak acceleration was
Mr. Harvey: I think you have to have the slightly below 0.5 g and there were a few iso-
road profile to simulate it using a machine. lated peaks near 0.5 g.
What did you have in mind?
Mr. Zell (Picatinny Arsenal): What were
Mr. Forkois: Rotating drums with some the relative suspension frequencies of the basic
sort of an axis-moving device to change the vehicle itself and the secondary suspension of
profile. the missile? Do you feel that the frequencies
chosen would have any bearing on the overall
Mr. Harvey: If one had a profile and the approach for designing secondary suspension
vehicle, that could be done, but we did not have frequencies for equipment to be transported by
either. We decided on this approach. wheeled vehicles?
Mr. Wursche: The predominant suspension
Mr. Forkois: It might be valuable in labo- frequency of the trailer was slightly over 1.2
ratory testing. One would not have to go out cps and the frequency of the missile suspension
in bad weather. was about 3 cps. We tried to stay above the
very lowest frequency of the trailer suspension
Mr. Harvey: I think that the tapes we have as much as we could so that we would not have
now would be very useful in laboratory tests. interaction. We could not make it too stiff be-
For example, there is a ride simulatbr in the cause we had to meet our 1/2-g rms accelera-
land locomotion laboratory. These tapes could tion to the payload during the continuous deploy-
be used to force the ride simulator, and one ment. We tried to make it at least twice as high
could probably get very useful human engineer- to get away from the basic frequency of the
ing information, trailer suspension. The next frequency of mis-
sile rotation was about 3-1/2 to 4 cps. The
Mr. Rice (Goodyear Aerospace Corp.): next predominant one was the wheel hop fre-
What did the power spectral response look like quency, about 10 cps.

85
 A
PROGRAM 624A-TITAN III-C
TRANSPORTATION TESTS
Louis A. Molinari
United Technology Center
Sunnyvale, California

and
Jack R. Reynolds
Space Systems Division
Los Angeles, California

This paper reports on the development of Program 624A transportation

systems, and in particular, the major tests contributing to their de-
velopment.

INTRODUCTION Container Systems

Award of the Air Force contract for Pro- Containers engineered to protect Titan rn-C
gram 624A (Titan rn-C) presented United Tech- components during shipment or storage consist
nology Center (UTC) with some very new and of three major assemblies: cover assembly,
interesting transportability problems. Trans- side enclosure, and base. Base and side closure
port of Program 624A Solid Rocket Motor (SRM) are equipped with ports for connection of air
components, many of which would be of excess conditioner discharge and supply hoses. Pro-
weight and dimension, by truck and rail mode tection caps cover the openings when not in use.
over long distances in substantial volume would Lifting eyes provided on the cover base and side
be required. In addition, special transport closure are used when assembling and lifting the
equipment and systems would have to be devel- empty container.
oped for protection of shipments against en-
vironmental extremes of shock, vibration, and The loaded container is lifted by use of a
climate to insure that each components' operat- hoisting fixture attached to the packaged motor
ing reliability would not be jeopardized during component. Seals at the interfaces between
transit. major assemblies and on access doors assure
a leak-tight enclosure when assembled. Four
machine pads are provided on the container
BACKGROUND base to index the container to transporter deck
interfacing hardware and resist shear loads.
The following descriptions of shipment
characteristics, container system, and Program Enclosures are provided on the outside of
624A environment requirements are provided the base for mounting and protecting temperature,
as a background for test particulars contained humidity, and sbock recording instrumentation.
in the body of this paper. Interfaces for attachment of shock accelerom-
eters are provided on the outside of the base.

*Shipment Characteristics
Environment
Physical characteristics of the SRM com-
ponents and contractual/environmental require- Climate and humidity controls are required
ments for specific components and aerospace during transit to assure the technical integrity
ground equipment (AGE) is reflected in Table 1. of segments, closures, and nozzles. The

87
 -4 4, -IC 4 CJ
(aeuapcoaddV) oasl pav
r-
.ao

suoysuaeTtI Vo.4 D
C4 -f
lotCI CCDto ppol'i ot4v C.CD Cq,
IBqUUOo2UldditlS r-414 1-4-4 14 -4 V-4 V-41

Ib
SO bl bb bb -1 bB-:8
-4
bOIt; b B ~
bbD bb- *bB
(oaumtxojddV) IL LIM LoD
suoisllatua uWouodtuoo ~ow e
co ci --qvo
0CD q qr f)Cq
cQ co v'om cq c
eq Cqa
o c o
Of)
coo
I o
,-aI w
eq to
v L
W44 pq -4I

SPUnl uj (*xoiddV) 0 Cj
U01 aU9tunils 21 HDy 00 C)E

2rndpl Tejoi

(oetumixddV) -

40!qtA 2UMdd1S CD 00 CCD Lr. Ceq

luouodruoo 00 Cq

1-C

CA*ON IIVd T-1 eq 0 E- Co 0 CD eq Cq T-4

pum 0 0D
v v
eqI
co
"I
CD
M~ eq 0 0 0 o
co to wo q 0 0D 0
*ON2ulaw~aafDi 2 4 2 4: -

C4
. -t 4 .

-4 0~
E.i -m-4
Va4 ,,- o. Q'- ;4 >

0. 0~ C)> 0 0~ 0
0V -4 4 w~ -4 t.
1-4 4 0 ~1-

Cq
eq
s-4 $4

;4 0
0 ~ E W 0U
. - .- 4J 1-4

34
0 - ~ C)0 En'
bO0 0 LI
Q) 0 0
W 4)

0 0 0

88
temperature range of 600 to 900 F, with a maxi- which the AGE and tooling would be subjected
mum of 50 percent humidity which must be during packaging, loading and transport; and
maintained, is provided by temperature control (c) to test the engineering calculations and
units (TCU) powered by diesel engine genera- predictions governing the developing and han-
tors (DEGS) which accompany the shipments. dling of transportation procedures and tech-
Since the repair and maintenance of DEGS and niques.
TCU's require special in-transit services to be
performed by the railroad or trucking firm The test configuration consisted of the
transporting the shipments, special pamphlets TPGR segment, a shipping and storage container-
of instructions have been issued for carrier base assembly, and th( transporter. The trans-
guidance. porter consisted of a low-bed trailer (60-ton
capacity) 11 ft. wide, a 16-wheel jeep dolly, and
Shock limitations for segments, closures, a standard truck tractor. The jeep dolly was
and nozzles will be restricted to a maximum of equipped with a universal fifth wheel, and the
3 g on either the longitudinal, transverse, or tractor was equipped with a standard fifth
vertical axis. Accordingly, shock mitigating wheel.
protection is required. To provide special
shock protection required on transcontinental Test instrumentation on the TPGR segment
rail runs, specially built cars were fabricated consisted of 24 strain gages placed in the verti-
by the Western Pacific and Southern Pacific cal axis and 8 strain gage accelerometers
Railroads. The car features a 20-in. travel mounted, as shown in Figs. 1 and 2. Container
hydracushion underframe and a 200,000-lb pay- base instrumentation consisted of 8 strain gages
load capacity. located as shown in Fig. 3. Another set of 8
strain gages was located on the trailer, as V
Sensing devices and recorders installed on shown in Fig. 4; 4 gages were rgounted on the
containers housing each component provide a trailer beam in the vertical axis, and 4 on the
continuous record of conditions encountered in trailer-beam flange in the horizontal axis.
transit. If any prescribed limits are exceeded,
the component is subject to special inspection Test activities consisted of: (a) lifting the
procedures at destination to insure technical segment clear of the ground and positioning it
integrity and acceptability, on blocks, (b) lifting the segment off the blocks
and installing it on a container-base assembly,
(c) installing the segment package on the trailer,
TEST PROGRAMS (d) measuring transporter trailer bed deflections
before and after loading with segment, and (e)
Transportation test programs related to simulating highway movements.
shock and vibration environment consist of:
Results
1. Normal handling acceleration and strain
test, Segment Package Build-Up - Oscillogram
data indicated that the strain on the segment
2. Highway transportation test - 400 miles, case during package build-up was insignificant.
Maximum strain was estimated at 0.00030 in./
. Railcar impact test, and in., as compared to the allowable 0.00420 in./in.
As the segment/ring assembly was lowered onto
4. Transcontinental rail test. the container base, oscillogram data indicated a
maximum acceleration of approximately 0.4 g
in the X axis, 0.3 g in the Y axis, and 0.26 g in
Normal Handling Acceleration and the Z axis.
Strain Test
Maximum accelerations and strains re-
The purpose of the test was to demonstrate corded during installation of the segment pack-
the effects of normal handling and movement by age on the trailer occurred in lowering the
highway transporter in preparation for an actual package onto trailer bed; g readings were X
400-mle road test. In this case, transportation axis, 0.22; Y axis, 0.10; and Z axis, 0.20.
conditions were simulated by artificial means.
The primary objectives were (a) to demonstrate Deflection Measurements - Maximum
the normal strains and accelerations to which trailer bed deflections with the segment pack-
a thermal gradient (TPGR) segment would be age installed were 1.64 in. with trailer in line
subjected during packaging, loading and trans- with jeep dolly and 0.94 in. with the jeep dolly
port; (b) to demonstrate the normal strains to at a 45-degree angle to the trailer centerline.

89
 N4J

< U

p C4

I:x:N
04

H
4.)

on 0
94C,
E.

011) i
'N,4

InI

C,

90
 ARROW INDICATES TOP
DEAD CENTER

STATHAiI 20812 TRI-AXIAL

MOUNTING BRACKETS
(3 PLACES)

X-AXIS, STATHAM A69TC-25-350

PROPELLANT ACCELEROMETER
.Y-AXIS, STATHAM A5-50-350
Y- AXIS ACCELEROMETER
Z-AXIS, STATHAM4A5-50-350
ACCELEROMETER

X-AXIS, STATHAMA69C-25-350
ACCELEROMETER
Z-AXIS, STATHAM A69TC-25-350
ACCELEROMETER

X-AXIS, STATHAM~ A69TC-25-350

ACCELEROMETER
Y-AXIS, STATHAM A5-50-350
ACCELEROMETER
Z-AXIS, STATHAM A69TC-25-350
ACCELEROMETER

Fig. 2. Location of accelerometers on TPGR segment

Figure 5 plots trailer bed ceflections with the certain point. This point was reached when the
jeep dolly at a 45-degree angle to the trailer, right rear corner of the trailer bed had jacked
A point by point breakdown of the deflection up to a height of 4 in. At this point, the left
measurements is given in Table 2. front corner of the trailer bed began to drop as
the right rear corner lifted. The drop and re-
Simulated Highway Movements - Simulation sulting tilt of the segment package were greater
consisted of: (a) lifting right rear corner of with the universal fifth wheel than with the
trailer bed to 5 in. (jeep dolly in line with standard fifth wheel.
trailer), (b) lifting corner of trailer bed to 4 1/2
in. (jeep dolly at 45-degree angle to trailer), The overall maximum acceleration occurred
(c) lifting corner of trailer bed 5 in. (tractor when the trailer was pulled off the blocks and
equipped with standard fifth wheel in line with brakes were applied. This acceleration was
trailer), (d) lifting corner of trailer bed to 4 1/2 0.21 g in the X axis, 0.14 g in the Y axis, and
in. (trailer at 45-degree angle to tractor), (e) 0.13 g in the Z axis. The overall maximum
pulling trailer off 2-in. blocks with braking on strain on the trailer bed also occurred when the
impact, and (f) pulling trailer off 2-in. blocks trailer was pulled off the blocks and the brakes
with no braking on impact. were applied. This reading was 0.00901 in./in.
During the lifting tests, it was found that Overall maximum strains on the segment
the trailer would not continue to twist beyond a case and the container base were recorded
91
 315

1350
/ ARROW INDICATES TOP
DEAD CENTER
I
(oI
BLII FAB-50-35-56 STRAIN GAGES
EQUALLY SPACED WITH 4 ON THE
TOP RIM AND 4 ON THE BOTTOM RIM
I

SYBL DCTS STRAIN GAGE 4IN

Fig. 3. Location of strain gages on

container base

CONTAINER BASE ADAPTERS

=7BH FAB-50-35-56 STRAIN GAGES (8 PLACES)

Fig. 4. Location of strain gages on trailerI

92
 TRAILER AND JEEP IN-LINE
1"4
+0
1,' ,,r8 -.-
2" - -

RIGHT SIDE OF BED LEFT SIDE OF BED-----------

TRAILER AND JEEP AT 450 ANGLE

O 'A - ..,25
li-7

-! 374" 472"

TABLE 2
Trailer Bed Deflection

Deflection, in.
Point on
Trailer Bed Jeep Trailer Jeep at 45-Degree
in Line Angle to Trailer

1 -1.50 0.30
2 -1.64 -0.51
3 -0.58 0.29
4 0.24 -0.42
5 0.34 -0.80
6 -0.58 -0.94
7 -1.53 -0.45
8 -1.49 0.43

during lifting of the right rear corner of the As a rule, strains on the segment case in-
trailer bed to 4 1/2 in. with the trailer attached creased when the trailer bed was lifted; how-
and jackknifed to a 45-degree angle to the ever, there was a decrease noted in two areas
tractor. Readings were 0.00208 in./in. corn- on the side of the segment case farthest from
presslon on the segment case and 0.00563 in./in. the lifted corner of the trailer bed. These two
tension on the container base. areas were near the bottom weld 195 to 285
degrees from TDC and near the top weld 165 to
The greatest overall strain occurred at the 225 degrees from TDC.
top weld of the segment case, 105 degrees from
top dead center (TDC). The greatest strain on Conclusions - During normal handling,
the container base was in the vertical axis 225 strains and accelerations on the segment were
degrees from TDC. less than half of the allowable limits; however,

93

.m~-- ~ ~ - -
 VL

strain was expected to be greater during the SRM/AGE package with the transporter and the
actual highway movement test. The recorded transportation route.
data indicated that strains on the trailer bed and
the container base were greater than that on the Special permits issued by the California State
segment case, but these strains were not criti- Division of Highways permitted the test load to
"- cal. Use of the universal fifth wheel between travel between sunrise and sunset, Monday through
the trailer and the jeep dolly will result in less Friday. The route followed was established by a
strain of the segment, container base, trailer, previous road survey which considered road
and hitch point because of the design of this characteristics, capacity of structures, terrain
type of fifth wheel. traversed, and other items pertinent to stability
and safe travel of the load. The route selected
was authorized as an "explosive route" for Pro-
Highway Transportation Test - 400 Miles gram 624A loads by the State of California Fire
Marshal. Approved stops located away from
The purpose of this test was to demon- large towns or densely populated areas were
strate the feasibility of transporting live solid authorized for overnight layovers. The composi-
rocket motor segments from UTC, Development tion of road test convoy was as shown in Fig. 6.
Center (near San Jose, California) to Edwards The data acquisition system operated continu-
Air Force Base, over approximately 440 miles oualy during the load movement, and one hour
of highway. Data collected during this test before and after each day's travel, to obtain data
were used to develop an acceptable method of on the varying wind conditions and their effect on
transporting live segments via highway. the temperature of the segment.

Objectives of the test were to: (a) demon- Composition of the Test Load - The test
strate the feasibility of transporting a live SRM load consisted of a live propellant segment
segment and associated AGE by means of stand- (Class "B" explosive), associated AGE (container
ard commercial heavy duty equipment, (b) to assembly, TCU, and DEGS), special trailer inter-
monitor and record the environmental effects to faces to accept the load, and a data acquisition
which the SRM would be subjected during the system. Locations of the load, AGE and data
test, and (c) demonstrate the capability of the acquisition system are shown in Fig. 7. TCU was

Fig. 6. Road test convoy

94

i/

* i
I'|
- -,--------
-- - - - -4

to

En X

In ;

Cd 0

- to

9 Cd

95d
4~
- -*- --.-.- - r
 dimensions were an 80-ft. length, ll-ft. 4-in.
?T>T:T-:;width, and 17-ft. 5-in. height. Trailer clearance
~~ from ground was 11 3/4 in.
S:Instrumentation - The total of 117 thermo-
couples were used. Thirty-six of these were
previously cast into the segment grain, and the
remainder were located as shown in Figs. 9 and
"1 10. Twelve accelerometers were used, with 11
- -. - . mounted on the segment handling rings, located
* .. as shown in Fig. 11. Twelve strain gages were
mounted near the bottom weld of the segment
case, all in a vertical axis as shown in Fig. 12.
Two humidity sensors were used. One was lo-
cated over the core of the segment and the
other, an anemometer, was located on the
electrical equipment shelter.
Results - Representative zero to peak ac-
celerations and force frequencies were recorded
for each major segment of the route, as well as
those experienced by inducing a deliberate stop
on an incline. A sample of such recordings is
given in Table 3.
Tests were also conducted to determiie the
natural frequency of the loaded trailer. These
data, which were obtained by driving the loaded
a i m trailer over 4-in. blocks, indicated that the loaded
Fig. 8. Loaded transporter trailer had a natural frequency of 1.95 cps and
that shock was mitigated between the trailer bed
and the bore of the segment by a factor of ap-
proximately 3.0. Table 4 reflects these data. A
mounted on the trailer bed aft of the load, and trailer deflection test indicated the maximum de-
the data acquisition system was installed in a gree of trailer deflection under weight of the
shelter mounted on the trailer gooseneck. Slid- container assembly was 2.70. Figure 13 provides
ing boards were placed over the electrical a point by point breakdown of the trailer deflection
shelter and load to enable the transporter to measurements and plot.
clear low wires and branches. The-loaded
transporter (Fig. 8) weighed approximately Conclusions - Maximum acceleration dur-
160,000 lb. gross, and its loaded maximum ing the road test were less than 1/5 of the

TABLE 3
Representative Zero-to-Peak Accelerations and Forcing
Frequencies from UTCDC to Mendota

Route Area Acceleration Forcing

op eterai (g) Frequency (cps)

UTC Development Center main gate 0.20 - 0.31 2-3

San Felipe Road and Silver Creek
Road 0.20 2-3
King Road, Aborn Road, and Loupe
Avenue triangle 0.29 - 0.38 2-6
U. S. Highway 101 0.10 - 0.15 2-3
Pacheco Pass 0.20 - 0.25 2-3
San Luis Dam construction site 0.20 - 0.25 2-6
Entering Los Banos 0.20 - 0.25 2-5
Dos Palos railroad tracks 0.20 - 0.40 2.5-3
Entering Mendota 0.20 - 0.25 2-3

96

- - -

,(
 ____
all

LAI

al UP r

8K 1
In cc

*0 In

010

cd~

Vc. -

147

C,~~-
o
0- =
E-
 IIT
v
,J

4 CI
1.
0

4)T
-. 1 -0,

-n-

4--

Q01

Vol,

980

'-4L
 TABLE 4
Natural Frequency Test of Loaded Trailer
(4-in. Block Test)

Location Acceleration Response

Accelerometer
LoIation (g) Frequency (cps)
Accelerometer
A5 - Y axis Trailer bed 0.36 1.97
A6 - Y axis Segment top 0.3 1.97
A9 - Y axis Container base 0.95 1.95
A10 - Y axis Segment bore 0.28 1.97
All - Y axis Segment bore 0.3 1.94.
A12 - Y axis Trailer bed 0.95 1.95

maximum allowable 3 g. Measured strain on route, and road hazards such as narrow road-
the segment and the case were insignificant, way, low wires, over-hanging tree branches,
Examination indicated that the propellant was and road dips were minor and not critical.
not broken or cracked, and there was no failure
of the bond between the propellant and liner. It was concluded that on this type of load,
The shipping package was compatible with the the standard fifth wheel has a distinct advantage
transporter and the established transportation over universal fifth wheel. The standard fifth

STAT.A A69TC-25-350
A7 %.i ACCELEROMETER
;2 PACS)
_ 6 ...
\

X-__AXIS + ARR INDICATES

TOP DEAD CENTER

ACCELEROMETER
AI
A X
A2 Y
A3 Z

A7 X
A8 Z Z-AXIS
A9 Z
A10 Y
All *f-AI -FEMALE CLEVIS-IONT
A12 _Y HNDLIF3 KI
A4, AS --- .-

A6

.- PROPELANT

A8 \ \ -,,

All \,\\
', \\ ,,MALE CLEVIS-OINT

. CONTAINER ASE

T_.A ________ A12

LA.. A2 A3\\-TRAILER 'ED

Fig. 11. Accelerometer locations

99

--
--------
 HYDRODYNAMICS A22-56100
HUHIDITY SENSOR

CABLE ASSEMBLY LEADS TO TOP DEAD CENTER

DATA ACQUISIMON SYSTEM POELN

CAGES EQUALLY SPACED AND PLACED STRAIN CAGE

APPROXIMATELY 1 INCH FROM THE
BOTTOM WELD

Fig. 1Z. Strain gage and humidity sensor locations

/rch P/# 4-'wrANr 9A56 AAAPf"(,

( ,M6,/')

1 ... _,_ -
_... ... Uh!

-~~ - .L*i- -~~--.-- r--

0.881WWfWM(
Fig.13.Tralerdeflection measurements and plot

10
wheel held the loaded trailer in line with the mounting pads provided; (c) simulated segment
tractor and jeep dolly on superelevations, filled with sand, weighing 85,000 lb weight, 11 ft
whereas the universal fifth wheel allowed the 4 in. in diameter, 13 ft 8 in. high, mounted in
trailer to travel and "lean" on steep supereleva- AGE container base and secured to freight 'car
tions. bed on the interface mounting pad; (d) air con-
ditioner mounted on the bed of the railcar powered
by the DEGS with air conditioner hoses leading
Railcar Impact Test to the test shipment; (e) two DEGS secured to
mounting plates provided on bed of the railcar
The purpose of this test was to demonstrate (one DEG furnished power for the shipment
the effect of couplings at various speeds on the environment system and a second set furnished
special 100-ton capacity hydracushion railcar power to the data acquisition system); and (f)
and on a test load consisting of an inert SRM shock recorders and accelerometers mounted
segment and a simulated segment. Test objec- in their routine shipping positions on the inert
tives were: (a) to demonstrate (in part) the container base. Figures 15 and 16 illustrate
transportability of SRM components by rail, (b) the test setup.
to demonstrate the accelerations to which the
SRM components and AGE would be subjected Test instrumentation consisted in part of
when the special car was impacted at varying two accelographs, each with 8-channel capacity.
speeds, and (c) to prove the adequacy of design These accelographs and two AGE shock re-
of the car deck interfaces securing the ship- corders were used to record the impact accel-
ment to the car. erations. Sensor illustration consisted of 14
accelerometers located as shown in Fig. 14.
The reports, recommendations, and con- Five accelerometers were mounted on the
clusions accumulated as a result of the impact railcar and nine on the segment, handling ring,
test were to be used to finalize an acceptable and shipping container base.
method of transporting live components by rail
to the Eastern Test Range (ETR). The test was Pre-Test Procedures - The railroad crew
also to be witnessed by an American Association was briefed on the test procedures. Five empty
of Railroads (AAR) Bureau of Explosive repre- buffer cars were placed at the far end of the
sentative who could recommend that special In- 200-ft track, with brakes set and couplers ex-
terstate Commercial Commission (ICC) permits tended. The test car was placed between the
could be issued approving the proposed method switch engine and buffer cars. Track torpedoes
of shipment if results were satisfactory. Due and electric timer switches were placed on the
to the considerable movement volume projected test track for use in computing test car's speed
(each shipment consisting of 14 cars per ship- of traveling.
ment carrying about one million pounds of ex-
plosives plus 300,000 lb. of AGE), the railroads Test Procedures - Test runs consisted of
and the ICC were anxious to have proof that the accelerating the test car with a switch engine to
shipment method and mode of transport would a speed approximating a previously designated
be safe and compatible, with regular freight speed, releasing the car from the switch engine,
train service operations. and allowing it to roll into the five buffer cars.
After each run, the draft gear was extended on
A general layout of the railcar impact test the buffer cars and the brakes were reset. A
is shown in Fig. 14. Equipment furnished by complete inspection of segments, railcar inter-
the rail carrier consisted of: (a) 100-ton ca- faces, and AGE were conducted after each test
pacity 20-in. travel hydracushion underframe run.
freight car equipped with special interfaces for
Program 624A shipments; (b) five empty rail- The first series of tests were run with the
cars to be used as buffers; (c) one switch engine; test car fully loaded, as shown in Fig. 17. Total
(d) 200 ft of reasonably level track; and (e) track weight of the test item was 247,800 lb. In
electric timers, stop watches, and track tor- this series of tests, the car was impacted on
pedoes. the A end at 5.50, 5.65, 7.50 and 10.30 mph.
Then the B end of the car was impacted at 9.70
The test configuration consisted of: (a) the mph.
data acquisition system mounted on a truck
connected by cabling to the instrumentation On completion of the first series of tests,
and DEGS on the freight car; (b) inert segment, the simulated'segment and container base were
weighing 82,000 lb, 11 ft 4 In. diameter, 13 ft removed from the car for a second series run.
8 in. high, mounted in AGE container and se- Total weight of this new configuration was 162,000
cured to the freight car bed on the interfacing lb. During the second series, the car was

101
 '.4l

: 0 4. $04 4
14[
40 a. a2w2w2wg 4 4
64-

1 3

102
 Results - Rail impact accelerations result-
ing from the test were as given in Table 5 for
accelerometers A2 through A12. No accelera-
tions over the maximum tolerance of 3 g, with
the exception of accelerometer Al located on
S. the railcar coupler, which sustained a load of
.- , 25 g, occurred on the first test run. This ac-
celerometer was damaged beyond use and was
- removed for the remainder of the test. Highest
accelerations obtained from the test were in
the X axis, 2.27 g; in the Y axis, 2.90 g; and in
the Z axis, 0.94 g. Highest impact speed was
10.30 mph.

The AGE shock recorders monitoring ac-

celerometers A13 and A14 registered no ac-
celerations of 3 g or over.
Fig. 15. Railcar impact test setup,
side view Maximum draft gear sill travel was 18.5 in.
at an impact speed of 10.30 mph. Maximum
buffer car movement along the track was 192
impacted on the A end at 5.26, 7.32, and 8.45 in. at an impact speed of 9.70 mph. Table 6
mph. The B end of the car was then impacted summarizes rail impact test sill travel and buf- "
at 6.11, 7.50, and 9.70 mph. On completion of fer car movement.
the second series of the tests, a third series
was run with only the diesel engine generator Conclusion- Accelerations encounteredwere
on the car. Test item track weight for this below the maximum allowable accelerations
configuration was 69,400 lb. During this third of 3 g. Propellant grain and container base ac-
series of tests, the car was impacted on the A cumulations were less than 2 g, and in the ma-
end at 5.00, 7.89, and 10 mph. The B end was jority of runs less than 1 g. Results of test runs
not impacted. when converted to impact forces compare favor-
ably with the results of a hydracushion car of the
Draft gear sill travel was measured for all impact test plotted in Fig. 45.41 in the Shock and
test runs. Buffer car movement along the track Vibration Handbook (1).
as a result of impacts was measured for the
first and second series of runs. The railcar impacts had no visible effects
on the railcar, interfaces, aerospace ground
equipment, and attachment fittings. It was con-
cluded that a special car and the method of
securing the AGE to the railcar are adequate to
sustain normal impacts which result from nor-
mal railcar humping operations. Based on test
results, special ICC permit was issued to cover
Program 624A shipments by the tested method
and transporter mode.

Transcontinental Rail Test

The purpose of the test was to accumulate

data and provide information to establish the
..- parameters for Program 624A standard rail-
road transportation system procedures re-
'+! +.++.:.+
"+'
.+ quired for delivery
UTC, Coyote, of SRMto components
California, from
ETR, Florida.

Objectives of the test were to: (a) prove

that SRM segments can be safely transported by
rail; (b) to determine the characteristics of vi-
Fig. 16. Railcar impact test setup, bration and acceleration of the SRM segment,
end view forward closure, associated AGE and the special

4=i
 CI

104
 TABLE 5
Rail Impact Test Accelerations

Impactar
Speed CarI Accelerometer (g)
\I
(mph) EndjA 2 -Y A3-Z A4-X A5-Y A6-X j jA7-Y jA8-X [A9-Y A1O-Z jA11-X A12-X

5.50 A 2.32 0.14 0.70 0.50 0.38 0.30 0.20 0.31 0.19 0.19 0.43
First 5.65 A 1.60 0.27 0.89 0.38 0.31 0.25 0.46 0.21 0.09. 0.29 0.25
Test 7.50 A 2.12 0.29 1.21 0.53 0,41 0.23 0.61 0.21 0.12 0.24 0.17
Series 10.30 A 2.00 0.29 2.11 0.96 0.72 0.60 0.72 0.28 0.09 0.52 0.56
9.70 B 2.32 0.49 2.27 1.29 1.00 0.62 1.34 0.36 0.17 0.45 0.40

6.11 B 2.59 0.22 0.64 0.32 0.70 0.32 0.07 0.52 0.35 0.50 0.08
7.50 B 2.55 0.23 1.39 0.69 1.00 0.86 1.21 0.82 0.94 0.96 1.85
Second 9.70 B 2.70 0.39 1.41 0.67 1.20 0.81 1.82 0.72 0.47 0.48 1.14
Test 8.45 A 2.80 0.29 1.59 0.40 0.60 0.61 1.09 0.72 0.71 0.59 0.78
Serie 5.26 A 2.25 0.12 0.67 0.39 0.51 0.20 0.15 0.31 0.19 0.24 0.25
7.32 A 2.90 0.22 1.13 0.25 0.72 0.15 0.91 0.26 0.19 0.38 0.57

Third 5.00 A 0.75 0.33 1.14 0.61

Test 7.89 A 1.48 0.41 1.98 0.38
Series 10.00 A 1.27 0.59 1.88 0.59

TABLE 6
Rail Impact Test Sill Travel and Buffer Car Movement

I c Speed
Impact Car 1d
j Sill Travel Buffer
MovementCar
End (in.) (in.)
(mph)

5.50 A 14 36
First 5.65 A 14-1/8 51
Test 7.50 A 16-1/2 89
Series 10.30 A 18-1/2 180
9.70 B 18 192

6.11 B 9-1/4 36
Second 7.50 B 11-9/16 58
Test 9.70 B 15-7/8 101
Series 8.45 A 15-1/16 76
5.26 A 11-1/4 36-1/2
7.32 A 15 50

Third 5.00 A 6-5/16

Test 7.89 A 10-1/16
Series 10.00 A 10-7/8

100-ton capacity hydracushion flatcar during Special transcontinental routes, as shown in

transport; (c) establish availability of rail sys- Fig. 18, capable of accepting the excessive
tern facilities and services for the cross- weight of car with lading (approximately 200,000
country transport of SRM components; (d) verify lb) and load dimensional clearance of 17 ft 1 in.
the reliability of the environmental control sys- in height and 11 ft 4 in. in width had been
tern in maintaining the temperature and humid- previously established. A railcar numping test
ity environments within acceptable tolerance previously conducted had verified that the lad-
during rail transport; (e) verify the accuracy ing would not be exposed to shock during normal
of the AGE recorders during rail shipment; and rail service handlings which would be in excess
(f) verify the car interface compatibility between of 3-g tolerances allowed. This test was re-
the railcar and AGE. quired to verify that the shock and vibration

105

-
~~~~~~~ 7 . -7---
4- I -- -~ - -- -- ~-- - -

II
'I

V
iJ

co
*1
II
:1
* I
* 0 ~

~ 0
* H U

0 P.~

.1
U

106

,2'

- -- ~ - - - -.. - ~
 tolerances would not be exceeded during actual for each phase. Test plan for the phase cover-
cross-country travel over the selected routes, ing New Orleans, Louisiana, to Meridian, Ala-
when shipments were moving iriregularly bama, is reproduced in Fig. 19. While the test
scheduled freight train service. Since the char- train was enroute, the location of the test car
acteristics of the track, trackbed, and terrain within the train was varied to determine the
traversed, as well as traveling conditions en- effect of such changes, especially the "whip"
countered, would have a bearing on the test re- effects experienced at the end of a long train,
suits, rail track profiles secured from each of as well as effects of slack, run-out, and shock.
the railroads involved in the 6,000-mile round
trip route were analyzed, to establish those Test load transporting equipment consisted
portions of the track over which testing and data of: (a) 100-ton capacity hydracushion railcar,
accumulation were to occur. In analyzing the interfaced for carriage of Program 624A com-
rail track profiles and establishing test loca- ponents; (b) a 70-ft long railroad baggage car
tions, consideration was given to such items as to house the data acquisition system; and (c)
track weight, condition of roadbed, travel over a personnel railcar to accommodate the test
structures, travel through classification yards, crew, AAR Bureau of Explosives representa-
grades, and curving track, so that in the final tive, and Government Agency representative.
analysis conditions representative of all of the
routes' characteristics would be included in the The test load consisted of: (a) a live SRM
test. segment, weighing 97,000 lb, 11 ft 4 in. in diam-
eter and 13 ft 8 in. high, and forward closure
Twelve test phases were established with weighing 58,500 lb, 11 ft 4 in. in diameter and
mile posts and stations reflected on the test 13 ft 8 in. high, in containers which included
plan to control the start and stop of monitoring monitoring instruments to sense and record

0 -

II

.d .. 7 .J i
...... .... * .
Nwruu
M 10 170 10
... P.A1
50 340 0 0 no 3.00 90
=40*
80 70

70 60 50 40 ,30 20 10 0
1
Stop Testint

Fig. 19. Test plan -- phase III - Southern Railway System --

Seabrook, La. to Meridian, Miss.

Fig. 20. Transcontinental rail test setup

107

~ . ~ - -. .
 temperature, vibrations arid accelerations; and car, (b) air duct hoses to container and air
(b) AGE consisting of a TCU linked by air ducts conditioner, (c) diesel engine generator to
to test shipments and a diesel engine generator freight car, and (d) air conditioner to freight
power unit. The data acquisition car and test car.
:, load are illustrated in Fig. 20.
leResults - The highest acceleration recorded

The data acquisition system was installed were caused by inherent slack built into the coup- i
in the baggage car and coupled, as shown in Fig ling system of each car. While the slack per car[
21, to the test car. The system was calibrated is small, the total slack for an entire train can 14
immediately before start of phase I test and
required. Eighty-one thermo-
thereafter as threfersriird
couples were installed as shown in Fig. 21.
be quite large and produce a "whip" effect, es-
pecially noticeable, when traversing rolling ter-
anEdrngghty-onain.DuineheI
rain and during dynamic braking. During the
1
Forty-eight were installed on the segment con- test, train lengths varied from 16 cars with a
tainer side enclosure, 24 to measure air tem- total weight of 193 tons to 185 cars with a total L
peratures and 24 to measure surface temper- weight of 8,946 tons. By adding the slack I
atures. The rest were distributed, as shown in produced per coupler, the engine would travel
Fig. 22, to measure surface and air temperatures a distance of 61 ft 8 in. to move the last car
on the segment case, segment core, handling of the 185-car train; therefore, the fewer cars
ring, container coverwere and container base. Eight- in the train, the less time required for the
een accelerometers used, as shown in engine to accelerate, ith a lower resulting
Figs. 21 and 23, with five installed on the seg- shock to a car at the end of the train, such as
ment, two on the forward closure, two on the SRM component cars. Track and roadbed con-
igniter case, four on container bases, four on ditions were minor factors because of speed
the flatcar bed, and one on the bend car coupler. restrictions imposed on the trains by the re-
One anemometer was installed on the baggage spective common carriers over areas of ad-
car. verse road conditions.
The maximum accelerations recorded were
Test Operations - The test cars proceeded 13 g on the test flatcar coupler in the X axis,
from the origin railhead near San Jose, Califor- 1.8 g on the test flatcar bed in the Y axis, and
nia, to Los Angeles by special train making 1.8 g on the SRM components in the X axis.
three braking stops at speeds of 15, 25, and 35 There were no accelerations over 3 g in any
mph on route. From Los Angeles to Jackson- axis, with the exception of those imposed on the
ville, Florida, via the Southern Route, and re- test flatcar coupler. Test flatcar and SRM com-
turn to San Jose via the Northern Route, the ponent sway, as reduced from Z axis accelerom-
test train was moved in regular freight train eter data, occurred more frequently in rail yard
service and normal operating schedules. and switch yard movements at low speeds, as
opposed to mainline movement. The frequency
Instrumentation records made were cross- range of the three orthogonal axes was 2 to 100
referenced as follows: (a) oscillograph accel- cps. The fundamental frequency of the axis was 1
eration records with train speed tapes, geo- for the Z axis, 2 to 7 cps; for the X axis, 5 to
graphic location, time and mile posts; and (b) 10 cps; and for the Y axis, 5 to 20 cps. In addi-
Bristol recorder temperature records with tion, a frequency of 40 to 100 cps was imposed
oscillograph records of wind direction, wind on the basic forcing frequencies. These im-
velocity, train speed, geographic location, time posed frequencies resulted in a complex fre-
and mile post. quency pattern. Tables 7 and 8 are two samples
of the twelve test phases recorded, indicating
maximum accelerations and frequencies im-
termine the external and internal heat absorbed posed on the test items during the trip.
by the segment container. A directional vane
and anemometer were used to correlate heat The SRM component grain temperatures
transfer with wind velocity and direction. Vi- were held between 600 and 70'F throughout the
bration and shock data were recorded to re- trip, within allowable limits of 600 to 90 F.
flect the degree of shock and vibration en- Ambient temperatures ranged from 260 to 820 F.
countered during the test phases, under the Weather conditions included rain, snow, and
varying ranges of physical road bed, train wind. AGE humidity recorders, which indicated
speeds, and terrain conditions anticipated by relative humidity of 70 percent during loading
study of rail track profiles. (in the rain) at origin raihead, slowly dropped
while moving enroute across country, and re-
At each stop and railroad division point, corded 40 percent in the closure container and
the following car interfaces were impacted by 30 percent in the segment container on arrival
the test crew: (a) shipping containers to freight at Jacksonville, Florida.

108

I-~
~~czI o0 N,

I.. b.
3 . 03

C C .
0 0 0 0

v0 0 C C

a I
0 u

.0 Cd

0 C; W

1t v rn
 M %a

low 0 ~orol--- V
~jw~)Ovi

4Q4

LjL
N~k 0___

IN''n EN co 0 co

P-3 0

121

co~~~D(
 0UI0 00

00

- .. , 0 0Q
P. I~ --

e Wo

Itz

S. 0 0

1 0

m 0 0

6=nwm--, , --- cz.

 II

TABLE' 7
! Test Phase VI

~Maximum Response
Acceleration Remarks
Location Amplitude (g) Frequency (cps)

A1-X Coupler 00 6.10 Shock Test phase conducted

between Kansas City,
A2-Y Flatbed 1800 1.80 5-10 Missouri and Gibbon,
Nebraska. !
A3-Z Flatbed 00 0.23 10-15
Flat terrain. Train
A4-X Flatbed 0* 0.26 2-6 speed, 70 mph.

A5-Y Flatbed 00 0.37 5-10 Little or no slack

action.
A6-X Grain 0.27 8-10
Midplane 1800 Average train speed
during test phase -
A7-Y Grain 0.36 3-8 47 mph.
Midplane 1800

A8-X Segment Case, 0.19 10-15

Top 180

A9-Y Segment, Case, 1.40 3-5

Top 1800 V
AIO-Z Segment Cas6, 0.34 4-8
Top 900

AII-Y Igniter 1.20 3-6

A12-X Igniter 1.10 2-4

A13-Y Forward-Closure 0.50 5-12

Adapter Ring 1800

A14-X Forward-Closure 0.70 10-15

Adapter Ring 1800

The anemometer indicated that winds caused system, rather than the entire train system, was
by the train movement were 18 to 22 mph less used by many engineers. An example of engineer
than actual train speed. This condition was control was seen between Birmingham, Alabama,
caused by a partial vacuum created by the and Jacksonville, Florida. On the trip east, the
trains when moving at mainline speed. The ride was the best encountered to that point. On the
wind appeared to have little effect on component return trip over the same track, the slack action
conditioning. was so severe during braking and runouts that
numerous components of the personnel car were
Conclusion - The maximum accelerations loosened from their mountings and required re-
received by the SRM components were less than pair. At times, it was impossible to stand or
2 g, as compared with the maximum allowable walk in the car without being thrown to the
of 3 g. The greatest accelerations were attained floor.
during the slack action of the train, which could
be severe enough to break coupler parts on the The SRM components and AGE were com-
runouts or buckle center sills on runins. The patible with the transporter and transportation
most common cause of slack action noted was routes. Interfaces between the rail flatcar and
the manner in which the train engineer handled AGE were compatible.
the braking system. The dynamic braking

112
 TABLE 8
Test Phase IX

Acceleration Maximum Response

Location Amplitude (g) Frequency (cps)

AI-X Coupler 00 3.40 Shock Test phase conducted

between Salt Lake City,
A2-Y Flatbed 1800 0.15 5-15 Utah and Wells, Nevada.

A3-Z Flatbed 00 0.25 2-3 6 in. to 8 in. snow storm.

Long grades with
A4-X Flatbed 00 0.20 5-10 numerous tunnels.

A5-Y Flatbed 00 1.10 5-20 Maximum train speed

during this phase was
A6-X Grain 0.20 3-5 50 mph.
Midplane 1800
Average train speed
A7-Y Grain Bad. Cal. during test, 29 mph.
Midplane 1800

A8-X Segment Case, 0.20 3-7

Top 1800

A9-Y Segment Case, 0.80 10-15

Top 1800
A10-Z Segment Case, 0.20 2-3
Top 900

All-Y Igniter 0.90 4-8

A12-X Igniter 0.10 2-3

A13-Y Forward-Closure 0.20 5-15

Adapter Ring 1800

A14-X Forward-Closure 0.40 5-10

Adapter Ring 1800

TEST PROGRAM. CONCLUSIONS range of equipment includes equipment of stand-

Test results indicate that the special rail ard make which can be modified to accept ship-
and highway transporters, containers, AGE and ments satisfactorily of excess weight, dimen-
environment systems fulfilled the requirements sion, and sensitive nature. Thus, these test
of the Program
Either 624A (Titan Ill-C) Program.
rail or highway results can lead to the fulfillment of the need of
transport system will both the purchaser and supplier of missile and
provide protection of the SRM components space-age equipment to substitute transport
against all normal transportation environments systems composed of combinations of modified
and assure that the maximum allowable 3-g standard components, common to industry and
shock, 50 percent humidity, and 600 to 900 F tem- readily available in the market, for systems
perature range will not be exceeded., Results requiring individual design and special fabrica-
also illustrate that the common carrier industry tion at the cost of considerable time and money.

REFERENCE
1. Sergei G. Guins, "Shock and Vibration in edited by Cyril M. Harris and Charles E.
Road and Rail Vehicles, Part II, Rail Ve- Crede (McGraw-Hill, New York) 1960, pp.
hicles," Shock and Vibration Handbook, 45-21

113
 DYNAMIC ENVIRONMENT OF M-113 ARMORED
PERSONNEL CARRIER
G. M. Pomonik and N. G. Tinling
Hughes Aircraft Company
Culver City, California

A measurement program was conducted to determine the shock and vi-

bration environment of the TOW weapon system installation in the
M-113 armored personnel carrier. Data were obtained over a range
of vehicle speeds and road conditions and at various vehicle locations
to provide general information on the M -113 dynamic environment.
Vibration data are presented for vehicle speeds between 5 and 35 mph
on both dirt and asphalt roads. Data for significant conditions is pre-
sented as acceleration spectral density and amplitude distributions.
The broadband results are summarized as functions of vehicle speed
and measurement location. Shock data were obtained for vehicle speeds
from 7 to 13 mph over wooden ramps on a dirt and asphalt Munson
course. Shock spectra of the most severe data are presented. The re-
sults are summarized as functions of vehicle speed and measurement
location.

INTRODUCTION 2. The most severe shock occurs when the

suspension bottoms.
A measurement program was conducted to
determine the shock and vibration environment These conclusions were arrived at after com-
of the TOW weapon system installation in the prehensive testing under a variety of vehicle
M-113 armored personnel carrier. Data were maneuvers and road conditions.
obtained over a range of vehicle speeds and
road conditions and at various vehicle loca- Based on these findings, the test was con-
tions to provide general information on the ducted to provide an abundance of acceleration
M-113 dynamic environment, data for conditions of (a) high speed on both
concrete and dirt roads, and (b) suspension
The M-113 armored personnel carrier bottoming.
(Fig. 1) is a high-production multipurpose
vehicle. Its proposed uses include mortar
carrier, command post, communications cen- DETAILS OF TEST
ter armored cargo carrier, and special pur-
pose weapon carrier for applications such as Most of the vibration tests were conducted
the TOW weapon system. This report and at the test track of the Food Machinery and
those listed in the references provide prelim- Chemical Corp. (FMC), San Jose, California,
inary design information for future equipment manufacturers of the vehicle. The shock tests
installations in the M-113. were conducted on a Munson track at the Hughes
The shock and vibration environment pro- Aircraft Company Fullerton facility in California.
duced by M-113
by various motion
agencies has The
(1-5). received attention
results obtained in Mhe vehle eri n erFle was used
ithe San Jose tests. This vehicle was used
provided a guideline for a more detailed inves- previously for general testing by FMC and had
tigation of critical conditions. In summary, an accumulated hull mileage of 14,500 miles.
previous tests indicate that: The tracks on this vehicle were worn. The
1. The most severe vibration is obtained M-113 used at Fullerton was a relatively new
on a level paved road; and vehicle with tracks in very good condition.

~115
 j7+ -

4B E

, jI,
W,

Fig. 1. Accelerometer locations

Instrumentation suspension system bottomed at 13 mph while

going over ramps spaced 14 ft apart. In addi-
Figure 1 shows the accelerometer locations tion, vibration data were recorded while trav-
for both tests. Piezoelectric accelerometers eling over an asphalt surface at 20, 25, and
were used for the vibration tests, and strain 30 mph.
gage accelerometers for the shock test. Block
diagrams for the instrumentation are shown in
Figs. 2 and 3. All instrumentation was carried VIBRATION TEST RESULTS
on the M-113 during the FMC tests. At Fuller-
ton, an instrumentation truck was driven next Broadband acceleration levels for all
to the M-113 during the tests. measurement locations are presented in Fig. 5.
Acceleration spectral densities of selected data
points are shown in Fig. 6. Included are spec-
Measurement Conditions tra for the most severe vibration for all loca-
tions, for vertical accelerations at position B
In the vibration tests, measurements were for a range of road speeds, and for position A
made on a concrete and dirt track at speeds of on a road test of a new vehicle. Amplitude den-
5, 10, 15, 20, 25, 27 (maximum on dirt), 30, anC sities of certain narrowband data are presented
34 mph (maximum on paved road). in Fig. 7.

The shock test was performed on a Munson The vibration produced by vehicle motion-
course containing wooden ramps as shown in specifically by the track engagement with the
Fig. 4. Tests were run at vehicle speeds of ground-has many strong features. As might
7, 9, 11, and 13 mph on the shock course. The be expected, the vibration levels increase with

116
ACCELEROMETERS CALIBRATION AMPLIFIERS TAPE RECORDER
ENDEVCO MODEL 2213 I PANEL EDVOAPXC-0

POWER SUPPLY
PORTABLE
GENERATOR

MICROPHONE

Fig. 2. Instrumentation, vibration test at FMG Corp.

STRAIN GAGE I AMPLIFIERS

ACCELEROMETERS (14) SA-

f I GENERATOR

M-115 VEHICLE 1IrUErr1, RC

Fig. 3. Instrumentation, shock and vibration tests. at Fullerton

117j
 30
2 28 2814..ju114' 14!-Iw:

'~'CONCRETE, i

00 0 LiE[111 11 M0
/
MUNSON COURSE WOODEN BLOCKS DIRECTION OF TRAVEL:-

29!-. 14"

48 BLOCK

Fig. 4. Size and location of wooden ramps used in shock tests

vehicle speed. There were exceptions to this in all of the data in varying magnitudes, usually
rule as the measurements made on the vehicle of higher value at higher speeds. The amplitude
deck and floor reveal (positions B and C). These distribution of the spike revealed that in the
locations exhibit their highest vibration levels vertical direction it was strongly sinusoidal.
at 29 mph. Otherwise, the levels are propor- Above 25 mph in the other directions, it tended
tional to the velocity. At almost all conditions, to be only slightly sinusoidal. At lower speeds
the longitudinal vibration was lowest. The ver- and on dirt tracks it appeared to be nearly
tical and lateral vibration showed equal prom- Gaussian in all directions.
inence, depending on the location. Positions A,
B, and C exhibit higher levels in the vertical A comparison of the vibration data taken on
direction; positions D and E had higher levels a used vehicle with that taken on a new vehicle
in the lateral direction. Postions A, B, and C reveal that the levels are slightly greater on
were located on structure that was stiffest in the the former. Acceleration spectral densities for
lateral and longitudinal directions, and positions position A were computed from data taken on a
D and E were on structure that was stiffest in new vehicle during the Fullerton shock tests.
the vertical and longitudinal direction. Exam- Because of high-frequency response limitation
ination of the vibration spectra reveal that there of the accelerometers used during this test, the
is significant high-frequency content in the data broadband acceleration levels cannot be directly
in the vertical direction for positions A, B, compared to the levels reported for the used
and C and in the lateral direction for positions vehicle. However, examination of the spectra
D and E. reveals that, in most instances, the level of the
tread frequency spike is lower for the new
Another significant feature of the vibration vehicle.
is the appearance of a strong spectral peak at
the tread engagement frequency. This frequency
is equal to the velocity divided by the tread SHOCK TEST RESULTS
length. Because of the strong excitation asso-1
ciated with this frequency, its characteristics The shocks which occurred during suspen-
must be understood. The spectral peak appeared sion bottoming as the vehicle rode over the

118 [ {
 30 _______ .0 -. ETIA.-
- VERTICAL 5. - VERTICAL
-- LATERAL II LATERAL
LONGITUDINAL LONGITUDI NAL
2.5 2.5

-IHARD POSITION A

2.0
TRACK
2.-
POSITION C
HARD TRACK
--

I N

o - - - - --. -L-

1.0

0. 0.5 L 1-

0 i ETCL30 1.

-- VERTICALVERTICAL
LATERAL -V LATERAL
2. 5 LONGITUDINAL - - .5s- ----- LONGITUDINAL - - - - 0- - -
POSITION A POSITION D /
DIRT TRACK HARD TRACK 01 Vf,
2.0 - - 2.0

C6 1.5 1.5
0 01

0 --

30 310
-VERTICAL VERTICAL
-- LATERAL -- LATERAL
.--- LONGITUDINAL --- LONGITUDINAL 0
2.5 - - 25 ,_

POSITION 8 POSITION E00

HARD TRACK HARD TRACK
2. 0 - -2.0 - -

1.51.5

1.0 - -. 0-

0.5 0.5

5 10 15 20 25 30 35 0 5 10 15 20 25 30 35

VELOCITY, MILES PER HOUR VELOCITY, MILES PER HOUR

Fig. 5. Acceleration vs velocity, all measurement locations

119
 01

U)

I.,
z

g r s 1 .4

U)

z
0.0011
_j

M-II3~~~~M- VEIL13AINMII EIL VEIVIBRATION

29~~~ ~2
MPH-HARD TRCK2RMHHADTRC
POSITIO VETICVBATIO POSITION A-RAL L

grms .4 1gis.1

0.0 100 1 000 10,00 10 100 1000

01.
 0.01

* a.
u
CL
Q

0.001

.- 00001

w
ma
U
o M-113 VEHICLE VIBRATION
15MPH -HARD TRACK
POSITION 9-VERTICAL
grins ,0.9

000001 I 1 fi ll,
10 t00 1000
FREQUENCY CPS

a.

>" 0.01
U-

4
U WI. ,
a.

Vw S0001

mai

M-113 VEHICLE VIBRATION M-113 VEHICLE VIBRATION

25 MPH-HARD TRACK 32 MPH-HARD TRACK
POSITION B -VERTICAL POSITION B-VERTICAL
grins -2.2B rms2.4?
II I
I1 fi |1 ! ! | I I
0.0001 I
10 I00 1000 10 100 1000
FREQUENCY, CPS FREQUENCY, CPS

Fig. 6. Acceleration spectral densities, selected data points (Cont'd.)

121
' '4
 S 001

- i -

CLQOO1

WIL
-I L
UM

M-II3 VEHICLE VIBRATION M-113 VEHICLE VIBRATION

25 MPH-DIRT TRACK I 32 MPH-HARD TRACK
POSITION B-VERTICAL POSITION C-VERTICAL
grms- 2.31 grmsz2 36

0.001 .1, , , ,, *.... ......... i ,ig 9

0.1 1

M-113 VEHICLE VIBRATION

32 MPH-HARD TRACK

POSITION C-LONGITUDINAL
grm$ 1.18
eg.
4N

0.01

z
WJ
_- A

I- I
o I
w

0.
0.001
U
wI

, M-113 VEHICLE VIBRATION

32 MPH-HARD TRACK
POSITION C -LATERAL
gfrms. 2.08

00001 -nn1 l!
10 100 1000 10 100 1000

FREQUENCY,CPS FREQUENCY,CPS

Fig. 6. Acceleratibn spectral densities, selected data points (Cont'd.)

 0.1

u)

I.-

z
o jh jT
w

J
w
-J
w F
w

M-113 VEHICLE VIBRATION M-113 VEHICLE VIBRATION

28 MPH-HARD TRACK 28 MPH-HARD TRACK
POSITION D-VERTICAL POSITION D-LATERAL
grins. 1.42 grms.2.74

0.0001 I1 1 i I 1 I I I oil, I I I IIIII , I I

0.1

a.
IL
0

0.04

z
w
I,

Iaa

InI

<M-11I3 VEHICLE VIBRATION -M-I113 VEHICLE VIBRATION

28 MPH-HARD TRACK I28 MPH-HAR TRACK
- ~~POSITION
D- LONGITUDINALI OIINEVRIA
grins, 1.02 POmSITO E-.22 CA
1
0.0001 I 1

10 100 4! 1000 10 100 1000

EREQJ.IENCY, CPS FREQUENCYCPS

Fig. 6. Acceleration spectral densities, selected data points (Cont'd.)

1233

. . .
 06 -
POSITION A- VERTICAL
85 CPS
28 MPH
HARD TRACK
NEW VEHICLE

0.5
-4- 1
SINUSOID

0GAUSSIAN [74

I-

02

II I

0.6 POSITION B - VERTICAL - -- ________

I
94,CPS
32 %IPH
0.5 HARD TRACK

0.4
z SINUSOID

F-u GAUSSIAN
., 0.3 A

0 0.2
I-

0.1

-4.0 -3.0 -2.0 -,rv -1.0 0 1.0 V2 2.0 3.0 4.0

/grms

Fig. 7. Amplitude densities

124
 Munson course are less predictabIe than the vi- 1. Generally, the vibration levels are pro-
bration environment. Shock levels did increase portional to vehicle speed;
with velocity, with the highest levels recorded
at 13 mph. Generally, the shocks recorded 2. The vibration levels are of comparable
from test to test represented the same velocity magnitude in all directions;
change with different combinations of duration
and level. The highest shock level measured 3. At all speeds, the vibration ha.3 a broad
was 32 g and had a pulse duration less than frequency spectrum and a strong peak at the
2 ms. Longer duration shocks had correspond- tread frequency;
ingly lower levels. As in the vibration tests,
measurements on the floor and deck of the ve- 4. The spectrum at frequencies other than

I
hicle had levels which were highest in the ver- the tread frequency has a Gaussian distribution;
tical direction, and measurements made on the and
side wall had their highest levels in the lateral
direction. Shock spectra of son.e typical pulses 5; At high vehicle speeds, the tread fre-
are shown in Fig. 8. quency spike was strongly sinusoidal.

The level and pulse duration of the shocks

CONCLUSIONS which occurred during.suspension bottoming
varied from location to location and test to test
As the results show, the vibration of the M- 113 and, therefore, cannot be predicted with accu-
vehicle is significant. Important characteristics racy. However, significant levels were meas-
which can be concluded from the test are: ured in all directions.

REFERENCES

1. V. A. Eddington, "Road Shock and Vibration Food Machinery and Chemical Corp., San
Test of Armored Personnel Carrier M-113 Jose, Calif., Rept. ORD 619, Dec. 27, 1960
(Vehicle Response Investigation for Mauler
Weapon System)," Aberdeen Proving Ground, 4. "Vibration and Shock on the XM474E2 F1
Rept. DPS-391, Dec. 1961 Missile Equipment Carrier," Food Machin-
2. R. P. Lenert, "Noise and Vibration Test of ery and Chemical Corp., San Jose, Calif.,
M-113 Armored Personnel Carrier with Rept. ORD 621, Jan. 25, 1961
Standard and Short Pitch Track," Aberdeen
Proving Ground, Rept. DPS/PTA-166, Sept. 5. "APC M-113 Full Vibration Induced by En-
1962 gine Operation with Vehicle Stationary,"
3. "Vibration and Shock Environments in Ve- Food Machinery and Chemical Corp., San
hicles of the Airborne Multipurpose Family," Jose, Calif., Rept. ORD 557, Jan. 25, 1960

Appendix

NOTES ON DATA ANALYSIS

ACCELERATION SPECTRAL AMPLITUDE DENSITY

DENSITY
In performing the amplitude density analysis
~A swept filter analysis was used with the the data are passed through a broadband filter
analysis system consisting of a Spectral Dy- (half-power points at 17.5 and 7000 cps) to a
namics SD101A for the analyzer and a Balantine 1/10-octave bandwidth filter. The filtered data
true rms meter for the detector. The following are fed to a bank of 10 amplitude discriminators.
were the analysis parameters: filter bandwidth, These discriminators are biased in equal in-
5 cps; loop length, 10 seconds; sweep time (20- crements of full scale, and the time that the in-
2000 cps), 15 minutes; and averaging time con- stantaneous amplitude of the data is within each
stant, 3 seconds. of these discriminators is determined. These

44

125
 25

-. ~- - ---------

\-----------

,
..

-25 1 _ 1__ _ 1__ _ 1__ _ 1__ _ 1__ _ _ _ _ _ 1__ 1__

S 25
POIIN -LAEA APN
w3P 1

-25

< -12

-n20 -

- - - - e------------
 POSITION
D-VERTICAL '%

-10

-0

-30

-501

O.LATERAL
POSITION 1

S60

-20

-10

560

-OSTI0 D-___
LONITUINL _IN

I'
~~~ 30____
10 - - _
------- 7.-

-30

.501- T
0 50 100 150 200 250 300 350 400 450 500
FREM)ENCY, CPS

Fig. 8. Shock spectra, typical pulses (Cont'd).

127

00I
times, divided by the total analysia time, are SHOCK SPECTRA

normalized with respect to the amplitude of the

equal discriminator increments. The amplitude The signal from the magnetic tape record-
increments are previously normalized with re- ing of the acceleration is fed into a series of
spect to rms value of the data. Therefore, the electrical analogs of lightly damped, single de-
area under the amplitude density curve equals gree of freedom systems. The peak negative
the probability that the instantaneous value of and positive outputs from each of these analogs
the data is within the amplitude limits of that is the response used to plot the spectrum. The
area. damping values used in the analysis were 1, 2,
and 5 percent of critical damping. These result
in amplification factors (Q) of 50, 25, and 10,
respectively.

DISCUSSION

Mr. Scott (Sandia Corp.): What was the load Mr. Tinling: When the vehicle was driven
inside the M-113 during these tests? over the wooden ramp the suspension would
bottom and there would be an input velocity
Mr. Tinling: It was loaded with a dummy change.
of the weapon system that was going to be in- Mr. Forkois: This is what I suspected. Do
stalled. you know approximately the g value attained?
Mr. Tinling: It varied quite a bit. A maxi-
Mr. Scott: Was it heavily loaded, lightly mum level of 39 g of about 2 ms duration was
loaded, or practically empty? observed. Most levels were lower-10 to 15 g I
with 5 to 8 ms duration. X
Mr. Tinling: I do not-know how to classify Mr. Marshall (Aeronutronic Division): What
it. was the relation between the bed measurement
and the deck measurement in terms of vibration
Mr. Forkois (U.S. Naval Research Labora- level?
tory): What was the phenomenon connected with
the shock inputs? Mr. Tinling: Fairly similar.

128

I
 INTERPRETATION AND APPLICATION OF SPECIFICATION
REQUIREMENTS THAT SIMULATE VIBRATION RESPONSES
OF EQUIPMENT BEING SHIPPED BY COMMON CARRIER
James E. Rice
Goodyear Aerospace Corporaticn
Akron, Ohio

The object of this pap-r is to encourage investigations which will show

that vibration resonance dwell test requirements that simulate loads
related to transportation by common carrier are not realistic for cer-
tain conditions. It is felt that the test frequency ranges and g inputs
are too high for items weighing 1000 lb or more. Of the various cri-
teria currently used to establish transportation load criteria, MIL-
STD-810 was chosen for discussion, since it clearly displays the
problem.

, By examination of the dynamic characteristics of the common carriers,

it can be shown that steady state vibration loads at values large enough
to cause damage are not present at frequencies greater than 30 cps for
items weighing 1000 lb. It is felt that the criteria as now presented
can encourage poor designs and introduce unnecessary costs of pro-
ducing military hardware for the weight range considered.

INTRODUCTION 5. What is the purpose of introducing the

problem at this symposium, rather than directly
A currently accepted standard used by gov- contacting the Aeronautical Systems Divisiun,
ernment agencies and industry is MIL-STD-810. Wright-Patterson Air Force Base, Ohio?
It is felt that the criteria governing vibration
testing to simulate shipment by common car-
rier can be interpreted in such a manner that WHAT IS THE PROBLEM?
poor designs are encouraged and good designs
will be penalized by increased costs which are A situation frequently encountered is that
felt to be unnecessary. Since the Department of personnel who are not dynamics specialists
Defense is making a determined effort to reduce have the responsibility of monitoring contracts
unnecessary costs, every avenue should be ex- or awarding contracts, either in government
plored so long as reliability is not jeopardized. agencies or as prime contractors. Their guide
Also, any interpretation that could encourage must be the current standards. The criteria
poor design must be carefully studied. that reduce input frequency with increasing
T g r i c qweight do not provide guidance to show that
To bring the problem into focus as quickly 1000-lb objects will respond dynamically in a
as possible, the following questions must be fardifferent manner than400-1b objects. Fig-
raised: ure 514-8 of MIL-STD-810 (1) permits a de-
crease in test frequency with weight, but for
1. What is the problem? items weighing 300 lb or more, the upper fre-
M-r quency level is 50 cps. Should this be so for a
2. Why is it a problem? 1000-lb item or a 5000-lb item? Figure 514-8
3. What are the consequences of the inter- does not discriminate, and nontechnical moni-
Spretation to be discussed? tors can only assume that 50 cps is the cutoff
limit. It is felt that this upper frequency cutoff
4. Will analysis or measured data support should be reduced to 30 cps for items weighing
the viewpoint to be expressed in this paper? 1000 lb or more.

129

-r
 The design problem is the resonance dwell are contained in frequencies below 26 cps for a
requirement of 1/2 hour at each of four natural 1000-lb object, so the soft structure could con-
frequencies for each of three axes with an input tinually be excited at resonance in service,
as large as 5 g. Generally speaking, large whereas a stiff structure would not. This en-
complex consoles, such as ground support courages poor design. Obviously, those aware
equipment, will respond at the same resonance of the input frequencies keep the structural nat-
frequency in each of the three axes, due to cou- ural frequencies above 26 cps, but with the re-
pled modes, so it is not uncommon to subject suiting penalties in cost. P
the equipment to 1-1/2 hours of resonance at
50 cps with a 5-g input and a 25-g output. Only
those who have had the experience of seeing WILL ANALYSIS OR MEASURED
this response for a 1000-1b console, loaded with DATA SUPPORT THE VIEWPOINT
heavy pumps, chassis on slides, dials, gages, EXPRESSED IN THIS PAPER?
and mazes of hydraulic tubing, can appreciate .

the severity of such an environment. The con- Newton's law, F = ma, requires a steady
crete floor shudders, and 130 db of acoustic state alternating force of 5000 lb to induce a
noise is generated by the radiating sheet metal 5-g response for a 1000-lb object. Also, action
side panels of the console. equals reaction, so the item being shipped will
apply these loads to the carrier. The common
carriers do not supply such large steady state
WHY IS THIS A PROBLEM? forces at frequencies above 30 cps. Shocks can
induce a high-frequency 'shock spectrum on
The problem created by specifying such equipment being shipped, but these are decaying
high g levels at such high frequencies (1) is that transients and do not justify resonance dwells
the development costs to design a structure to for such long periods with a 5000-lb input. If
withstand the loads are substantial. Also, it is shock spectrum response is the problem, we
usually not possible to buy "off-the-shelf" should require proper shock testing in place of
equipment that has been qualified to vibration resonance dwells at high g inputs.
levels greater than 15 g, so the cost of requali-
fication must be borne. This involves large A number of excellent papers have been
sums of money. presented at recent Shock and Vibration Sym-
posia which show that both the input g levels
and the cut-off frequencies are questionably
WHAT ARE THE CONSEQUENCES high when 1000-lb objects are being shipped.
OF THE INTERPRETATION Two papers presented at the 31st Shock and
TO BE DISCUSSED? Vibration Symposium describe investigations to
establish dynamic loads in cargo airplanes and
A serious consequence can be the design- trucks. In the test reported in the first paper
ers' response to the requirement. If the pri- (2), pickup's were placed on a 1000-lb bomb and
mary structure has four ratural frequencies on structure adjacent to the bomb. No response
less than 26 cps, the specified vibration input greater than 0.5 g was measured on the bomb at
levels are 1.3 g. Even with an amplification frequencies greater than 30 cps. The pickups
factor of 5, the induced vibration level on corn- on adjacent structure responded at 1 g at fre-
ponents is only 6.5 g for the resonance dwell quencies up to 500 cps. This 1 g merely re-
periods. This simplifies the procurement flected the fact that light-weight structure will
problem for component assemblies. Also it is respond at high frequencies but with little force.
much simpler and less expensive to design a The second paper (3) described an analysis of
large structure with natural frequencies less data to try to correlate the vibration require-
than 26 cps than to design the structure to have ment of MIL-E-4970 by test. The tested item
no natural frequencies below 50 cps (to avoid under consideration was a heavy refrigeration
resonance dwells). Some agencies define a unit. The conclusions reached were that even
resonance as a dynamic response with a Q = 2. over the severe Munson Road course, the max-
This means if the primary structure had a low- imum value at any one frequency did not exceed
est natural frequency of 60 cps, and was fairly 1.1 g, and even the 3 sigma values did not ex-
well damped, a Q = 2 could be experienced at ceed 2.5 g. The requirements of MIL-E-4970
50 cps because of the broad response curve, Procedure VI are almost identical to MIL-STD-
thus necessitating a resonance dwell. 810 for the common carrier vibration test.

It will be shown later that the steady state The tests at Aberdeen Proving Ground and
forces from common carriers that excite many similar tests show that a truck responds
shipped items at their resonance frequencies with a steady state vibration input at one of two

130

-Awl
 natural frequencies which are the "sprung be more readily identified. Also, if further
weight" which varies from 1 to 5 cps and the testing is conducted on trucks and trains, in
'"unsprung weight" (or wheel hop) which varies which both the response of the shipped item and
from 12 to 17 cps (4). Since the road surface nearby structure is measured, the cutoff fre-
is best described by power spectral density quencies can-be established for trucks and
(PSD) measurements (5), the truck responds trains as they were for airplanes (2). With this
only at these two natural frequency ranges, as information, it is felt that test requirements to
is characteristic of response of dynamic sys- simulate shipment by common carrier can be
.tems to PSD inputs. The Munson Road course examined to see if the present requirements
excites the same frequencies. There is a third should be changed.
frequency range, the body natural frequencies,
for trucks from 60 to 120 cps. These frequen- The use of shipping containers has not been
cies have been measured, but the force to ex- discussed so far. If properly designed, they are
cite 1000-lb objects is not present. Usually the an excellent solution to the transportation prob-
input forces at these high frequencies are lem. Special shock mitigating rail cars are
shock-induced transients with little power to also very effective.
drive heavy items.

Steady state vibration inputs from ships WHAT IS THE PURPOSE OF

and trains are at frequencies less than 30 cps. PRESENTING THIS PROBLEM
As previously stated, the dynamic characteris- AT THIS SYMPOSIUM INSTEAD
tics of all common carriers that could cause OF CONTACTING ASD DIRECTLY?
failure from steady state inputs occur at fre-
quencies less than 30 cps. The answer is as simple as this: it is a
serious responsibility to change a specification
The Shock and Vibration Handbook (4, pp. or standard, and personnel with this responsi-
45-29, 45-30) indicates that the steady state bility need sufficient facts and agreement from
vibration input of trains is in the frequency experts to warrant a change. The attendees of
range less than 10 cps with a maximum steady this Symposium represent experienced dynami-
state input of 1.3 g. The vibration input com- cists, familiar with all aspects of the problem,
mon to ships is shown in Fig. 46.2 (4, p. 464)" - and heads of government agencies that perform
and is 0.3 g at frequencies less than 20 cps. testing that can show the needed facts and pro-
vide expert interpretation of the data. In addi-
There is little doubt that shock loads are tion, those who disagree with the basic premise
experienced by items being transported by of the paper will present other viewpoints which
trains and trucks. The shock loads from trains should be considered. It is hoped that sufficient
are due to humping and cargo shifting resulting interest has been generated to initiate such
from sudden velocity changes. The large ac- healthy discussion, so that studies can be initi-
celerations experienced by items being shipped ated to clarify areas of disagreement.
by trucks can be either shocks due to chuck-
holes or bumps in the road, which cause the It is not the purpose of this discussion to
shock struts to bottom, or to the vibration loads criticize MIL-STD-810, which represents an
occurring 1 percent of the time. Aberdeen advancement in a most difficult field. Junker (6)
Proving Ground places the measured accelera- made the following statement at the 31st Sym-
tions on trucks into three categories: single posium which is very appropriate to the prob-
maximum acceleration, accelerations exceeded lem: "It is essential that constant surveillance
1 percent of the time, and rms accelerations. and liaison across the vast Military-Contractor
effort be maintained so that MIL-STD-810 can
It is felt that the damage experienced by be kept up to date. If, through the use of this
lading is due to these large loads whether they Standard, deficiencies are noted or if new cri-
be clearly defined as the single maximum shock teria and methodology are developed for which
or accelerations experienced 1 percent of the there is a general need, it is suggested that
time in trucks. Either of these can induce high- recommendations be made for upgrading the
frequency response on the carrier or in the Standard."
shipped item. After the high-frequency compo-
nents are separated from the major pulse (3), It is with this thought in mind that this
the signature of the damaging frequencies can paper is presented.

131
 REFERENCES
1. MIL-STD-810A, "Military Standard Envi- 4. C. M. Harris and C. E. Crede, "Shock and
ronmental Test-Methods for Aerospace and Vibration Handbook" (McGraw-Hill, New
Ground Equipment," June 1965, pp. 514-2, York), 1961, Vol. 3, Ch 45
514-11 5. J. R. Harvey and R. A. Wursehe, "Rough-

2. C. E. Thomas and P. G. Bolds, "Effects of ness Measurements and Missile System

Weight on the Frequency and Amplitude of Response Evaluation for Highway Environ-
Vibration Test Items," Bulletin No. 31, ment," 35th Symposium, October 1965
Part II, March 1963, p. 221
6. Virgil J. Junker, "Environmental Testing
3. R. D. Baily and J. W. Apgar, "Preparation Standardization Via MIL-STD-810 Environ-
and Analysis of Munson Road-Test Tapes mental Test Methods for Aerospace and
for Laboratory Vibration Tests," Bulletin Ground Equipment," Bulletin No. 31, Part
No. 31, Part II, March 1963, p. 64 I, March 1963, p. 36

DISCUSSION

Mr. Schell (Air Force Flight Dynamics people, we just do not have enough time to de-
Laboratory): Although it is not clear in the vote to these things. If you have some good
specification, it was never intended that the vi- suggestions we will be glad to try and incorpo-
bration test requirement of MIL-STD-810 be rate them. As far as the test levels in the doc-
applied to equipment weighing more than 1000 lb. ument are concerned, they are based on envi- 2

ronmental measurements. At that time we did

Mr.Rice: Itistrue that people who moni- not have any impedance techniques and I do not
tor programs frequently do not intend that the think impedance techniques today are well
test be extended to these weights. I know this enough developed to apply to this problem. If
is true because I have talked to Mr. Thomas there are some who disagree with me, I would
and Mr. Earls and others at Wright-Patterson like to hear their arguments.
Air Force Base. There are people, however,
who interpret the specification as allowing its Mr. Rice: Thank you, Mr. Schell; I am so
application out to 1000, 2000, or even 5000 lb. happy to hear your comments. I do not want to
It is for these people that I wish a definition, fall into the category of those who merely criti-
cize. I on!7 want to help. I have had to ship
Mr. Schell: I wish that we could write a many big objects, and I want to try to pass on
perfect specification, but I do not know of any- the experience I have had. It confuses my most
body who has managed to do this yet. As these learned friends when they do not have the meas-
points come up we try to take care of them. ured data in their grasp. There is something in
This might be a point that we can take care of the system that vexes and frustrates me a great
in the present revision for tri-service coordi- deal. If I know that a specification must be
nation on the document. In my opinion, MIL- changed and can get to my counterpart I could
STD-810 ought to be divided into weight cate- change it in one day. But there is a rule that
gories, such as 1 lb, 1-10 lb, 10-50 lb, 50-100 you must spend $50,000 and 3 months to prove
lb and so on, for dynamics testing purposes. that God and Newton are wrong, and then they
Since we are limited, however, in money and will talk to you. (Applause.)

132
 SHOCK AND VIBRATION ISOLATION

NEAR-OPTIMUM SHOCK MOUNTS FOR PROTECTING

EQUIPMENT FROM ACCELERATION PULSES
Ralph E. Blake
Lockheed Missiles and Space Company
Sunnyvale, California

In this theoretical study, the properties of a mount are defined by the

dynamic force-deflection curve. To relate mount properties to strength
required in equipment, an "index of severity" is developed from the
normal-mode theory of structural response. The best mount is the one
which minimizes the index of severity of the shock transmitted to the
equipment.

The mount force-deflection curve was fixed except for three shape pa-
rameters. The problem of finding the optimum set of parameters was
then programmed for a computer. Computer input consists of the dis-
turbing shock pulse, the maximum allowed mount deflection, and some
assumptions as to expected dynamic properties of the equipment. The
resulting solution is "near-optimum" in that the shape was constrained
except for three parameters and typical properties of the equipment
were assumed. However, comparison with a theoretical lower bound
on the index of severity suggests that there is not room left for great
improvement.

Shape parameters, indices of severity, and shock severity spectra are

reported for several cases. Shock severity spectra are similar to
shock spectra but are modified to give more emphasis to vibratory re-
sponse in equipment.

The extent to which assumptions can be relaxed to deal with similar

problems of interest has to be examined. The shock disturbance was
assumed to be a pulse, without reversal of sign. No modeling of human
comfort or injury criteria was included. Mount deflection was as-
sumed to be limited by system design problems, but trade-offs of
space, weight, cost, reliability, etc., involved in system optimization
were not introduced.

INTRODUCTION and nylon line. Sources of shock include blast

from explosions, parachute drops, accidental
A shock mount is a flexible nonlinear de- drops of shipments, lunar landing, aircraft
vice used to reduce the damaging effects on an landing, freight car humping, and bumpy roads.
equipment of a shock or impact. It achieves With so many alternatives, the question arises
this by trading space to reduce the force neces- as to whether one mount design is better than
sary to accommodate the equipment motion to another.
the base motion. Mounts now in use can be
found in an endless variety of design concepts, More than 10 years ago, this question was
materials, and applications. They include rub- attacked directly at the Naval Research Lab-
ber pads and bumpers, metal springs, air and oratory by testing mounts with a special ap-
hydraulic springs and shock absorbers, friction paratus on the Navy shock machines (1). Figure
brakes, yielding and buckling elements, plastic la shows dynamic load-deflection curves ob-
foam, crushable honeycomb, rubberized hair, tained for four mount types, and Fig. lb shows

133
 0 A:
.A.

Fi. /MOUNT B
As

FREQUENCY
(o) DEFLECTION Z

Fig. 1. Force-deflection-curves and shock spectra

from Navy shock machine

the corresponding shock spectra of the motion NOMENCLATURE

of masses supported on the mounts. Mounts A
and B were overloaded and showed force peaks ao Acceleration of rigid box caused by
due to bottoming, Mount C was fairly linear, and constant-force mount
Mount D produced a constant force over much of
its deflection. Judging these mounts by their c/c, Fraction of critical damping
shock spectra, Mount D was the best. But why
was it best? Would it be possible for some d Geometric progression ratio in index
other force-deflection curve to produce an even computation
lower spectrum? What is the best possible
mount load-deflection curve? F Mount force

About a year ago, we started a study at fn Frequency in cps of nth normal mode
Lockheed for the Air Force Ballistic Systems
Division to define the potential benefits of g Acceleration of gravity
shock-mounting equipment. The results to be
reported concern a theoretical investigation to I Index of shock severity
define the best possible load-deflection curve
for a mount and the severity of shock trans- J Static influence coefficient
mitted. The design of hardware to approximate
the ideal will not be discussed here; a complete k,K Spring stiffnesses
report is given in Ref. (2).
,L Lengths
First, the theoretical situation and param-
eters of the shock input and the mount are de- m,M Masses
scribed. Then normal mode theory is employed
to justify a new quantitative measure for the P Force on base
severity of shock transmitted to equipment.
This "index of severity" is a weighted average P1 ,P2 ,P 3 Parameters of mount force-deflection
of a modified shock spectrum which takes ac- curve
count of structural "tuning" effects in the equip-
ment. Next, the relation of some possible mount q "Shock-resonance" or "tuning" factor
characteristics to the index of severity is in- in index
vestigated. On the basis of these results, some
key munt parameters were selected and a r Any response of a linear elastic
computer program set up. The computer finds structure
the combinations of parameters which minimize
the index of severity for a given input shock. Rn Normal-mode participation factor
Some optimum load-deflection curves and their
corresponding severities are shown and com- S Maximum deflection of shock mount
pared with a theoretical limit on the possible
protection. So Displacement of base at time TO

134
 t Time BOX Y
S
T Time of maximum deflection
F
To T of constant-force mount
K Vn Vibratory response factor of nth
normal mode
v Velocity of base at time T BASE 9

Xx, X Displacement, velocity, acceleration

of base PWt

j Y,Y, ,Y Displacement, velocity, acceleration, Fig. 2. Forces andmo-

and jerk of box tions in shock mount
problem
z Deflection of shock mount

INPUT SHOCK AND ASSUMPTIONS 11ASEX

All shock motions and forces will be as-
sumed to be parallel to one direction. This 0
does not seem to be a serious restriction; the
notation is much simpler and most cases are .J Box V
-essentially one directional.
Figure 2 shows an equipment "box" attached _

to a "base" by a mount. Some exciting force T!ME t T

P(t) causes an acceleration R(t) of the base. It v
will be assumed that the force F(t) exerted by so
the mount has no effect on R(t). Furthermore,
the box will be assumed to bi- rigid only for the U
purpose of describing the changes in mount I Box
height z(t). These assumptions could be re- >
laxed somewhat by assigning local flexibilities
of the box and base to the mount and by consid-
ering the effect of box mass on the rigid-body (b) TIME t T
motion of the base. But the general problem of Fig. 3. Acceleration and ve-
including mechanical impedances of the sys- locityhistories of box and base
tems is not dealt with here. Thus, the mount
K must be far more flexible and deform much
further than the structures. It is also expected
that a flexible mount will help to justify the as- SHOCK MOUNT PARAMETERS
sumptions by having a minimum tendency to
excite or transmit structural vibrations. Figure 3 also shows rigid-body accelera-
tion Y(t) and velocity Y(t) time histories that
The base motion R(t) is an acceleration a box might have as the result of the force F(t)
pulse; that is, it has one sign and does not re- of the mount. The deflection of the mount is the
verse or oscillate. Although it may be possible integral of the difference in the two velocities,
to deal with some oscillation, as in the case of
the Navy shock machines, this has not been ex- t
plored; the study by Eubanks (3) appears to bar Z(t) = (k(- i)dt(1
us from handling the most general cases. Thus,
earthquake shocks are excluded but impacts, 0
drops, blast pulses, etc., are included.

t An acceleration which does not change sign

causes a continuously increasing velocity. Thus,
At time T, the two velocities are equal and the
maximum deflection S occurs. Since it will be
seen that the severity of shock to the box can
in Fig. 3, the acceleration x(t) corresponds to be reduced indefinitely if S can be increased
the velocity (t). indefinitely, S is an important parameter.

135

I"1111 7"
 In most designs, S cannot be increased solution in the mathematical sense, since this
without penalty in space, weight, cost, etc., so does not require that the assumptions be the
the optimum mount will be the one which makes best possible approximation to reality.
the best use of an allowed deflection, s, to min-
imize the shock effects of F(t) for a given in- The next task is to set up a general mathe-
put X(t). Since the weight, reliability, cost and matical model to represent the response of the
space for possible mount hardware designs will majority of equipments. This model is tentative,
be different, the mount characteristics may not allowing for adjustment to suit improved infor-
be optimum from a broader system viewpoint. mation on typical box characteristics and also
In this first attempt, such system trade-off to see how sensitive the answer is to errors in
problems will be omitted. assumptions.

.After meeting its obligation to bring the The basic assumption is that the box has
box velocity up to that of the base without ex- distributed mass with linear elastic springs up
ceeding S, the mount force is still subject to to the point of failure. With the further assump-
some requirements. But these are usually tion that damping is small, viscous, and uni-
minor and can be met without exceeding the formly distributed, it can be asserted that the
strength required of the box up to time T. response of the box to the force F(t) is de-
Hence, the basic problem is to find the box scribed by the normal-mode theory. The as-
motion Y(t) up to the time T (when i-- k and sumption of linear behavior is reasonable for
z= s), such that the chance of damaging the box most structures up to the point of buckling,
is a minimum. Given Y(t), it is a simple mat- yielding, brittle fracture, slipping of friction
ter to plot force F versus deflection z. bonds, etc. Although there are some exceptions,
the great majority of failures can be related to
the exceeding of some critical mechanical quan-
INDEX OF SHOCK SEVERITY tity such as force, stress, strain, and displace-
ment. Many of these failures are preceded by
To pick out the best Y(t) curve from the elastic linear behavior of the box structure, as
infinite number which satisfy the space require- in the fracture of brittle materials, opening of
ment, some criterion for preferring one to relays, contact or arcing of electricai elements,
another is required. The criterion of minimum permanent set, slippage of mechanical joints,
peak acceleration I( t)l max is overly simple and collision of adjacent parts. In the cases of
and leads one straight to a constant-force mount, fracture after plastic deformation and elastic
This mount produces severe vibrations of the buckling, the region of the box involved in the
box due to the initial step discontinuity, as will failure will be nonlinear, while the remainder
be seen later. is elastic. In th(" ases of fatigue and gyro drift,
failure is due to an accumulation of effects.
The shock spectrum was used for the Navy However, fatigue is not important, since only
shock mount evaluation (1). This takes vibra- about 10 oscillations occur in response to shock,
tion, as well as a peak acceleration, into ac- and gyros are special cases. For plastic de-
count, but it will be shown that a modification formations and elastic buckling, the conserva-
to increase the relative importance of vibration tive assumption can be made (as in static stress
will be desirable. analysis) that the start of nonlinearity is synon-
ymous with failure. Thus, it is considered that
The problem to be considered here is not an assumption that the box is linear and elastic
that a specific box has been designed and tested up to the point of failure will be sufficiently
and all of the needed dynamic characteristics realistic for practically all cases of interest.
and conditions when failure occurs are known.
This is a rare situation. Instead, we must con- The assumption of linear behavior is, of
sider a whole population of possible box char- course, the starting point of the well-known
acteristics and ask for a mount having the high- normal-mode theory of structural response
est probability of protecting a randomly selected (3, 4). In the appendix, normal-mode theory is
equipment from failure. This would then be the used to show that the response of any of the
best standard mount to try when an exhaustive quantities (force, stress, relative displacement,
tailoring to the box peculiarities is not worth- strain, or acceleration) at any point in the box
while. Such custom-fitting can improve on the is related in the disturbing force F(t) by an
standard mount, but not very much, as will be equation of the form
shown later. Thus, the mount has been called
"tnear-optimum" because it could be improved O
on when details of the box and systems trade- r(t) - (t) E 1n - R Vn (f 2 . t) (2)
offs become known. But it will be an optimum n=1 n1=

136
 where, with small damping, we can regard a particular box design and its
R. values as a random sample from the popu-
lation of all possible designs, and we can rea-
Vn(fnt)
j0
(- T / e soal rfrY()t 2
that (ri)max will be less than
(t) if it is probable
(r2)m, for abox

w
.+ence
cos 2-nfn(t

and Y(t) is the acceleration that the force

Swould have produced if the box were rigid,
- -)d"
selected at random. Random selection means
simply that the box was chosen without refer-
to its actual values of R n and fn.

The next problem is to muster all the in-

r(t) is the response quantity in the proper formation possible concerning the various pos-
dimensions, and f0 , c/c, and R, are deter- sible values of fn and Rn . The assumptions
mined by the details of the box structure and selected here represent our judgment and can
the locations where r is observed and F is be revised to suit. Little can be said about how
applied, the natural frequencies are scattered over the
frequency spectrum above the fundamental
As seen in the appendix, R. is very diffi- (lowest) natural frequency, but it is generally
cult to compute, even when a box design is avail- true that a box is fairly compact and rigid.
able. But the problem at hand is to consider Thus, it could be assumed that the box has no
how to change the mount to minimize the peak natural frequencies below 100 cps; or to be
value of r(t). In Eq. (2) the values of R, are precise, natural frequencies below 100 cps are
unknown but fixed by the box design; Y(t) is due considered very improbable. Since the results
to the mount force and can be manipulated by will be dimensionless, the lowest natural fre-
the mount design. The approach to be used here quency, f , need not be selected until more is
is to ask what shape of Y(t) would be best for known of the box.
minimizing response in the box. Minimum re- As one looks at successively higher nat-
sponse is considered to be synonymous with the ural frequencies, it should be reasonable to
minimum strength requirement for the average anticipate some trend in the value of R and the
box, maximum chance of a given box surviving, probability that a frequency range will be asso-
or maximum shack severity that can be with- ciated with failure. Since the lower natural fre-
stood. quencies generally involve flexure of the chassis

structure, whereas failure to function involves

According to Eq. (2), the response at some some component, the typical natural frequencies
t is made up of terms directly proportional to of components are considered to be more sig-
(t) and terms in which the variations of the nificant. On the basis of very limited and un-
have an integrated effect. Examination of the systematic data, we anticipate that the range
integral V shows that whenever Y(t) is not 100 to 500 cps is most important, 500 to 1000
varying (that is, (dY/dt) o),
o the integral is eps is next, and over 1000 cps
ps netn
A ov e ws isslect
least impor-
i ch
simply a damped sinusoidal vibration at the
frequency fn. Thus, if one waits a sufficiently tant. A weighting scheme was selected which
lon cyme af"th ifbecoes cotasufnt, gave diminishing value to vibratory terms from
long time after Y( t) becomes constant, the 100 to 1000 cps and no weight above 1000 cps.
oscillations will decay to zero and The final result is not expected to be very sen-
sitive to the weighting scheme because vibra-
Ction values tend to rise or fall together as Y(t)
r(t) =i(t) R. = J Y(t). (3) is changed and good 1(t) shapes cause rela-
n- tively little vibration above 1000 cps.

Actually, it is unnecessary to use normal-mode Finally, we come to the question of whether

Stheory to compute the r(t) of Eq. (3). The static or vibratory terms are more important.
stress distribution does not vary with time and A common assumption is to consider that a
could be computed by the methods of static mount producing the lowest peak value of Y(t)
stress analysis. J is just an influence coeffi- is the best (3, 6, 7); vibratory response is ig-
cient. Thus, we shall call Y(t) J.the "static nored. This assumption has been disproved by
'1 response" and each term of the right-hand studies of the simple two-mass model of a box
series of Eq. (2) the "vibratory response of the shown in Fig. 4. Since the box has only one
nth mode." natural frequency, Eq. (2) reduced to

Since, in general, the terms in Eq. (2) reach r(t) = R1 ((t) - V( f, t)} , (4)
maxima at different times and the Rn are not
known, it cannot be proved in an absolute sense and Y(t) and the integral, v, will have equal
that some curve Y1 (t) is better than V2 ( t). But weight. A plot of the peak values of the term in

137
 M depended on the ratio m/M and the ratio of fre-
. quencies (the ratio vq7m to /v~k-7). The largest
M R, values occurred when the frequencies were
Fig. 4. Two-mass box equal and the upper part of the system was said
MOUNT on shock mount to be "tuned" to the lower system. When tuning
existed, the small mass had a "shock resonance"
SBASE with the large system. The ratio of peak load
in spring to the peak of static load was calculated
for the case of a step-force input. This ratio
depended on the mass ratio m/2M and the system
percent of critical damping c/cc, as shown in
brackets as a function of natural frequency Fig. 6. The factor for a two-mass analysis
(e.g., Fig. 1) is known as a "shock spectrum." (k = c) would be 2 or less.
Studies of various possible mounts have shown
that shock spectrum values less than twice the
peak value of Y(t) are easily achieved, pro- 20
vided the natural frequency fn is well above the
frequency dominating the vibration of the box
on the mount. Several possible mounts give a
vibratory component in Eq. (2) of less than 10 15
percent.

However, a slightly more complex model is

sufficient to show that Eq. (4) is misleading and rmax 10
that vibratory components deserve more weight. rsttic0
The three-mass system in Fig. 5 has two oscil-
latory modes, so the response of spring k will
be c/crO
5 .005
r(t) = Y(t)(R 1+R 2 )-RV 1 (fl,t)-R 2V 2 (f 2,t) . (5) .01
.05
Now, it happens that the R's may be either
plus or minus in sign; for load in the spring, k, I I I I I I
one is plus and the other mninus. The signs of 0 .01 .02 .03 .04 .05 .06
the oscillatory terms tend to alternate with time m/2M
so that they can add together with the same sign
even though R1 and R2 are subtracted in the Fig. 6. Shock response ampli-
static term. It is easy to see that, for any num- fication by structural tuning
ber of degrees of freedom greater than the one
represented in Fig. 4, the vibratory terms can
overwhelm the static term. Another illustration of this phenomenon is
a rod subjected to an axial step force at one
end (Fig. 7). It is well known that a step wave
of force propagates along the rod with unchanged
C "
magnitude; that is, at a point distance t from
m
k[
Sk the far end, the force is F(,t) = F. The static
MY] force is obviously F t/L, so that the ratio of ac-
K tual force to the static component is L/t.

Fig. 5. Three-mass box

MOUNTE - on shock mount L

BASE
P(t) Fig. 7. Rod subjected to
axial step force

The relative importance of static and vi-

bration response was indicated in Ref. (8). The Both the rod and the tuned three-mass sys-
factors R1 and R2 for the three-mass system tem are special cases which are not typical of

138
 actual structure. It was also shown in Ref. (9) criteria must be added, and judgment and corn-
that the high response due to tuning occurred in mon sense must be applied to arrive at final
a system of 20 masses, when the small system recommendations. It is considered better to
was tuned to a mode of vibration of the large use approximations and judgment with a realis-
one. Even so, tuning is not a typical situation; tic model of the situation than to achieve mathe-
it requires the accidental matching of natural matical precision through an over-simplified
frequencies. On the other hand, failure of equip- model.
'4! ment is not a typical situation and it is plausible
to suggest that tuning is a common cause of
failure. Unfortunately, evidence for or against CONSTANT-FORCE MOUNT
this hypothesis is lacking.
The mount which exerts a constant force
We have indicated that the shock spectrum from time 0 to T provides a valuable basis for
is an unsatisfactory index of shock severity be- comparison of other mounts. Figure 8 corn-
cause it gives equal weight to the static and vi- pares the constant-force mount (dotted) with
bratory components of response. The vibratory another Y(t) having the same deflection S. The
component should have greater emphasis; this shaded area is zero so the dotted curve must
can be provided by a factor q in somewhere have higher slope (acceleration)
than the constant-force line. It follows that any
mount with the same deflection must have a
r(t) = R,{(t) ' + qVn( fn' t)) (6) higher peak acceleration than the constant-
n= force mount. Furthermore, since any index of
severity exceeds the peak acceleration, the
There is little to indicate the most suitable constant-force acceleration is the lower bound
value of q to be used in the comparison of pos- of the index for all possible mounts.
sible shock mounts. For the present, provision
is being made in the computer analysis for var-
ious values of q.
: ~~X(t)
In using Eq. (6) as a basis for comparing CONSTANT-FORCE Y "
.shock mount Y(t) curves, we have already de-
cided to give weight to the possible Rn and f s
values by making all Rn 's equal but concentrat- VELOCIY
ing fnIs toward 100 cps. Thus, as our index I
reflecting the severity of (t) curve, we have
selected
TIME

b=ax
lY(t) + qVn(ft) , (7) Fig. 8. Constant-force mount
minimizes peak acceleration
n= 1

with
The index of severity for a constant force
fn 100,100d, 100d 2
..., 100db -l cps, F0 applied to a box of mass m is simply
and F

i, and

q =1, 5, or 10.
.- (1+ q) = a(1+q)

Thus, limits on the

index for any mount
can
(8)

The geometric progression of f, with fac- be set by the step-force result. The index can-
tor d was also selected to avoid coinciding with not be lower than a and need not be higher than
the set of peaks and valleys that occur in many I0. Since the constant-force case is so easily
spectra at uniform intervals of fn. calculated, it enables one to make quick analyses
which are accurate enough for many purposes.
In view of the many assumptions and ap-
proximations necessary to reach a numerical
index of severity, the index should be used only SOME BETTER MOUNTS
to suggest Y(t) curves that deserve further at-
tention. These curves must be tested and ex- In seeking the optimum shape of (t) curve,
amined to assure that they satisfy the thinking one is looking for an infinite number of points or
on which the index was based, nonmathematical parameters. This number could be cut down if

139
we expressed Y(t) by a Fourier series and ig- in both peak force and vibration, while the
nored frequencies above some limit, such as versed sine produced the least oscillation. The
1000 cps. This still leaves many more param- versed sine had the highest peak acceleration,
eters than the two or three which would be suit- apparently because it rose too slowly at the
able for computer optimization. So the approach beginning.
used was to study some candidate ii(t) curves
by hand calculations as a guide to the selection
of the i(t) curve, with two or three undefined COMPUTER OPTIMIZATION
parameters to be tried on the computer. Again, PROGRAM
the result will be less than optimum, but we can
use judgment to try again-unless comparison On the basis of the preliminary studies, it
with the constant-force mount suggests that appeared that some compromise between the
there would be little profit in going further. exponential and versed-sine shapes might be
best. The load-deflection curve of Fig. 10 was
The strong vibratory response of the con- selected with the three parameters, P,, P2 , and
stant force appears to be due to the step dis- P3 . The curve is made up of two segments of
continuity at the beginning, which is a spike on parabolas and a straight line. The slope is zero
a Y(t) plot. Because of the resemblance of the. at Z= 0, and from P2 up, it has no discontinui-
integral of Eq. (2) to the Fourier coefficient, it ties. Since P1 and P 2 are not necessarily less
appears that lower vibratory response would be than S, considerable variation of shape is pos-
obtained by making Y(t) a smooth shape of some sible, as shown in Fig. 11.
duration. This move, of course, increases the
peak acceleration. So the problem is to find a
shape which is the best compromise.

Three more shapes are shown in Fig. 9. P3

Each has a time parameter, T, which had to be
selected to get the best index. To do this it was
assumed that the input was a very short-duration FORCE
F
acceleration pulse and caused a net velocity
change of 10 fps and S was 1 in.

P P2
SHAPE CURVE NORMALIZED INDEX DEFLECTION Z
OF SEVERITY

CONSTANT 0f Fig. 10. Shape parameters of force-

FORCEA 1.00+1.00q deflection curves
0

LINEAR
1.16+O.19q
RAMP
0 7 t

EXPONENTIAL 1.12 O.3q

RISE Y q

o " tFORCE
F
VERSED
SINE 123+0.10q
0 '7" t--

Fig. 9. Indices of severity for some

acceleration-time shapes
DEFLECTION Z S

Fig. 11. Shapes permitted

The indices of severity were normalized by by parameters
dividing by the constant-force value of ao, which
was simply v2/2S. Thus, the linear ramp had a
16 percent higher peak acceleration than the A block diagram of the present. computer
constant force, but it produced only 19 percent program is shown in Fig. 12. Inputs for the
of the vibration. The exponential rise was better program include:

140

( I-~ "'
 INPUT DATA

Weight Factor }
Ranqe of Frequencies

Response SPECTRUM
Base Acceleration RESPONSE Acceleration &
CALCULATION [ SEVERITY
INDEX
Parameters
Allowable Deflection Shoc Mount

Deflction Revised
ShckMonShock Mount
rParameters

OtmmValues of: I_.

ParametersPRGA

Shock Mouht Force

Time Curve

Severity Spectrum
and Index

Fig. 12. Flow diagram of computer optimization

1. The acceleration-time history of the it will be shown later that the results can apply
base, i(t); to many longer duration pulse shapes.

2. The mount deflection, s; By plotting points for dimensionless pa-

rameters, the number of computer runs can be
3. The vibratory response tuning factor, q; mhinimized. The inputs to the computer are v,
S, f, c/co, and the fnrs. The fn's are defined
4. Vibratory damping factor, c/cc; and by the lowest frequency f1 and the geometric
progression factor d. Outputs of interest in-
5. Natural frequencies and weighing factor clude the optimum parameters P 1, P2 and P3,
for computing severity index, fn and d. the index of severity I(q), and the shock sever-
ity spectrum I(f,.q). Results will be given in
The program operates by computing a se- terms of the dimensionless parameters P1 /S,
verity index for an initial set of mount param- P2 /S, P3 2S/v 2 , 12S/v 2 , 4fnS/V, q, d, and c/c .
eters. It then moves in a series of steps toward
sets of parameters producing lower severity The combinations v2/2S and v/2S were
indices. When it can find no better set, it stops chosen because they are, respectively, the ac-
and the last set is the optimum. celeration and the time TO for a.constant-force
mount. Thus 12S/v 2 will be between 1 and 1 + q,
To aid in understanding the result and to as was shown earlier. The range of the fre-
allow the study of sensitivity of the result, the quency parameter 4f 1S/v is limited at present
program also prints out the severity spectrum to values somewhat above 1 so that the vibratory
for the optimum or any other mount character- response at the lowest frequency f will reach
istic of interest. a peak prior to TO . This limitation also insures
that mode deflections will be small compared to
mount deflections.
PRELIMINARY OPTIMIZATION
RESULTS Figure 13 shows the input conditions, opti-
mum parameters and indices for the velocity-
The computer program has not been in op- change shocks studied to date. The trend is
eration long enough to explore all of the possi- toward a higher severity parameter (I/a ) with
bilities of interest, but results have been ob- lower frequency parameters; this was to be ex-
tained for some interesting cases. The first pected. But the shape parameters (P2 /s and
studies were of very short duration pulses so P3/a,) reached a maximum in the middle of the
that only the resulting change in velocity, frequency range. It is easy to see that large
v X R(t), was significant. This step-velocity frequency parameters should have a near-
change is the most important shock input, and constant force as optimum. Only a small ramp

141
 (a) ACCELERATIONS

0 INDEX/o
" " " 3 1 PEAK ACCELIa
o
o

RATIOq

--

* pa/S (b) DISPLACEMENT

00
RATIO I '"

0 P/ q5
q =
q I)." I I '

0 1 2 3 4 5
FREQUENCY PARAMETER 4f, S/v

Fig. 13. Optimum indices and peak accelerations

(p2/S) would be needed to avoid vibratory dis- probability of causing a failure varies with am-
turbance of the relatively high frequency modes. plitude and frequency of the spectrum. Since
Hence, the index approaches 1 as 4fI S/v ap- data for the probability distribution will be very
proaches infinity. But as 4f 1 S/v gets smaller, hard to gather, a more promising approach is
the mount shape reaches a limit on the con- to study the sensitivity of the optimum to error
cession in peak acceleration that is worthwhile, in probabilities. This is planned for the near
For low values of 4fS/v the shape again tends future.
towards the constant force. Values for 4f 1 S/v
less than 1 are not indicated because this would
require definition of the mount characteristics SENSITIVITY OF OPTIMA
beyond T. TO ASSUMPTIONS

For the "shock resonance" factor q = 5, the Preliminary indications are that the sever-
mount shape tends to avoid vibrations at the ex- ity index is not very sensitive to the shape fac-
pense of peak acceleration; that is, P2 /s and tors or the assumption as to q, the tuning factor.
P3 /a, are higher. P1 /S is greater than 0, unlike As the computer moved toward the optimum
the q = 1 result. shape, the index did not vary much for quite
large ranges of P 2 and P3 .

SHOCK SEVERITY SPECTRA Furthermore, the optimum shape for q 5

was almost as good as the optimum for q = 1,
Plots of shock severity spectra for the six iH q were actually 1. The reverse was also true.
velocity-change cases studied are shown in Hence, errors in q had little effect on the index
Fig. 14. These curves all show a rather sim- in the range q = 1 to 5. Although the index did
ple trend with the magnitude of vibrations being not vary much, the shape did, suggesting that the
relatively small at higher frequencies. The index is not sensitive to details of shape.
index is always lower than the low frequency
end of the spectrum and higher at the other end. The difference between the indices ob-
The correctness of this depends on how the tained and the lower limit a. suggests that it

142
d

6-
4f%S/v

0 MARKS VALUE OF 4f, S/v,THE

\%- LOWEST FREQUENCY IN INDEX OF
SEVERITY CALCULATION
4 -

W
N

:__ 2-

01 Al A I
2 5 10 20 50
FREQUENCY PARAMETER 4 fn S/v

Fig. 14. Shock severity spectra of optimum mounts

would seldom be worthwhile to tailor a moumt To find a load-deflection curve of the opti-
l to suit box. mum mount, one must compute, z(t) = (x - Y) and
plot it against simultaneous valueo of ir.

OTHER INPUT PULSE SHAPES It is evident that a velocity step is the most
severe shape of input motion. More gradual ve-
The results for velocity-step shocks can locity changes, in effect, allow the mount a
be extended to a large class of acceleration "free" increase in deflection, So . This, of
pulses without using the optimization program. course, allows a reduction in severity index.
The base motion k in Fig. 3 has reached a con-
stant value, v, prior to the maximum mount
deflection, S,occurring at T. The box motion LONGDURATION INPUT
would have produced a mount deflection of PULSES
(S + S,) if the input had been a velocity step v.
It is not difficult to prove that if i is the opti-
mum box motion for the velocity step v and For acceleration pulses which have not
displacement (s + s,), it is also optimum for * reached zero by the time of peak mount deflec-
and S. tion T., the above theorem does not apply. The
theorem can be modified to say that a i will be
Suppose the optimum Y for v and (S + SO) an optimum for all input k curves having the
were used as the first trial solution to the com- same slope at T and appropriate deflection.
puter for the inputs (t) and S. The computer But to be useful, optimum solutions would be
would try alternative curves near i and find that needed for a variety of slopes.
they had higher indices-because the same al-
ternatives were tried and found higher in the For the present, it seems more practical
original optimization. Thus, if k isnot so to perform the computer optimization for each
oddly shaped as to allow two optima, y is opti- (long-duration) input pulse of interest. This has
mum for i. It is required that * be zero in the been done for a half-sine shape and an exponen-
neighborhood of T. tial decay pulse.

143

.i-4r ; P
 OPTIMIZATION FOR OTHER Finally, consideration could be given to in-
REQUIREMENTS cluding some modeling of human comfort or

.
The cases studied thus far are only a be-
ginning of those which are of practical interest.
No change would be needed to consider other -
acceleration pulse shapes; the step acceleration
injury criteria in a severity index.

ACKNOWLEDGMENTS
V
would
mount be of general
force interest.
after time T wouldDefinition
be neededoftothe
montfoer ler timel rwoulbeneded tvaluable

The given deflection condition could be re-

It
I issapesr
a pleasure tooakoldetemn
acknowledge the many
contributions of my co-worker in the
shock mount project, Vincent R. Paul. In addi-
tion to providing constructive criticism of the
f
_

placed by a more sophisticated weight-cost- theory, Mr. Paul collaborated with computer
penalty criterion related to a system's effec- programmer Ruth A. Short in obtaining a prac- j
tiveness trade-offs. tical program and the results presented here.

REFERENCES

1. J. P. Walsh and R. E. Blake, Mechanical En- 6. A. B. Burns, "Guide for the Selection and
gineering, 77:3, March 1955 Application of Shock Mounts for Shipboard
Equipment," American Machine and Foundry
2. "Hardening Technology Studies Report (U)," Co. Greenwich Conn 1
LMSC B130200, Vol. IV (SRD), 30 Sept. 1965 96 !

3. R. A. Eubanks, Shock and Vibration Bulletin, 7. L. A. Schmidt, Jr., and R. L. Fox, "Synthesis
No. 34, Part 3 (Dept. of Defense) Dec 1964,
p. 157ofaSmlShcIsltr"NSC-5,- of a Simple Shock Isolator," NASA CR-55,
June 1964
4. J. W. S. Rayleigh, The Theory of Sound, 2nd
Ed. (Dover, New, York), 1945 8. R. E. Blake and T. R. Ringstrom, Shock and
5. J. J. O'Hara and P. J. Cunniff, "Elements of Vibration Bulletin, No. 28, Part IV (Dept. of
Normal Mode Theory," NRL Report 6002, Defense), Aug. 1950, p. 15
Nov. 15, 1963

Appendix

NORMAL MODE-THEORY OF STRUCTURAL RESPONSE

Since mathematical derivations of the system or the mass system. There are no lim-
normal-mode theory are given in several texts, itations to the distribution of exciting forces
they are not repeated here. The required as- over the structure, their direction, or their
sumptions are stated, and the usual results of manner of variation with time. Excitation by
the theory are given. Various terms in the arbitrary motions, which any parts of the struc-
equation are explained and examined. It is ture must follow are also included. Of course,
shown that the expressions for stress, strain, there are practical limits to the degree of com-
deflection, .andacceleration are all of the form plexity of a situation which can be handled, but
given by Eq. (2). they are not due to a limitation of theory.

The theory concerns the response of any The linearity of the system makes it possi-
shape of linear elastic structure to any disturb- ble to use the principle of superposition. The
ance whose variation with time is given. The response of the structure resulting from appli-
structure has distributed mass, damping, and cation of a force at one point at one instant of
elastic (spring) elements. The system of elas- time is independent of response to other forces;
tic elements could be described by stating the thus, one need only add (superpose) the effects
stiffness of all the springs or by a set of influ- of individual forces to obtain the total response.
ence coefficients, but deflections must be small If r .(t - r) is the response at time t at point i
enough that this description is sufficient. Damp- on the structure due to a unit force at point j
ing is viscous and must be distributed in the which acted for the time interval T to T + dr,
same manner in a fixed ratio to the spring then the response to many forces Fk(T) is

144

* ~
'-..9---~-S-_
 _____ ____ n] "Z'

ri(t) =E Fk(-r ) rik(t -,")dT (A-1) ritkt) f F .0~e

The response usually derived is the deflec-

tion, the distance that the point on the structure x sin 2nfn( t - -r)dT. (A-3)
has moved from the point itwould have occupied
ifno forces had been applied. A complete de- The corresponding equation for disturbing mo-
scription of the deflection of all mass points on o
report.
the structure requires the statement oi as many
numbers as the system has degrees of freedom
v. All real structures have distributed mass
and thus v = w. For analysis, structures are Damage is not associated directly with de-
often represented by point masses, each able flection, but it may be due to excessive strain
to move inthree dimensions. The number of or relative motion between two parts, that is,
degrees of freedom is then three times the with [ri(t) - r.(t)]. When Eq. (A-1) is used to
number of masses, and there will be v values find this expression, the result is similar to Eq.
of i inEq. (A-I). (A-3) except that 4i , is replaced by [ i- in]
and, of course, ri(t) isreplacedby [ri(t) - rj(t)].
Ithas been found that a structure will vi- Since stress and internal -forces are proportional
brate after all disturbing forces are removed to strain and relative deflection, a similar re-
incertain special patterns called normal modes sult is obtained for them.
and at certain frequencies called natural fre-
quencies. Any arbitrary deflection of the struc- Forces can also be calculated from the ac-
ture can be represented as the sum of normal celerations of masses. The expression for
modes of suitable amplitude-in close analogy acceleration is the same as Eq. (A-3) except
to= the representation of any arbitrary function that 2nf. is in the numerator.
by a Fourier series. The advantage is the same;
a satisfactory approximation can be obtained in
fewer terms than the number of degrees of free- The integral of Eq. (A-3) can be expressed
dom. Thus, we have said that if 9in defines the in two parts:
shape of the nth normal mode, the most general
jt motionpossible in the absence of disturbances is
-5
-t A e / JFk(r)e(t-r)
k sin 2nfn(t- T)dr
-/c c 21rfnt
ri(t) = E An 0j.e- sin (2nf nt + 0 ) .

- Fk(t) 1 f(dFk.\ -C/Cc 27fn(t-r)

The An's may have any value; in fact, all (2iif.) 2

(27fn) 2
J & e
An's may be zero except one and the structure
will vibrate at a single frequency. Methods of
computing mode shapes 5in and their corre- x os 27 fn(-t )d-. (A-4)
sponding natural frequencies f, are given in
the references. Letting

The solution for r. (t -), the so-called

indicial admittance, following an impulsive dis- (2fn)2M- ekiJ R(A-)
turbance, turns out to be (for small damping)

rjj(t-T
- Jntin e C/cc 277fn(t -r) and
2.fnf~ MEOinn(dFk e-C 2wfn(t-')
er
e cos 2nfn(t -r)dr
x sin 2nfn(t- ), (A-2)
V
Vn(fnt), (A-6)
where Me is the mass associated with the de-
flectiori hn; therefore, the form used in Eq. (2) is obtained.

145
I

DISCUSSION

Mr. Snyder (General Dynamics/Electric curve of a single pad, cushion or resilient "
Boat: Are you planning to compare your ana- element ?
lytical data eventually with experimental data,
or have you done this in your efforts to deter- Mr. Blake: Yes, in this case the problem
mine the optimum design for a mount to put was in one dimension. There is one shock mount
under some of your chassis, boxes, and so on? and all parts move in one dimension.

Mr. Blake: Not in the immediate future. Mr. Johnson: Does your method apply to
We would like to do that eventually when we mounts presently being developed with some
have some ideas of what hardware would be medium for a low-frequency vibration mount, a
best to use as a mount. - then leading into a stiffer system for shock
cushioning? Will your method apply to this
Mr. Snyder: Will these data eventually be type of mount?
available ? Mr. Blake: You mean could we put in se-

Mr. Blake: Yes, if we get the information ries, for example, a soft element then a stiff
I am sure it will be available. If it is done for element?
the Air Force it will be in our reports to them, Mr. Johnson: Yes.
of course.
Mr. Blake: At this point, we do not care
Mr. Johnson (K. W. Johnson & Co.): Are how one achieves the force time history that we
your data based on a single force deflection find to be optimum.

* * *

146
 SHOCK ATTENUATION USING PASSIVE ELEMENTS
V. H. Neubert and D. L. Pyke
Pennsylvania State University
University Park, Pennsylvania
and
D. F. Poeth
HRB-Singer Corporation
State College, Pennsylvania

In ships or other vehicles, it may be desirable to add mass or passive

dynamic absorbers to protect mounted equipment from shock inputs.
Also, the equipment itself will act as a dynamic absorber with respect
to motion of. its base, as is known from the shock spectral dip phenom-
ena. Studies were carried out to investigate the magnitude of the effect
of using masses, compared with masses on springs, as passive ab-
sorbers. Also, a nonlinear absorber was studied where the mass was
supported by a bolt with relatively low preloads so that separation of
the mass from the base could easily occur in one direction only.

Response of a thin free-free bar of circular cross-section is compared

with that of the same bar with masses or mass and spring combinations
attached at the end opposite the struck end. Masses alone may be quite
effective in reducing amplitude of motion at the attachment point, but
amplitude at other points may be increased. Using the same mass on a
spring as a passive dynamic absorber may be more or less effective
than using the mass by itself, depending on the spectrum of the input
pulse and the effect of the absorber on the system mobility.

For the second system, a disc was attached to a short, chick steel cyl-
inder using a single preloaded bolt, simulating in ship's the use of hold-
down bolts to attach equipment bedplates to foundations. By measuring
bolt strain and base acceleration a fairly clear physical picture of the
motion occurring was obtained. While the disc is in contact with the
base cylinder it has the ability to act as a high-frequency dynamic ab-
sorber. During separation or partial separation, it suddenly becomes
a low-frequency system consisting of a mass supported primarily by a
bolt. The usual shock spectrum of base motion indicated some absorber-
type action at low frequencies. A special shock spectrum, calculated
from base accelerations during intervals of separation compared with
a spectrum from accelerations during contact, indicated to some ex-
tent the directional variation of the base motion.

INTRODUCTION only may bo more desirable than adding both a

mass and spring to form a dynamic absorber.
The attenuating effect of passive dynamic
absorbers in systems experiencing steady In this paper, examples are presented to
sinusoidal vibrations has been thoroughly dis- demonstrate the effect of adding masses or
cussed in the literature. Den Hartog (1,2) dynamic absorbers to attenuate transient
presented the classic case of an absorber at- vibrations.
tached to a base mass and spring system.
Neubert (3) considered a uniform bar having Two different systems were used in the
one or two absorbers tuned to attenuate steady experiments. One was essentially a one-
vibrations. Optimum tuning and optimum dimensional bar, 30 in. long by 3/8 in. in diam-
damping on displacement and velocity or eter, to which either masses or mass and
mobility.curves was emphasized. It was also spring combinations were attached as shown in
demonstrated that in some cases, use of mass Fig. 1. The other system is called a cylinder

147

,$
and was 13 7/16 in. long by 6 in. in diameter. For the thin bar, u is the particle displace-
A mss was attached to the cylinder by a single ment and x is the position along the bar meas-
preloaded bolt as shown in the bottom of Fig. 1. ured from the struck end. If E is Young's
The transient force was applied by using strikers modulus, p the mass density and a 2 = E/p, the
as indicated by the dotted lines. The effect of wave equation in one dimension is
the addition of passive elements on maximum
2
displacements at the point of attachment of the a U
elements is discussed. a2 t12

To investigate mobility, the form of the solution

h--4"1 __________for steady sinusoidal vibrations is

STRIKER BAR- U(x,t) ; B cos!I+ C sin e (2)

Here B and C are constants determined from

1J boundary conditions. Ifthe area of the free-
STRIKER BARWITH MSS free bar is A and the force feiwt is applied at
x =0, the boundary conditions are

~3u
AE-(,t) = feJ~
STRIKER a MASS
BAR WITHSPRING and (3)

13~*-
71 ," 3-"--
16 12.8 AE-y-(t~t) =0.
T IPRELOADED
6- BOLT Evaluating the constants B and c in Eq. (2)
-based on Eq. (3) and differentiating with respect

r
STRIKER CYLINDER ENDMASS to time, we obtain the particle velocity
Fig. 1. Systems studied

Aa sin
i.(xt) - Apa Cd', (4)
Primarily, experimental results are pre-
sented for the cylinder, which is much more
complex than the bar. For the bar, both ex- Mobility Mi j is defined as the velocity at i
perimental and analytical results were obtained due to a force at j. If position I is defined to
and agreement between the predicted and meas- be at x = 0 and position 2 at x t, then
ured values was excellent (4).
, u(0,t)
fe ji t
THEORETICAL BACKGROUND FOR and (5)
MOBILITY OF THIN BARS

fejw&t
To introduce notation, some of the the-
oretical relationships for the thin bar is first
reviewed. To predict the transient response of Primes are added here to distinguish from
the thin bar with end attachments, the one- mobilities of combined systems.
dimensional wave equation may be applied. Then from Eq. (4), using o/a
Two types of solutions are discussed: one
based on the use of normal modes of free vi-
bration, and the other based on the use of ap- , Cos p (6)
propriate mobilities and the Fourier transform M Apa sin p
relationship between mobility and transient re- and
sponse. Each method has some advantage in
helping to visualize the physical behavior and , - 1
both approaches are outlined briefly. M21 - pa sinp (7)

148
 The M 1 and M' are shown by dotted lines The mobility M 3 is the mobility of the dynamic
in Figs. 2 and 3. absorber driven at the base of the spring or, if
k is the stiffness of the absorber spring and m
If the bar has a dynamic absorber attached the absorber mass,
at position 2, the mobilities of the combination
are jk
7 M33
M;3
=
0'jk . ( o
(10)
(8)
MM21 -- jm
+ M3 3 or
I nP
~p2
and 2 ~(1
' N M 33 Apa 'CV
M22 M21
M,2 1 '21 - (9) Here /3 = m/Ap is the ratio of the absorber
M22 + M33
mass to total bar mass and p, = %c/a is a

13-Q50

10
+ l1
'/i/
\\ /
\ I
\ \ I
P
-T I \1I
to

I I
-- 1 0--0 /5\0
001 .. . ...
0 2 4 5 6 7 8 9

i#\
t- II \ /I'l
I \'
/ \ \Il
/ \/
/ \
P

(b)

0 I 2 3 4 5 6 7 8 9

" (b)

Fig. 2. Point and transfer mobility for free-free

bar (3 = 0) and bar with concentrated mass at x e
(=0.50)

149

. ---. 4 . - --
 _0__~050
ItPo=3Oo Ii
tO Il

~~ FREE..
BARI

I0 / \ /

_' II\ I I\

01- oI 1 IIi
001 .....
I I/
I, II ,
0 I 2 3 4 5 6 7 8 9
P

(a)

10o
"05
[t,-0
101
*

IIl
1FREE
I

300
0,
I'
I
I

S\ BAR I
';/ \ / /

0.I -

001 0 1 2 3 4 5 6 7 8 9
*
(b)

Fig. 3. Point and transfer mobility for free-

free bar (,6 = 0) and bar with dynamic absorber
at x = (f8= 0.50)

nondimensional natural frequency. From Eqs. If the absorber spring is very stiff, pn, co and
(6) and (7) it is obvious that resonances occur the resulting frequency equation is, for the bar
at P = Pn = nr or on = niia/C for the free-free with a rigid mass attached at x t2
bar.
tan p = -6p. (13)
For the combination indicated by Eq. (8) or
(9), the resonances will occur at m 2 + M3 3 = 0. If the end mass is zero, 63= 0 and the fre-
Since M' 2 = M, the frequency Eq. (12) is ob- quency equation for the free-free bar is
tained by adding Eqs. (6) and (11) and cancelling
some constants: sin p = O0 (14)
3
'P
tanp p 2
- (12) If the mobility is known, the transient re-
sponse may be obtained from

150

'
. . . . .. . . -----.- C
 the free-free bar. The magnitude of transfer
21 fJ
J t
vi(t) e Mj(&) Fj(oj)dw. (15) mobility is significantly decreased.

In Fig. 3 typical mobility curves are shown

The numerical evaluation of Eq. (15) is possi- for a dynamic absorber attached at x . Here -

ble. Here F(w) is the Fourier transform of the 6 = 0.50 and Pa = 3.00 is the nondimensional
actual forcing function. natural frequency of the absorber. The point
mobility curve shows the usual attenuation range
The pressure pulse applied to the bar may near p = 3.00, with an additional resonance ap-
be deduced from the strain record at x = /2 in pearing. Again the transfer mobility shows a
Fig. 4. The first negative pulse is assumed to decrease in amplitude.
represent the shape of the input pressure sup-
plied by the impact of the striker bar. This Although Eq. (12) indicates that knowledge
pulse was used in a numerical analysis as of M1 .(&) and Fj(c) will aid in anticipating
transient input to the bar at x = 0 with the dotted transient response, it is also helpful to inves-
line in Fig. 4 showing the resulting predicted tigate the normal mode solution.
strain at x /2. The case shown in Fig. 4 is
for 63= 0.62 and p, = (k/m) 1 / 2 'i/a = 1.575 where
pa is the nondimensional natural frequency of NORMAL MODE SOLUTION
the absorber.
The normal mode solution for a bar free at
x = 0 and having attachments at x = t is of the
Theoretfal form
30--- Experimental
paoz1575
/ P~c -
U( t
()
Cos
~
q(t)
20 =

10 I' k- pnx
=20' cos - qn(t). (16)
40r59 =
0 19 C! 0
time The p,, are the natural frequencies determined
J -10 for various cases from Eqs. (12), (13) or (14).
The mode shape is cos (pnx/t) for each case.
-20 If the kinetic and potential energies are
written for the various cases in terms of the q
-30 coordinates and Lagrange's equations are used,
then the decoupled equations will be
-40
+ =
Mn 4n Mn (on2 q n Qn n > 0
Fig. 4. Strain vs time at x C/2 (17)
for bar with dynamic absorber
=
at x =? (/3= 0.62, p. = 1.575) M 0 qo Qo n = 0.

For all cases, for a force F(t) acting at x =0,

=
The input force is not exactly a half-sine Qn F(t)
"I'. wave. If it were a half-sine wave of frequency
W
.,the Fourier transform F.(&J) of the input For the general case (4) of a bar of mass
would have a broad spectrum of input primarily mb Apt, with an absorber at x
from w = 0 to e,= 3 wf (5). For a half-sine wave
of 150 Isec duration, the first frequency cutoff mb m k(k+ %' m)
would correspond to p =9 for the bar studied. M= - + 2 - 2M) 2 cos 2 n n >0. (18)
(kcom) a >.(8
In Fig. 2 the solid curves represent mo-
bil.ty of a bar with a mass rigidly attached at
x =-tand /3 = 0.50. As expected, the resonant If k.is very large, Mn for a bar with an end
or natural frequencies are less than those for mass only is

151
 mode analysis. For the greatest possible accu-
Mn 2 + cos 2 n n >0 (19) racy, the input force-time curve was approxi-
mated by a series of step functions, and the re-
Also, if m= 0, the value for a free-free bar is sponse obtained as the total response to these
step-function inputs. In this summation process
mb much of the physical understanding of the prob-

I
Mn n >0 (20) lem is lost. On the other hand, the input may be
approximated to some extent by a half-sine wave.
For n =0, for all cases, M = mb + m andPn = O. Thus, Eqs. (23) and (24) at least show qualita-
tively the parameters involved. Obviously a
A general form of Eq. (17) is modal effective mass term may be identified as

2
M~o~ F(t) n >0. (21) / Pnb)Ph

If F(t) were a steady sinusoidal force

F(t) = fei - t the response is obtained by sub- where I, is the position of the force and x the
stituting the solution of Eq. (21) into Eq. (16) or position at which displacement is measured. In
this case h = 0 so the effective mass is
u(x, t) : Cos PX fe2 Pn x
:I L Mn( - 2) M /cos .

t
fe Jw The impedance curves and Eqs. (19) and (20)
(mb + m) W2 (22) show that Mn decreases when the concentrated
mass is added to the tip. The other important
If F(t) is a half-sine wave, parameter is (f/C%. For a pulse of 150 ,sec
duration, pf = wft/a is about 3.14. A decrease
7( in wn particularly for the first mode results in
F(t) ff sin cf(t) 0 < t < an increase in modal amplitude, even though Mn
increases slightly.
F(t) 0 t > -

TRANSIENT RESPONSE OF BAR

In this case, the solution for elastic displace- WITH END ATTACHMENTS
ments is
From the preceding discussion it is obvi-
OD fpx ous that the transient response may be pre-
ff P dicted using calculated or measured mechani-
u~x, t) -- M(.-2) cal mobility. This approach has been discussed
in detail by Belsheim and Young (6). In the
present example of a simple bar, a traveling
-<" f < (3 wave solution could also be obtained. A normal
Wn -(23 mode solution was actually used with ten modes
summed to predict elastic displacement to
and within- 1 percent and 100 modes summed to pre-
dict strain.
f_
ff / Pnc In Fig. 4, the calculated strain at x = V2
u(x, t) :n Mn(c -c2 ) o is compared with the measured strain. The
applied pressure pulse applied at x = 0 was as-
sumed to be the same shape as the first nega-
x - f sin 71 sin con t 7tive strain pulse which appears on Fig. 4 be-
(n -W 'f/ginning at 80 /,sec and ending at 230 tisec. The
magnitude of the applied pressure pulse was
7""n.)- taken as Young's modulus E times the strain
1+'0
-7-os n (5f
,- tL> -. (24) magnitude of the first pulse. The compari-
son of measured and calculated strain in Fig.
4 is typical of many presented in Ref. (4),
The actual prediction of response for the confirming the validity of the analytical ap-
cases presented was carried out using a normal proach used.

152
 For this study, the maximum elastic dis- x = -C,
the value of Pa = 1.575 is of particular
placements were calculated along the bar for interest because the first mode node is at x =.
various cases using the same input as in Fig. Other cases would certainly be of interest in
,4. Some results for maximum elastic displace- addition to those examples considered here.
,ient are shown in Fig. 5 for a mass rigidly
attached to the bar at x = 30 in. The dotted
A

line inFig. 5 shows the maximum elastic dis-

placements for a free-free bar where ,3 0. _
These displacements generally do not occur at
the same instant of time. Other values of 3 4 -
were considered as shown. Addition of concen- 0 . .
trated mass causes considerable reduction in - 0
displacements at the attachment point. How-
ever, the amplitude at the drive point is in- _u "0.62

creased. This may be anticipated to some 2

extent by observing that W

Pnx -
u(x.t) COS q3

n= 0 5 I0 I5 20 25 30

for the elastic modes. The modal amplitudes

at x = 0 and x = will only be equal in magni- Fig. 6. Maximum elastic displace-
tude when Pn - '7, or when the.bar is free- ment for bar with dynamic ab-
K' free. Addition of mass causes the values of pn sorber at x = f and various values
to decrease, which causes nodal points to shift of /3, with p = 1.575
to the right on the bar.

5515

_ - Po.O

4-

Pa '6.0
. .2 0

i - . ' iV

%
Ck.
I. %

060

0.60

0 5 10 15 20 25 30
* , X Onctei

X(inchWs Fig. 7. Maximum elastic displace-

ments for bar with dynamic ab-
Fig. 5. Maximum elastic displace- sorber at x = f and various values
ment for bar with concentrated of P., with 8 = 1.53
mass at x= andvarious values of 6

CYLINDER WITH NONLINEAR

Results for a bar with a dynamic absorber ATTACHMENT
' iat x =,- with frequency p. = 1.575 are shown in
Fig. 6. The same values of / were considered. The cylinder having the end mass attached
Comparison of Fig. 6 with Fig. 5 indicates that by a preloaded bolt is shown in Fig. 1. The
addition of the absorber may or may not be gage length of the bolt was 1 1/8 in. on a sec-
more beneficial than using the mass by itself. tion machined to a diameter equal to the root
This is confirmed in Fig. 7 where, for /8 = 1.53, thread diameter of 0.185 in. The area of the
maximum displacements are plotted for several bolt section was 0.0269 sq in. In the photograph
values of p,. With respect to attenuation at (Fig. 8) strain in the bolt is recorded as the

153

77-1 rr
lower trace in each pair of traces. The upper Preload strains in the bolt of 600, 1000 and
trace is the record of acceleration near a point 1400 uin./in. were used for the cases of Fig. 8.
on the axis of the cylinder at which the bolt was If E = 10.1 x 106 psi, the preload stress for a
attached. The accelerometer was mounted in a strain of 1400 min./in. on the bolt was about
small recessed hole. Details of instrumentation 14,000 psi or a total force of about 38 lb. The
are given in Ref. 7 or 8. percent of contact of the end mass and the cyl-
inder is not known, but if the contact area were
10 sq in., the pressure would have been only
600. in Pieload * about 4 psi. Thus, the contact area could sup-
Accelesation in port practically no superimposed tensile strain
MY and it might be assumed that separation of the
C" end mass from the cylinder would occur at ten-
sile strains greater than about 0.1 ,un./in. In
the cylinder the first strain pulse is compres-
Strain sive, but it is then reflected as a tensile pulse.
20%.in 200 -. Thus, it might be assumed that there is separa-
i--" 00 _.. tion of the end mass from the cylinder each
cm time the strain becomes negative. It is ob-
I000.,in Preload * served that the negative swings are nearly half-
,n- sine waves. As the preload is increased, the
length of these half-sine waves decreases. This
indicates that the separation may not be com-
plete and the stiffness of the contact region is
variable, rather than simply being equal to the
bolt stiffness during separation.

For the present study, it is of particular

interest to investigate the effect of motion of the
1400.in Preload * end mass on the base motion. Base velocity
in (Fig. 9) was obtained by integrating the accel-
eration. The effect of preload, as shown in Fig.
9, is that higher preloads tend to result in de-
creased base velocity during the early time in-
terval before the first apparent separation.
In Fig. 10, the shock spectra associated
Fig. 8. Oscilloscope rec- with base motion for two different preloads are
ords of base acceleration shown. The higher amplitudes of Fig. 9 are
and bolt strain vs time for associated with lower frequency components.
various initial bolt pre- The change in magnitude at about 800 cps indi-
loads cates that some absorbed action is taking place.

4-,

.........
End
- -- $00 Pro-looad
pin#,a.

S...... 1000 A Wmin Pro-lood

-- 1400 )ll
in/MkProl-load

50 T00 150
sec)
Time (10"

Fig. 9. Base velocity vs time

for various bolt preloads

154

- ~-'I,'-,
-
/ . .
 7 "residual" spectrum was obtained by assuming
for calculations that the base acceleration was
6- zero at the end of the oscilloscope records of
600,Lin/in. Fig. 8.
C!1

V>4,___________
4 -- Finally, an effort was made to determine a
-3 ,spectrum sensitive to the highly directive nature
g I4oo,~ifl~ of the system. This was done by calculating one
residual spectrum using as inputs the base ac-
celerations during intervals of contact. A sec-
0 2 3 t ond residual spectrum was calculated using the
0 1000 2000 3000 4000 5000 6000 base accelerations occurring during intervals
- cps of separation. This has some mathematical
justification for linear systems based on the
Fig. 10. Velocity spectrum for Duhamel integral. The resulting spectra are
two different bolt preloads shown in Fig. 13. It is interesting to note that
during contact the low frequencies dominate and
during separation the higher frequency compo-
nents contribute most to the spectrum.
maximum
For the 600 gin./in. preload, the
negative shock spectral velocities are compared
with maximum positive velocities in Fig. 11.
They are practically the same except at frequen-
cies near 3000 cps. 7

In Fig. 12, the initial shock spectrum is 6 6001L i inpro-load

compared with the residual spectrum. The C
- 4 contac

7 - 3

6 - 2

.5 600pin/minPRE-LOA0 I ,,%/_,.i , / / ..
%- ,
0 1000 000 3000 4000 5000 6000
5 3 - .cps
0-J
Fig. 13. Velocity spectrum calcu-
c lated from base accelerations meas-
ured during contact and from base
S 0 20 30 40 5 6 accelerations measured during sep-
00 2000 3000 4000 5000 6000aration

Fig. 11. Spectra of maximum nega-

tive and maximum positive velocities
The accelerometer outputs were filtered
using a low-pass filter with a cutoff of 8500
i !cps. Initially a 2000-cps filter was tried, but it

600
__ .in/in pro- Wad / was found that this eliminated much of the es-
6 6W ;L npro-Ioad sential information regarding base acceleration.
L -5
In Ref. 8, emphasis is placed on relating bolt
strain to acceleration measured at the base and
N4 on the end mass.
I I
2 \ ittol "2"2/ t SUMMARY AND CONCLUSIONS
Iresidual_I- d // VI The results show that, for an input pulse
0 . V having a broad frequency spectrum, concen-
0 1000 2000 3M 4000 500 6trated mass may be added to attenuate vibra-
tions at a certain point. The effect of mass has

Fig. 12. Initial velocity spectrum been known and accounted for in many shock
and residual spectrum spectra. The fact that using the same mass on

155
 -
'L- . -- . - -- ' 1 -. l - -

during separation as a separate input. The

a dynamic absorber may be more or less useful separation appears to have been only partial
than the mass by itself is of particular interest, separation probably because of elastic defor-
Thus, a machine having a certain dominant fre- mation of the surface of the end mass which
quency with respect to shock response may be contacted the base cylinder.
protected by having a system nearby having a
similar dominant frequency. This suggests also No signals due to sharp impacts were noted
that consideration might be given to having in accelerometer or strain records. These
items with simi" .r fixed base resonances might have been expected if the mass banged
mounted in th: same area in a vehicle, against the base. This may be another sign that
separation did not occur instantaneously, but
No consideration has been given here to rather a more continuous reduction in the area
increase in stress due to hard spots created by of contact occurred.
adding structure locally. Of course, additional
mass or dynamic absorbers could be distributed The problem arises as to how acceleration
over an area rather than placing one large mass data may be processed to detect that preloads
or absorber at a particular spot. have been exceeded during shock motions. If
sharp separation cccurs, the machine on the
Knowledge of the effect of the addition of bolts as springs will represent a low-frequency
passive elements on the point or transfer mo- system compared to the situation where the
bility is sufficient to predict resulting transient machine bedplate is in contact with the founda-
response. Visual inspection of the change in tion. On a usual shock spectrum, a dip caused
the mobility curve is generally not sufficient to by the motion during the outswing may be can-
decide on an optimum frequency for an absorber, celled by motion which occurs during contact.
because F(co), the Fourier transform of the tran- Due to the simplicity of the system used in the
sient input, and ei&t are also involved. Thus, present studies, special shock spectra were
each case must be calculated. However, even calculated from base motion which emphasize
for complex elastic systems, calculation of the directivity of the base motion. As a general
transient response is readily performed using procedure, accelerations measured on the ma-
a digital computer. chine itself may be useful. Information regard-
ing strains in bolts is probably the most essen-
The cylinder carrying the mass attached tial in determining that the desired preload
by a preloaded bolt simulates equipment bolted exists and in understanding response during
to a subbase. In the laboratory apparatus used, shock motions.
measurement of the dynamic strain in the bolt
furnished the most useful information regarding
behavior of the system during shock loading. ACKNOWLEDGMENT
Traces presented clearly indicate that the strain
during separation has primarily a low-frequency
content, and higher frequencies dominate during The results regarding the preloaded con-
periods when the mass is in contact with the nection were obtained as part of a study
base. This was difficult to detect in the usual supported by the Office of Naval Research,
shock spectra calculated using the base acceler- Contract No. Nonr-656(28)(x). Mr. Pyke's
ation as input. It did appear in shock spectra work was carried out under a National Sci-
(Fig. 13) calculated by using base acceleration ence Foundation Traineeship.
during contact as one input and base acceleration

REFERENCES

1. J. P. Den Hartog and J. Ormondroyd, Trans. 5. S. Rubin, "Concepts in Shock Data Analysis,"
ASME 50, 9 (1928) in C. E. Crede Shock and Vibration Handbook,
Vol. 2 (C. M. Harris and C. E. Crede, Ed.)
2. J. P. Den Hartog, Mechanical Vibrations (McGraw-Hill, New York), 1961
(McGraw-Hill, New York), 1954
6. R. 0. Belsheim and J. W. Young, Jr., "Me-
3. V. H. Neubert, J. Acoust. Soc. Am. 36, 673 chanical Impedance as a Tool for Shock or
(1964) Vibration Analysis," U.S. Naval Research

4. D. L. Pyke, "Investigation of One-Dimensional Laboratory, NRL Rept 5409, Feb. 15, 1960
Elastic Waves with Compound Reflection
Conditions," Masters Thesis, Pennsylvania 7. D. F. Poeth, "Response to Shock Loading of
State University, Dept. of Engineering Me- a Mass Supported by a Pre-Loaded Spring,"
chanics, Sept.'1965 Masters Thesis, Pennsylvania State

156
 University, Dept. of Engineering Mechanics, of Naval Research Contract No. Nonr-656
March 1965 (28)(x), Interim Rept.No. 3, Feb. 1965

8. V. H. Neubert and D. F. Poeth, "Shock Re-

sponse of a Pre-Loaded Connection," Office

157
 INVESTIGATION OF THE VIBRATION
DAMPING PROPERTIES OF VISCOELASTIC
MATERIALS USING A DELAY ANGLE TECHNIQUE
Saul A. Eller and Levi Cohen
U. S. Naval Applied Science Laboratory
Brooklyn, New York 11Z51

The vibration damping properties of 1/4-, I/Z-, and Z-in. thicknesses

of ML-DZ and a urethane viscoelastic damping material were studied
in the frequency range of 150 to 11,000 cps by disc and beam methods,
which measured the rate of vibration decay in the steel substrates onto
which the materials were adhered. Results were expressed in terms
of percent of critical damping of the material. In addition, the materials
were inve-tigated by a novel delay angle technique which measured ac-
celeration ratio, velocity of sound, and delay angle over a frequency
range of 50 to 10,000 cps. High vibration damping occurred at those
frequencies where there were resonances in the thickness-wise direc-
tion, as indicated by high acceleration ratio values, confirming the work
done by another investigator in this field. The ML-DZ material had
better vibration damping properties than the urethane material at fre-
quencies above 1,300 cps, whereas the reverse was true at lower fre-
quencies. Analysis of the data obtained by the delay angle technique in-
dicated that a frequency is reached at which only those portions of the
material which are in contact with, or adjacent to, the substrate are
vibrating,,and thus absorbing energy. This is one explanation why vi-
bration dalmping at all frequencies is not proportional to the thickness
of the material.

INTRODUCTION specific damping energy several hundred times

that of typical structural materials.
The Navy has been conducting extensive
research programs to reduce ship noise and to Although satisfactory v~iscoelastic materials
minimize its effects.by the following methods: have been developed, there is comparatively little
(a) elimination of the causes of vibration in fundamental information regarding the manner
ship's components, (b) isolation of sensitive in which viscoelastic materials absorb vibra-
equipment (e.g., sonar) from noise sources, and tional energy and the frequency range in which
(c) absorption of vibrational energy by adhering they exhibit optimum damping properties.
energy-absorbing materials to bulkheads of - Some investigators attribute vibration damping
compartments, particularly those in the vicinity to tension and compression forces parallel to
of sonar equipment. the surface of the material, whereas other in-
vestigators attribute damping to forces in the
This paper concerns an investigation into thickness-wise direction.
the physical properties of viscoelastic materials
used to absorb vibrational energy. The use of The effectiveness of damping materials can
these materials has been so successful that be indicated by percent c/c 0 (percent of critical
more than 3 million square feet of ML-D2, a damping), the percentage of damping required
viscoelastic vibration damping material, have for a nonoscillatory return to original conditions.
been installed to date by the Navy in the Fleet This may be determined by either the disc or
Rehabilitation and Modernization Program (1). beam methods, described below.

The use of viscoelastic materials for damp- Both of these methods have the disadvantage
ing vibrations is understandable because, as of requiring a new setup for each discrete fre-
noted by Ruzicka (2), these materials have a quency at which the damping of the material is

159
determined. In addition, the quantity of mate- with an accelerometer. Damping is calculated
rial and the time required are relatively high. in terms of the percent of critical damping of
the system. When a mechanical system of
The delay angle technique to be described mass m, damping c, and flexibility k is sub-
rates damping materials in terms of "acceler- jected to an impact force, the response y or a
ation ratio" (gain in acceleration in traversing point on the mass at time t after the impacts
through the material). At high acceleration is given by
ratios, the material is absorbing more energy
and, therefore, is a better damping material. i 2y dy
Only one small specimen of each thickness is re- - - + c dt + ky = 0 (1)
of interest.
quired for the entire frequency range
The solution of this equation is given in the
VISCOELASTIC MATERIALS appendix for reference, and leads to

The viscoelastic materials studied were C Dm (2)

ML-D2 and NASL-D101 urethane material which
were developed and formulated by T. Ramos of where D is the decay rate in decibels per sec-
the U. S. Naval Applied Science Laboratory. ond, and to
These materials can be cured at room temper-
ature; their compositions are given in Table 1.
Samples in thiclmesses of 1/4, 1/2 and 1 in. c = 4-rmf o (3)
were studied with the 'disc method and the delay
angle technique, whereas only a 2-in. thickness where
of material was studied with the beam method. c= the minimum damping for nonoscillatory
return (critical damping), and
VIBRATION DECAY MEASUREMENTS f, = (1/277) -,/i is the natural frequency.
USING DISC AND BEAM METHODS Combining Eqs. (2) and (3), we get
The disc method, described by Kallas and c 1 0 0 1 83 D
Rufolo (3), makes use of a coated steel disc z-x =:f--f
The disc is given a
suspended at its center.
sharp impact and its vibration is measured or

TABLE 1
Composition of ML-D2 and NASL-D101 Urethane Vibration Damping Materials
1 Parts by 1 Order of
Component Function Weight Mixing

(a) ML-D2 (sp gr = 0.73)

Halowax 4004 Flame regardant 42 1

Epon 828 Co-reactant 20 2
Versamid 115 Co-reactant 100 3
DMP-30 Hardener 5 4
Antimony- trioxide Flame retardant 14 5
Aluminum oxide
(80 mesh, loose grains) Filler 400 6

(b) NASL-D101 (sp gr = 1.54)

Chlorinated paraffin Plasticizer and 55 1

flame retardant
Polyol Chain extender 25 2
Urethane prepolymer Rpactant 100 3
Ottawa River sand Filler 80 4
Organotin Catalyst 30 5

160

>
e- --
 C 1. D The beam method was similar to the disc
ff 0 (4) method, except that a 45 1/2-in. by 3-in. piece
of viscoelastic material was adhered to a steel
The method takes into consideration the inher- beam, 47 in. long by 3 in. wide by 1 3/4 in.
ent damping in the steel disc onto which the thick. The beam was suspended at its midpoint
viscoelastic material is adhered by subtracting and driven by a small shaker to vibrate at a
the vibration decay due to the steel disc from number of resonant frequencies. A view of the
that of the composite system. Therefore, the setup for the beam method is shown in Fig. 2.
When resonance was attained, the shaker was
of the material) 183 (D.-shut off and the time rate of decay of the vibra-
84 (DODB) (5) tion of the beam was determined using electronic
equipment and calculations similar to those used
where for the disc method.

Do = damping rate of coated disc, db/sec;

APPARATUS USED FOR DELAY
DB = damping rate of bare disc, db/sec; and ANGLE TECHNIQUE

fo = natural frequency, cps. The delay angle technique was designed to

determine the effect of thickness-wise (longi-
A more detailed development of the disc tudinal) tension and compression forces on the
method is given by Timoshenko (4). Another damping properties of viscoelastic materials.
analytical study of the vibrating disc which dis-
cusses the geometrical relationship to frequency Figure 3 shows a specimen mounted on a
response has been presented by House and Kritz specimen plate, with holder and accelerometers
(5). attached. Figure 4 is a view of the specimen
mounted on an electrodynamic shaker, and Fig.
Tests on each thickness of material were 5 is a schematic of the apparatus. As shown
made on a series of 3/8-in. thick discs which
measured 8, 6, 4, 3 1/2 and 3 in. in diameter
and had bare disc natural frequencies of 1750, ,.
3500, 7000. 9000, and 11,000 cps, respectively.
The specimens covered only one face of the disc, -
as shown in the setup for this method (Fig. 1).

FOLLOWERt .

SPECIMEN
I4

'M .

,f C ELEROMETER ,.

Fig. 1. Setup for disc method Fig. Z. Beam and shaker

161
 ACCELEROMETER
AC E..... , "-'-PREAMPLIFIER
SPEC'MEN'-'' EPOXY ADHESIVE

SPMENE CATHODE FREQUENCY

SPCMN-- RING ICONSOLE I
Ft- .,l E SHAKER TABLE

PHASE ANGLE TRUE-RMS DUAL-BEAM

METER VOLTMETER OSCILLOSCOPE

AA 1

Fig. 3. Specimen assembled for

delay angle technique

Fig. 5. Schematic for delay

angle technique
....... ........ . . . . :i~a :
......
... :,: ...
....... ........
................ The lower (or input) accelerometer is used

to maintain the shaker at a constant sinusoidal.

............ acceleration level of 10 g over the frequency
range of 50 to 10,000 cps. The accelerometer
outputs are connected to a phase angle meter
which is used to measure the delay angle, or
time required, for a wave to traverse the speci-
men. Acceleration ratio is measured directly
on the true-rms voltmeter by scaling the upper
-accelerometer output to 1.0 when the directly
measured gain is 1.0. This occurs at low fre-
quencies. In addition, the accelerometer out-
puts are connected to a dual beam oscilloscope
to permit comparison of the wave shapes and
their phase relationship in a Lissajous figure.
Fig. 4. Specimen on
shaker for delay angle
technique PROCEDURE USING DELAY
ANGLE TECHNIQUE

in Fig. 5, the specimen which measures 2 in. by Initially, acceleration ratio and delay angle
2 in. by the desired thickness is adhered with were measured over the frequency range of 50
an epoxy adhesive to the 1-in. thick 4-in. diam- to 10,000 cps for a specimen consisting only of
eter steel specimen plate bolted to the shaker. a double layer of epoxy adhesive. The acceler-
ation ratio remained essentially 1.0 at all fre-
An accelerometer holder is adhered with quencies for this specimen, and no corrections
epoxy adhesive to the upper surface of the were made in subsequent runs. However, there
specimen, and the upper accelerometer is was an appreciable error in delay angle, and
bolted to the holder. The lower accelerometer the values of delay angle obtained in this test
is bolted to the undersurface of the specimen were subtracted from the values obtained with
plate. Both accelerometers are on the center- the viscoelastic materials to correct for errors
line of the apparatus. The upper accelerometer introduced by instrumentation and epoxy adhes-
thus measures motion of the specimen plate ive layers.
after the motion traverses the specimen, where-
as the lower accelerometer measures motion of To reduce errors at high frequencies, where
the specimen plate directly. the absolute amplitudes of vibration were in the

162
 300

24 r.5K 0-

zz
41240 a.SK

01.0 \ N*0 ,K0K

a
".120 z

, . L 5K U.
~ 0.560

100 200 500, 1K 2K SK 10K

FREQUENCY, C.P.S.

Fig. 6. Typical acceleration ratio, velocity and

delay angle in 1/2-in. thick ML-DZ

order of microinches, the accelerometers were 360 fT

mounted with a torque of 20 lb-in, on polished v , (6)

and oiled surfaces.
where
Normally, readings of acceleration ratio
and delay angle were taken at 50, 100, 200, and v = velocity through the material, in./sec;
500 cps and thereafter at 500-cps increments
to 10,000 cps. Readings were also taken of f = frequency of vibration, cps;
resonance frequencies and of frequencies at
which acceleration ratio passed through 1.0 or T = thickness of the material, in.; and
the delay angle passed through multiples of 90
degrees. 0 = delay angle, degrees.

As seen on the graphs, the minimum velocity of

- RESULTS AND ANALYSIS the vibratory wave occurred when the accelera-
tion ratio was a maximum.
Typical results of acceleration ratio and
delay angle obtained on a viscoelastic material Examination of the Lissajous figures and of
are shown in Fig. 6. The acceleration ratio had the shifted waves indicated that both the upper
a value close to 1 at low frequencies, rose with and lower accelerometers were vibrating at the
increase in frequency to a maximum at resonance, same frequency throughout the entire frequency
and then fell off with further increase infrequency, range, and that there was no apparent harmonic
An acceleration ratio of 1 indicated that the mo- distortion.
tion of the specimen plate was transmitted through
the specimen without amplification or reduction. In Fig. 7, the acceleration ratio (delay angle
An acceleration.ratio greater than 1 indicated technique) and percent c/c (disc and beam
amplification of motion and, hence, resonance methods) obtained on 1/4-, 1/2-, and 1- and 2-in.
in the specimen; a ratio less than 1 indicated thicknesses of ML-D2 are shown. The acceler-
that the upper portions of the specimen were ation ratio curves obtained for the different
vibrating at an amplitude less than that of the thicknesses of ML-D2 are similar to the typical
plate. (In sinusoidal vibration, displacement acceleration ratio curve shown in Fig. 6. The
ratio equals acceleration ratio.) peak value of acceleration ratio occurs at pro-
gressively lower frequencies as the thickness of
The delay angle was zero at low frequency, the specimen increases: 5000 cps for 1/4 in.,
increased rapidly to reach 90 degrees at close 2100 cps for 1/2 in. and 1900 cps for 1 in. The
!: 'to to the
riseresonance frequency,
with further increaseand then continued
in frequency. percent c/c c curves
as the values alsoc/c,
of percent exhibit a similar
increase to a trend,
maxi-
mum and then fall off with further increase in
The velocity of the vibratory wave through frequency. In addition, the maximum value of
the material is also sho.-i In Fig. 6. The velocity percent c/c c occurs at a lower frequency as the
was calculated by thickness increases: 9000 cps for 1/4 in., 7000

163

t-
 . ---o__-- ,.I,--- ~ ______

d 2.0 ACC. RATIO~-

2.o 16I
V 0153. -I
40 21

RiM -1

'00 200 0
5G0 K 2K SK 10K

FREQUENCY. C. P.S.

Fig. 7. Acceleration ratio and percent

c/c. of ML-D2

cps for 1/2 in., 3500 cps for 1 in., and 1300 cps substrate are vibrating and thus absorbing energy.
for 2 in. Furthermore, the peak value of per- The upper portions of the specimen are vibrating
cent c/c C for a 2-in. thickness occurs at ap- at an amplitude much less than that of the substrate,
proximately the frequency that could be pre- as indicated by the low acceleration ratio obtained
dicted from examination of the acceleration at these frequencies.
ratio curves.
In Fig. 8, the acceleration ratio and percent
At frequencies below 7000 cps, the percent c/c 0 obtained on 1/4-, 1/2-, 1- and 2-in. thick-
c/c c is dependent on thickness of material, nesses of the urethane material are shown. The
whereas above 7000 cps this relationship does shape of the acceleration ratio curves are similar
not hold; 1/2-in. thick ML-D2 has a higher per- to those obtained with ML-D2. In addition, the peak
cent c/c c than 1-in. material; at frequencies values of acceleration ratio occur at progressively
above 9000 cps, there is comparatively little lower frequencies as the thickness of the specimen
difference between the percent c/cc obtained on increases: 850 cps for 1/4-in., 630 cps for 1/2-in.,
1/4-, 1/2-, and 1-in. thick specimens. At these and 270 cps for 1-in. thickness. At frequencies
frequencies, the vibrations are apparently oc- above 2000 cps, the acceleration ratio is very low
curring so rapidly that only those portions of and there is not much difference between the ac-
the specimen that are close to the metal celeration ratio obtained on 1/4-, 1/2- and 1-in.

-- , ACC. RATIO
5- .2rN %C/c,

IN

4. I / - 34
3~,-

kI. .IIN 2..

100 '200 504' 1K 164i;2 K 10K

FREQUENCY, C.P.S.

Fig. 8. Acceleration ratio and percent c/ce,

of NASL-DlO1 urethane vibration damping
material

164
 thicknesses. Above 2000 cps, the percent c/c However, despite these differences, the
is rather low (peak value of 4.8 percent com- frequency ranges in which the materials have
A pared to 13 percent for ML-D2), and there is high acceleration ratio values show good quali-
comparatively little difference in percent c, cc tative correlation with the frequency ranges in
of the different thicknesses. The acceleration which they have good vibration damping prop-
ratio curves indicate that the urethane material erties. The correlation between the two param-
would have high vibration damping at frequencies eters (acceleration ratio and percent c 'c c ) im-
below 1000 cps. This is confirmed by the peak proves with increase in thickness of the mate-
value of percent c/ce of 5 percent obtained by rials. Apparently, as the mass of the specimen
the beam method at 450 cps on 2-in. thick ma- is increased, the effect of the added mass of the
terial. Examination of the percent c/c curves accelerometer and holder is decreased.
also indicates that at frequencies above 5000
cps a 1/4-in. thickness of material has a higher A viscoelastic material absorbs vibra-
percent c/c c than 1/2- and 1-in. thicknesses of tional energy because of the hysteresis effect.
material. Above 5000 cps it should also be noted As noted by Van Santen (6), the area beneath one
that the acceleration ratio is low, and there is cycle of the hysteresis loop wpoXo sin , and the
little difference due to thickness of specimen. work per unit time converted into heat by the
Apparently, at these frequencies, only the layers damping is
of material at and adjacent to the metal substrate
are absorbing energy. Furthermore, the non- Polo
vibrating portions may be inhibiting motion of 2 sin Q,
the specimen and thereby reducing energy ab-
sorption. where

O OApo = maximum force acting on the material,

/ COMPARISON OF ML-D2 AND NASL-
D101 URETHANE MATERIALS x. = maximum deformation produced by the
force po,
the ML-D2 ma-
As shown in Figs. 7 and 8,
terial has better vibration damping properties = angle that the deformation lags the
than the urethane material at frequencies above force, and
1300 cps, whereas the reverse occurs below
1300 cps. Furthermore, high acceleration i. = maximum velocity of the material.
ratios, i.e., greater than 1.0, occur over a
broader frequency range for ML-D2 than for The work converted into heat by the material
the urethane material. The narrowness of the cannot be calculated from the above equation,
frequency range, wherein the urethane has high since the force, velocity, and lag angle for each
acceleration ratios and consequently good vibra- portion of the vibrating material are unknown.
tion damping properties, may be due to the uni- However, with all other factors equal, the greater
form size and weight of the Ottawa River sand the velocity of the material, the greater the en-
particles used for the filler. ergy absorption and, hence, vibration damping.
This is consistent with the observation reported
herein, that at frequencies where the acceleration
DISCUSSION ratio and, hence, velocity ratio of the material
was high, the percent c/c c was correspondingly
There is an apparent disparity between the high.
conditions under which viscoelastic materials
are tested with the delay angle technique and As noted in the introduction, some investi.-
those encountered by the materials in service. gators attribute vibration damping to tension
For example: and compression forces parallel to the surface
of the material, whereas other investigators at-
1. There is the added mass of the accel- tribute damping to forces in the thickness-wise
;='erometer holder and the accelerometer (total direction. The investigation reported herein
weight = 84.2 gin) adhered to the normally un- indicates that damping is due to forces in the
adhered surface of the specimen. thickness-wise direction of materials which
dhave
particles of filler in a viscoelastic matrix
Sa 2. The specimen is vibrated sinusoidally
at a constant 10 g with swept single frequency,
and have a thickness greater than 1/4 in.
Furthermore, this investigation indicates that
whereas in service, the material vibrates simul- optimum damping occurs at frequencies where
taneously at a large number of unknown accel- there are resonances in the thickness-wise
erations and frequencies. direction. That damping in thick layers of

165
 viscoelastic material may be attributed to reso- viscoelastic materials over a continuous fre-
nances in the thickness-wise direction is sub- quency range of 50 to 10,000 cps. This tech-
stantiated by Bolt, Beranek, and Newman (7), who nique has been used to determine the frequency
note that damping increases, apparently due to . ranges in which different thicknesses of two
* resonance phenomenon, have been observed in types of viscoelastic materials have optimum
viscoelastic layers, where previous theories did vibration damping properties. The technique
not predict damping peaks. Furthermore, Bolt, has advantageous features in that acceleration
Beranek, and Newman (7) conclude that thickness- ratio and, hence, resonances in the niaterial can
wise wave motion is relatively soft and thick be determined on a relatively small specimen
viscoelastic layers may be primarily responsi- over the desired frequency range within one
ble for the damping behavior of plates to which hour. It can also be used to measure the veloc-
the viscoelastic layers are attached. ity of sound in viscoelastic materials vibrating
The various methods of rating the damping at or close to their resonance frequency. It
perious mets a o rct t has a disadvantage in that the mass of the ac-
properties of materials are not directly com-
parable. The most common indicator, percent celeronicter and holder influence the frequency
of critical damping, is not an intrinsic charac- at which maximum damping occurs. However,
teristic of the material but is dependent on both this influence is reduced as the thickness of the
the method used and the actual geometry within specimen is increased.
a method. For example, use of discs of identi-
cal resonant frequencies but differing in diam- The delay angle technique has shown that
eter and thickness would lead to different values vibration damping of 1/4- to 2-in. thick
of percent c/c c . At best, any method commonly vi.-!,iclastic materials at frequencies from 50
used allows the individual investigator to rank to 10,600 cps is due to resonances in the thick-
materials, ness-wise direction. Additionally, analysis of the
data obtained using this technique explains why
Most methods lead to the same ranking, vibration damping at all frequencies is not pro-
i.e., results are qualitative. What is needed, portional to the thickness of material. The delay
of course, is a material parameter which would angle technique can be easily automated.
be independent of the geometry and would char-
acterize the damping properties of the material.
The delay angle technique offers a method for ACKNOWLEDGMENT
determining the behavior of vibrating visco-
elastic materials which is independent of sub-
strate geometry. The authors wish to thank the following per-
sonnel of the U.S. Naval Applied Science Labora-
CONCLUSIONS tory: D. H. Kallas, C. K. Chatten, and H. Strobel,
for their interest and sponsorship of this work,
The delay angle technique developed in this and E. Zangrillo, E. Kelly, and L. Benjamin for
investigation is a tool which can be used to study preparing and testing the materials.

REFERENCES

1. MIL-P-22581 (SHIPS) "Plastic Sheet, Vibra- 5. House and Kritz, "An Analytic Study of the
tion Damping, Type ML-D2," Aug. 1960 Vibrating Free Disk," Presented at National
Ultrasonic Symposium, Stanford U., California,
2. J. E. Ruzicka, "Damping Structural Reso- Aug. 1959
nances Using Viscoelastic Sheer Damping
Mechanisms, Part I,"ASME Paper No. 60-WA-73
o i n 6. G. W. Van Santen, Introduction to Mechanical
3. D.H. Kallas and A.Rufolo, "Damping of Hulls Vbain Mciln e
by Use of Viscoelastic Materials," U.S. Naval Vibrations (MacMillan, New York), 1958, pp.
Academy, 1961 184-187

4. S. Timoshenko, Vibration Problems in Engi- 7. Bolt, Beranek, and Newman, Inc., Rept. 938,
neering (D. Van Nostrand Co., New York), 1937 June 1962

166
 Appendix

The general solution of C

dy,-- Ae YM t (A-5)
d12y dy
. _[ 2 c-d
m t(It--- y= Al Dividing y, by y,, we have

is that of a sirausoidal curve of

y constantly de- -- t C
creasing amplitude, shown below, Yde A mc I by
SSinc t e p eof
t (A-6)

woY equo is lge (g ) (A-7)

All, pressed in decibels

The relationship by
between y, and y may be ex-

=
_V20 ,
logo d (decibels). (A-8)

~
(A ~ osa) ~ ~ ~ si Y t) '
Since the falloff of amplitude from yI to y,
occurs in a time period At = (t2 - t,) seconds
the decay rate, in decibels per second, is

20 logly.-
whose equation is
(t 2 t)(A-)
y z e t(A sin &t + B cos cet),(-2
The number of vibration cycles n which can
where A and B are constants and w,is the angu- occur during this time period is
lar frequency.

'I
W / 2 4mk
4m 2
- (A-3) Therefore,

When c approaches 0, (t 2 - t1 )

or
rk . (A-4) /

is equal to 277f, where f is the frequency of 20 log1 0 \-A-l

vibration of the system. Dn/f

F To calculate the magnitude of the damping

c, two points yI and Y2 are selected on the
envelope of the sinusoidal curve, so that wt,
Substituting this into Eq. (A-7) yields:

and (t 2 equal 90 degrees. Substitution of these

values in Eq. (A-2) yields: log It) ' (tI (n

\t
C
y, Ae 2m (A-5) (A12)

2.303 logi0 = M
and (Y 2 m

167
 Since measurements are made at the natural
C 1 D (A-13)
n/f 2m 2. 30320 '
2f
0 x.C bt kf 4 0 M. (A-18)
""C =
D = 20 2-mx 233 4"34 m ' (A-14)

or
Dm 2
c = 4m(4 f 2m) . (A-19)
c =_ 4 (A-15)

f.f the damping c for this system is assumed from which

to be small relative to m and k, the natural fre-
quency is given by c c = 2m (2nf 0 ) . (A-20)

2 n(- Dividing the value of c by c c we get

The damping effectiveness of materials
can be reported in terms of the critical damp- c Dm 1
ing cc. For this condition, c is 0 and occurs nc2 ii
when (A-21)
D,
/c. 2
- 4ink = 0 _0.0183
4.34 x 4 x7fo fo

or (A-17) or

C D
c2 =4mk. %-c .83. (A-22)

DISCUSSION

Mr. Melnick (TRG Inc.): Has any concern Mr. Melnick: You constantly have altera-
been given to the proximity of welding to these tions to ships during drydocking.
joints? Has a test been run or has field informa- Mr. Eller: You can remove the material and
tion shown that this material will stand up after replace it. Since we have worked with the ML-D2
welding within a foot or within inches of it? that I have described in this paper, a sprayable
form of the material has become available. It
Mr. Eller: You mean after the material is can be sprayed in place after you weld.
applied, can you weld?
Voice: What kind of elastic moduli to these
Mr. Melnick: Nearby at least. materials have?
Mr. Eller: They are relatively soft, softer
Mr. Eller: No tests have been made but it than rubbers. We have not measured it, but I
would be best to weld first if possible. estimate about 8000 psi.

168

2I
 ENERGY ABSORPTION BY DYNAMIC CRUSHING*
1C. V. David
General Atomic Division, General Dynamics Corpo-ation,
Special Nuclear Effects Laboratory
San Diego, California

The dissipation of energy by crushing a structure is well known. If the

crushing process is slow, the energy absorbed corresponds to the
static crushing force times the amount of crushing. If the load is ap-
tplied instantaneously and exceeds appreciably the static crushing reac-
tion, a smashing process occurs and the specific energy absorption of
the system increases. In this case, the reaction of the structure is
considerably altered. The investigation of the dynamic behavior of
such a crushable structure is described.
Honeycomblike cores, 6 in. long, 12 in. in diameter, and of 0.008-in. thick
stainless steel corrugated sheets, were assembled between two steel
plates 1 in. and 1/4 in. thick. The 1/4-in. thick plate was accelerated
to 600 to 700 fps and smashed and crushed the core against the 1-in.
plate. Accelerometers mounted on the back of the 1-in. plate meas-
ured the reaction from the core against this plate. Similar samples
were crushed statically, and the dynamic and static specific energies
were compared. For the plate velocities investigated, an improvement
by a factor of 2 to 3 of dynamic loading over static loading was obtained.

INTRODUCTION test program was initiated at General Atomic

to investigate the behavior of realistic designs.
The dissipation of energy by crushing spe- Objectives were (a) to ascertain whether (and to
cially designed structures (such as honeycomb- what extent) the specific energy dissipation of a
like cores) has been used for several years, and crushable core is higher under rapid loading
off-the-shelf items are available commercially. conditions than it is under static, or slow-
The crushing characteristics of such structures crushing, conditions; (b) to ascertain the behav-
have been extensively studied experimentally ior of the core in the vicinity of the primary
and analytically (1). But in all cases, the crush- impact plate after the impulse is delivered to
ing of the core takes place on a time scale such the plate; (c) to obtain information on the reac-
that local buckling of the cell walls can occur tion of the core against the supporting structure
ahead of the crushing plate. This produces a as a function of time, since it was believed that
uniform crushing pattern along the core length the relatively constant force that existed during
until bottoming-out is reached. If, on the con- static crushing tests would not exist under dy-
trary, the crushing plate moves appreciably namic loading; and (d) to demonstrate the ability
faster than the rate at which local buckling can of the system to function under actual dynamic
propagate,. a different process takes place and conditions.
Va the core metal is smashed into shavings. This
process produces two effects: the energy dis- To achieve these objectives, it was decided
sipated per unit weight of core increases, and to test scaled-down models of the full-size
the time history of the reaction transmitted to crushable-core system. In these tests, the pri-
the supporting structure is altered. mary impact plate was to be given sufficient
impulse to obtain realistic dynamic test condi-
The use of high-strength steel cores for a tions. The results could then be compared with
dynamic application has been considered for results from a static crushing test. The energy
cases where the crushing plate might have ini- source to deliver the proper impulse to the plate
tial velocities of 500 to 1000 fps. Therefore, a would be developed from data already available

*Research sponsored by Defense Atomic Support Agency, Department of Defense, under Contract
DA49-146-XZ-386.

169

.,- .
from high-explosive (HE) testing techniques, solid mass. With this plate attached to the core,
The experimental program was thus divided the energy dissipated in the core is about 82
into the development of the energy source, the percent of the absolute kinetic energy of the
dynamic crushing tests, and the static crushing 1/4-in. plate. Two eyebolts were mounted on
tests. the 1-in. plate to allow the assembly to be in-
stalled for the dynamic tests, as shown in Fig. 2.
The source development and dynamic
crushing tests were carried out at the General Provisions were made for mounting twc
Atomic Green Farm Test Site (2). At the HE accelerometers on the back of the 1-in. plate to
test facilities available at the site it is possible record the force transmitted by the core to the
to recover plates weighing several hundred plate during the crushing process as a function
pounds and moving at a velocity of a few hun- of time. Since the energy dissipation capability
dred feet per second without damage to the plate of the core would have been too large if the core
from ground impact (2). A number of plates had completely covered the area of the plate,
were accelerated to velocities of 600 to 900 fps. the core density was effectively decreased by
Following these experiments, three dynamic making six radial openings, as illustrated in
tests of quarter-scale core assemblies were Fig. 1. A total of five assemblies were built:
made. Each assembly consisted of a 1/4-in. three were tested dynamically, one was tested
steel plate brazed to a crushable core that was statically, and the fifth was kept as a spare. An
bonded to a 1-in. thick steel plate. Instrumen- assembled model is shown in Fig. 3.
tation consisted of two accelerometers mounted
on the 1-in. plate and a high-speed motion pic-
ture camera from which the velocity and orien- ENERGY SOURCE DEVELOPMENT
tation of the moving assembly could be ascer-
tained. In the static tests, a core assembly was A series of ten experiments was performed
crushed on a Baldwin compression test machine, to determine what combination of sheet HE and
polyurethane foam damper would give the re-
quired impulse. In these experiments, 12-in.
MODEL DESIGN diameter plates, 1/4 in. thick, were mounted
inside a heavy steel ring which acted as a bar-
The models were designed to provide quick rier and positioning device. The energy source,
preliminary results that could be used to pre- which consisted of circular sheets of HE and
dict whether the full-size system design was polyurethane foam, was placed in contact with
sound and what could be expected from a known the 1/4-in. plate and flush with one side of the
configuration. The choice of materials was lim- steel ring. The foam was positioned between
ited, since they had to be immediately available the HE and the plate to modify the impulse de-
and have properties compatible with the avail- livered to the plate. Combinations of three foam
able manufacturing and assembly processes. densities, two foam thicknesses, and two HE
The following combination of materials was weights were used. The plate velocities were
selected: measured from films taken by a Fastax camera
which viewed the first 20 ft of plate travel. The
Core: AM-355 CRT-XH stainless steel, test results are given in Table 1.
0.008 in. thick and having a strength of about
150,000 psi yield after brazing at 1250F (total Since the accuracy of the results is esti-
core weight of approximately 4 lb); and mated to be only within about 10 percent, one
can conclude from Table 1 that the total amount
Plate: USS TI for both the 1/4- and 1-in. of foam (thickness times density) is almost im-
plates, having a strength of about 100,000 psi material and that the magnitude of the plate ve-
yield. locity is determined primarily by the amount of
HE used. One obtains, roughly, 680 fps with
The original plans called for both plates A-3 and 900 fps with A-4.
and the core to be brazed at a maximum tem-
perature of 1250'F. But since the large mis- For a given amount of HE, the pressure
match in thickness between the core wall and peaks developed against the plate should in-
the 1-in. plate was inconipatible with an accept- crease for decreasing amounts of foam. These
able braze, the 1-in. plate was bonded to the pressure peaks have not been measured but
core with an adhesive. The parts were all built sho-,ld be in the 25- to 50-kbar range for the
according to the drawing in Fig. 1. 0.25-in., 2-lb/ft 3, A-3 combination of Table 1.
The 1/4-in. plate used for this test did not ex-
The 1-in. plate used as a back support for hibit any spallation. Only a slight dishing was
the assembly simulates the inertia of a large noticeable.

170
3/32-IN. DIA VENT HOLE THROUGH EACH TRIM CORE TO 12 IN. DIA 1.0 IN.6.IN
CANTED WES. LOCATE 3 IN. FROM END6.IN

-~~~ LOCATE CENTRALLY - -

I ~~SPOTWELD
ATTACH,___ __

DIA
DIA LOCATE
HOLES
L.. INI-N.
PLATE 'X

THI

0.75 (TYP) REF

STEEL
0.008-41N. .- ___- ___

SHEET AM-355 CRT-XH - - --

TYET TE LT
I IN. X 12 IN. DIA BRAZED OR BONDED

TYPE TI STEEL PLATE

0.25 IN. X 12 IN.
Fig. 1. Test model design layout

TABLE 1
Velocity (Impulse) Measurements

Foam Configuration Plate Plate Estimated

1H Velocity Velocity Initial Plate
Thickness
(in.) J(lb/it
Density
3) ____
at 12 Ft
(fps) j
at 6 Ft
(fps)
Velocity
(fps)

0.5 2 A-3 585 605 630

0.5 2 A-4 830 860 900
0.5 4 A-3 655 -- 700
0.5 4 A-4 865 -- 920
0.25 2 A-3 670 680 700
0.25 2 A-4 840 860 800
0.25 4 A-3 605 640 600
0.25 4 A-4 815 895 950
0.25 10 A-3 630 -- 660
0.25 10 A-4 820 860

*:A-.3 and A-4 represent sheet HE containing 3 and 4 gm/in.2 of HE,

respectively.

171
 -.... sample dynamic tests. An HE-polyurethane

'1 foam configuration of sheet A-3 and 0.5-in.

thick, 2-lb/ft 3 foam slab was adopted. The en-
ergy source and the test sample were installed
in a manner similar to that described for the
source development tests.

Three samples were tested: two were

slightly precrushed and one was not. The pre-
crushing was performed in such a way that the
core was crushed about 0.25 in. at the juncture
o ~with the 1-in. plate.

Non-Precrushed Test Sample

This first test was only partly successful

because of difficulties with high frequency os-
cillations in the accelerometer signals. As a
result of this experiment, the accelerometer
signals in the remaining two experiments were
filtered to attenuate the high frequency compo-
nents.

This experiment showed that the 1/4 in.

plate separated from the core at an early time,
probably at the time at which the plate stopped
' - /crushing the core. The motion picture film re-
i:.. _ _ - vealed that the 1/4-in. plate actually recoiled
from the 1-in.-plate-core assembly. Appar-
Fig. Z. Crushable sample installed ently, an appreciable amount of energy was
for dynamic testing stored elastically in the core at the end of the
crushing process. After separation, the 1-in.
plate had a velocity of 120 ft/sec and the 1/4-in.
plate a velocity of 96 ft/sec.

The test sample is shown in Fig. 4. The

crushing of the core is quite different from
that usually obtained in static tests. It is seen
to have occurred primarily or, the end to which
the 1/4-in. plate was attached. The core metal
is not crushed per se but is smashed into
chips, which are pressed together against the
1/4-in. plate. There appears to be no real
crushing except for half a convolution next to
the smashed portion. The remaining part of the
core was intact. The smashed length of core
was about 3.5 in. The complete destruction of
the integrity of that part of the core next to the
1/4-in. plate during the smashing process ex-
plains why the 1/4-in. plate could not remain
attached to the core.
Fig. 3. Crushable sample
First Precrushed Test Sample
DYNAMIC CRUSHING TESTS Since the core of the non-pre rushed test
sample (Fig. 4) did not develop the expected
For the full-size application contemplated, crushing pattern at the 1-in. plate connection
a plate velocity of 630 fps was fairly realistic, and, therefore, probably created much higher
so this velocity was chosen for the crushable crushing force peaks than were expected, it was

172
L

decided to precrush all cores at the 1-in. plate test, but it has not been resolved. It appears
connection. The dynamic behavior of the core that the calibrations of one or both of the ac-
was considerably altered by the precrushing, as celerometers had changed from their original
revealed by Fig. 5. values, but the tight scheduling of this task did
not permit a recalculation to be performed. It
has thus been tacitly assumed that while the ab-
solute values of the acceleration signals are
questionable, the shapes of the acceleration
versus time records are correct.

Fig. 4. Non-precrushed test sample

4 after dynamic test

.,' Fig. 6. Accelerometer signal, accel-

eration vs time (second dynamic test):
vertical scale = 1000 g/cm (positive
downward); horizontal scale 0.2 ms/
cm starting at time of detonation (each
division along scales = 1 cm)

- .
,, The average maximum acceleration indi-
cated in Fig. 6 is about 6000 g and is, therefore,
more than twice that expected (2500 g). But the
duration.of the signal was, as predicted, about
2.3 ms. The 1/4-in. plate again separated
from the core. The following velocities were
measured from the recording film: 1/4-in.
Fig. 5. First precrushed test plate, 150 fps; 1-in. plate and attached core,
sample after dynamic test 110 fps.

After testing, the core was found to be Second Precrushed Test Sample
z
'c smashed at the 1/4-in.-plate end and also
crushed at the 1-in.-plate end. The crushing In this test, good records of the acceler-
stroke was about 4.25 in. long, which was 0.75 ometer signals were obtained. Figure 7 shows
in. longer than that measured for the previous the signal from one of the accelerometers,
test. The magnitude of tie 1-in.-plate acceler- illustrating the shape of the acceleration versus
ation, as indicated by the two accelerometers, time curve. It is noted that the signal shows a
was larger than anticipated. The one good ac- peak acceleration which is quite high compared
celerometer signal record did not include the to the average value. Prior to this test, the
peak acceleration, although it can be estimated. precrushing of the core of this sample was ob-
This signal is shown in Fig. 6. Records of the served to be azimuthally unsymmetrical; i.e.,
signal from the other accelerometer indicate the core was not completely precrushed aruund
that it apparently sensed about twice the accel- the entire perimeter. One would, therefore,
eration measured by the first accelerometer. expect a rather high peak acceleration corre-
A difference in the two accelerometer indica- sponding to the force transmitted to the 1-in.
tions was also observed in the third dynamic plate before initiation of the crushing action.

173
 This interpretation is borne out by the fact that
in the first precrushed test, the core of the as- . ;f.-ro
sembly was observed to be rather uniformly
precrushed and the acceleration signal of Fig. 6
did not appear to have very high peaks. -'---'", (a)

Er4

. (b)
Fig. 7. Accelerometer signal, accel-
eration vs time (third dynamic test):
vertical scale = 5000 g/cm (positive
downward); horizontal scale =0.2 ms/
cm starting at time of detonation (each
division along scales = 1 cm)

Again, the 1/4-in. plate separated from the

core, and the following velocities were meas-
ured: 1/4-in. plate, 125 fps; 1-in. plate and
core, 105 fps.

Two photographs of the test sample are

shown in Fig. 8. The top of the 1-in. plate with Fig. 8. Second precrushed test
the accelerometers still connected is also sample after dynamic test
shown.

Figure 9 shows the three dynamic-test

samples after the experiments for comparison
with the sample before test. The 1/4-in. plate
is also presented to show the plate edge bending
caused by the uneven support of the core.
( --
STATIC CRUSHING

The fourth sample was crushed at a very

slow rate (1 in./min) on a Baldwin compressior.
test machine to determine the static crushing
characteristics of the core. The crushing
process was stopped when the 1/4-in. plate bot- 3 6 7 20.
tomed out. The crushing force was recorded (b)
versus displacement, and the results are pre- Fig. 9. Dynamic test samples before
sented in Fig. 10. A force of 72,300 lb was re- and after test
quired to initiate the crushing, and a constant
force of 61,000 lb was sufficient to continue it.
The initial length of the core was about 6-in. The crushing force was quite constant, ex-
and the crushing travel was about 4.4. in., indi- cept for some irregularities corresponding to
cating that about three-fourths of the core each local buckling convolution along the core
length is available for a compression stroke. length. The main disturbance, indicated by the

174

-4-..
J

340 _ _ _ _ _ _ _ _ _

320-

300

280

260
DYNAMIC
240- CHARACT'ERISTICS

220

~1200-
0'
-- oo

.I180.
Id

o 160

~STATIC
cr 120- CHARACTERISTICS

too-.

60 Fig. 11. Precrushed test :4

4 sample before and after
static testing
i201

0 _difficulties encountered with the accelerome-

CRUSHED LENGTH (IN.) ters, the brief pericd allotted for this task per-
mitted only the execution of the few tests just
Fig. 10. Crushing force vs core length described. These tests did yield a considerable
amount of information, however.

arrow in Fig. 10, was caused by the 1-in.

spacer clip shown in Fig. 1. The bottoming-out Influence of Precrushing
is indicated by the beginning of a sharp rise of
the force curve. Ifthe core is not precrushed, and espe-
Tcially if both ends of the core are restrained
p(brazed or bonded to plates), the onset of local
was 61,000 x 4.4/12 = 22,400 ft-lb, and since
the core weighed about 4 lb, the energy dissi- bucklang, which is required for crushing, occurs
pated per pound of core material was 22,400/4 = at a much higher load than the static crushing
pa,00ft-lb/lb. Since te core wall thickness force. Furthermore, if the load is applied very
o,600 rapidly and for a very short time, it is doubtful
was 0.008
10,00
pi,in. ne
andcn
the pedit a yield
material enegyabout
isspa- that the core will buckle even locally; in this
Stion capability for an optimum cell wall thick- case, loads up to and above its static compres-
150,000 psi, one can predict an energy dissipa- caeloduptanabvissaicomr-
oness of 0.030 in. and 300 series maraging steel sion strength migh.: be transmitted. This would
of 00 300,000/150,000 (0.030/ result in too high loads on the support structure.
of about 5,600 X 3t-l0/0. This loading peak phenomenon is evident in
02 fFig. 7. One advantage, of course, is that all the
The stati.c test sample before and after energy goes into smashing the core, which is a
crushing is
c i shown in i Fig. 11. One readily no-
1amuch more efficient way to dissipate energy, as
tices the difference between this crushing pat- will be shown later. For instance, the same
tern and that obtained in the dynamic tests. amount of kineti.c energy was dissipated in the
non-precrushed test sample as in the second pre-
crushed test sample, but in the non-precrushed
DISCUSSION OF RESULTS sample it took only 3.5 in. of core (with smash-
ing and negligible crushing) as compared to 5.5
Although it would have been desirable to in. of core (with smashing and appreciable
perform more dynamic tests and to resolve the crushing) for the precrushed model.

175

"4 R
 The precrushing, therefore, plays two TABLE 2
roles: Typical Plate Velocities

1. It reduces the early peak load to toler- Late-Time Calcu-

able levels; and Velocity of Late-Time lated
Dynamic 1-in. Plate Velocity of "Initial"
2. It filters out some of the compression Test with Attached 1/4-in. Velocity
stress-wave effects initiated by the early-time Core Plate of 1/4-in.
smashing of the core and tends to eliminate (fps) (fps) Plate
high-frequency, high-amplitude vibrations. I (fps)

The disadvantage of precrushing is that it de- 1 120 96 683

creases the efficiency of the energy dissipation
process, since part of the energy is used for 2 150 110 784
crushing the core at the connection to the su-
port structure. 3 120 105 668

The physical interaction during the dy-

namic crushing of a precrushed core assembly All of the calculated initial velocities are
appears to be as follows: higher than the expected 630 fps, probably be-
cause the interaction time between the plate and
1. The core end attached to the moving HE source was greater owing to the slow move-
plate is smashed and a compression stress wave ment of the more massive plate-core assembly
is transmitted to the core wall. This stress as compared to a plate alone.
wave, trapped in the core, accelerates the un-
smashed part of the core until it has reached
the velocity of the moving plate. Some crushing Energy Dissipation
of the core probably occurs near the 1-in. plate
during this process. The difference between the initial and final
kinetic energies of the core assembly is the en-
2. When the core and the 1/4-in. plate have ergy dissipated in the core. By using the con-
the same velocity, the core is crushed against servation of momentum equation, this kinetic
the 1-in. plate. energy dissipation can be expressed by

3. The end result is a core half smashed W ( W

and half crushed, as depicted in Figs. 5 and 8. Ek 2 g p W
The smashing is similar to that shown in Fig. 4,
and the crushing is similar to that obtained dur-
ing the static test (Fig. 11). where w is the weight of the primary impact
plate (i.e., the 1/4-in. plate), Wt is the weight
of the total assembly, and V is the initial ve-
Velocity Measurement locity of the 1/4-in. plate. Xs w, = 8 lb and
wt = -44 lb, the energy dissipations Ek in the
To evaluate the energy dissipation of the three dynamic tests are 48,000 lb-ft for the
core in the dynamic tests, it is necessary to first assembly, 63.000 lb-ft for the second as-
determine what energy is actually imparted to sembly, and 46,000 lb-ft for the third assembly.
the primary impact plates in each of the tests.
This can be determined if the "initial" velocity Since about 4 in. of the 6-in. core were
of the plate is known. Although the initial ve- smashed for the first assembly (no crushing),
locity of the plate was nominally 630 fps based the specific energy dissipation of this 4-lb core
on the results of the energy source development was 48,000/4 x 6/4 = 18,000 lb-ft/lb. For the
test previously discussed, it was necessary to second and third assemblies, the energy dissi-
ascertain what the velocities in the dynamic pation was 15,400 lb-ft/lb and 11,800 lb-ft/lb,
tests actually were. Data taken from the motion respectively. These figures indicate (a) that the
picture films of the moving assembly yielded smashing action dissipated almost 220 percent
the late-time velocities of the 1-in. plate with more energy than the static crushing; and
attached core and the 1/4-in, plate. This infor- (b) that for the plate velocities considered and
mation, given in Table 2, when inserted in the the core-to-plate weight ratio used, a combina-
equation for conservation of momentum, gave tion of smashing and crushing yields up to 175
the actual initial velocities of the 1/4-in. plates, and 112 percent more energy dissipation than
which are also presented in Table 2. static crushing. As a first approximation, then,

176

7 -77 *'----
 one can assume that for the plate velocity and where k is the correction factor and a(t) is the
the core density-to-plate weight ratio used in 1-in. plate acceleration given by the scope trace
these tests, the dynamic crushing yields up to of Fig. 7. The step integration of this curve
twice the specific energy dissipation of the gave k = 0.59 for t f = 2.2 ms, which was when
static crushing. the acceleration dropped to zero.

Since the total amount of crushing in the

Core Reaction Against the third dynamic test is known to have been 4 in.,
Support Structure the relative distance between the two plates,
expressed in terms of the initial velocity and
Two disturbing factors appeared during the normalized acceleration, can be set equal to
dynamic tests: the existence of high-frequency 4 in.:
oscillatory acceleration and the presence of
high-magnitude acceleration spike, as duration
shown in ~ f ()d n 3
V Fig. 7. Although the spike is of short )d

and might disappear if an adequate amount of

precrushing were present, the high-frequency
oscillation would
respond to the probably
small It may by
remain.exhibited
oscillations cor- Here, the 1/4-in.a plate
an acceleration is assumed
'(t), related to receive
to a(t) by the
the static crushing force, magnified many times mass ratio of the two plates. Thus,
because of the influence of the time element on
buckling. The influence of this oscillatory force a'(t)
on the response of the support structure should a t 4
be investigated.
The results obtained were
used to plot the dy-
Also, it appears that the smashing-crushing namic response curve of Fig. 10.
of the core does not result in a constant force
as does the slow application of the load during One can then integrate the dynamic test
the static tests. For instance, an approximate curve of Fig. 10 to obtain the work done by the
evaluation of the average reaction of the core core reacting against the 1-in. plate. It is
against the 1-in. plate versus core crushing
stroke is shown in Fig. 10. This does not com- f
pare well with the static crushing force. It ap- dl' R dx,
Fx) (4)
pears that after the peak force is reached, the JJ0
uncrushed core has accelerated enough that the
crushing against the 1-in. plate takes place and where F(x) is the curve just discussed and
produces a constant-level force comparable to xf 4 in. This yields 30,500 lb-ft. But
* but slightly higher than the static crushing load. since the core has a mass (equal to about one-
half that of the 1/4-in. plate), by inertia it re-
acts against the 1/4-in. plate with a higher
Smashing-Crushing Energy force than F(x). The core can be considered as
a lumped mass between two springs. In this
The dynamic response curve of Fig. 10 was case, part of the 1/4-in.-plate kinetic energy is
obtained from corrected readings given by th, trapped in oscillations within the core and does
accelerometer signal. The amplitude of the not contribute to the reaction against the 1-in.
signal was corrected so that the time integral plate. A similar study indicates that 25 to 30
of the acceleration, percent of the 1/4-in.-plate kinetic energy could
have been dissipated in the core only to accel-
erate it. Therefore, the total energy absorbed
a(t) dt by the core could be 30,500/(0.75 to 0.70)
40,600 to 43,700 lb-ft. These numbers compare
reasonably well with the 45,000 lb-ft calculated
was equal to the 1-in.-plate velocity as measured as the kinetic energy of the plate.
by the motion picture film. In the case of the
third dynamic test, for example, this would be
CONCLUSIONS

'(t a(cit 123 fps, (2) The experimental program described in

a1 this paper yielded very interesting results:

177

_ _ _ _ _ _ _ A -
- _ _

.. _./..--..
 1. The specific energy dissipation achiev- structures is extremely high and could lead to
able by honeycomb core smashing can be two to very compact and light systems.
three times as high as that obtainable by crush-
ing. This smashing occurs if the smashed end 5. Considerable work would still be re-
moves at high velocities (a few hundred feet per quired to determine more accurately the influ-
second). ence of the plate velocity on the smashing
process, of precrushing, of the core cell con-
2. The precrushing of the core at the sup- figuration, and of the core material. The pres-
port attachment affects noticeably the amount of ent investigation was only intended to probe this
specific energy dissipation (up to one-third) and new area and constitutes a spot check of the
the core reaction against the support. problem.

3. The smashing process completely de-

stroys the core connection to the moving plate. ACKNOWLEDGMENT
The latter always recoils after the interaction
is completed. The author wishes to acknowledge the con-
tribution to this work of Ronald E. Rinehart,
4. For a "one-shot" application, the en- who supervised the HE testing and performed
ergy dissipation capability of such crushable some of the analytical work.

REFERENCES

1. R. K. McFarland, Jr., "A Limited Analysis 2. "Technical Summary Report, Nuclear Pulse
of the Collapse of Hexagonal Cell Struc- Propulsion Project, Vol. IV, Experimental
tures Under Axial Load," Jet Propulsion Structure Response (U)," Air Force Sys-
Laboratory Tech. Rept. 32-186, Dec. 1, tems Command Rept. RTD-TDR-63-3006,
1961 July 1963

* *

178
 DRY FRICTION DAMPING WITH FORCE
1PROPORTIONAL TO DISPLACEMENT
Leon Wallerstein, Jr.
Lord Manufacturing Company
Erie, Pennsylvania

Conventional (constant force) friction dampers have certain shortcom-

ings which inhibit their use in vibration isolation systems. Among them
are a potential for infinite resonance amplitude, a limitation on force
isolation, the creation of static off-center conditions, and the introduc-
tion of harmonics not necessarily in the environment. If the friction
force is made proportional to displacement, these deficiencies disappear
or become unimportant, and exceptional vibration isolation can be real-
f ized. One form of such a damper is shown which exhibits both an elas-
tic stiffness and a friction gradient for which expref.sions are given.
The friction gradient is two-valued and depends on the relative signs of
displacement and velocity; its dual nature shows clearly in force-
displacement tests. Self-centering is assured when the elastic gradient
exceeds the minimum friction gradient. Analog computer solutions for
a single degree of freedom system are given as transmissibility-
frequency curves for several damping ratios. The decay rate at high
j frequencies is 12 db/octave, equal to that of an undamped system.
Resonance transmissibility depends only on the ratio of friction gradi-
ent to elastic gradient; it is (analytically) independent of input amplitude,
as confirmed by model tests.

INTRODUCTION friction force is proportional to displacement.

The force is, of course, directed against veloc-
Most vibration isolation systems incorpo- ity, and when such a damper is made to execute
rate some form of damping, among them dry, harmonic oscillations, it generates the wave
sliding friction which has a number of attrac- form shown in Fig. 1.
tive properties. It also introduces problems
which militate against its use. Our purpose is
to show how a modification of this common DESCRIPTION OF A DISPLACEMENT-
damper form can yield uncommonly good re- PROPORTIONED (DP) DAMPER
sults.
Form and Force-Displacement
What attracts us to friction dampers is Properties
that they are relatively simple, have no leakage
problems, and are quite insensitive to changes The single degree of freedom system of
in temperature, frequency, and strain ampli- Fig. 2 shows 'one configuration such a damper
tude. This. cannot be said of fluid dampers and might take. At static equilibrium the damper
of many elastomers. On the other hand, con- spring kd is unstressed; displacement in either
ventional (constant force) friction dampers direction creates a proportional tension in kd
have these shortcomings: (a) relatively high and a corresponding force on the sliding sur-
transmission of vibration in the normal "iso- faces.
lating" range, (b) the characteristic of infinite Thus, a friction gradient (A) is established.
:, resonance when input exceeds a critical value
reoacewe ipteced ciialvle Its dimensions are like those of an elastic gra-
i an it
{ the creation of static offset, and (d) the
(1), .(c) dient or stiffness, ee.g., pounds
diensifns pose per inch, and it
introduction of harmonics not necessarily in has two values, given by
the environment. This is a formidable list of
defects, but they disappear or become unim- Al 41 t a( + tan 2 a)
portant, and exceptional vibration isolation - 4 t ( tan a) (1)
results, if the damper is so designed that -d IL tan a)

179
 when x and x are of like sign, and 40

A2 41 tan a(I + tan2 ) (

k," (l ! tan '1) (

,,when x and .xare of unlike sign. o2o-- Slope

L. (K+A)

o (K-A)
isplcement
_Dcerier L I
r e -. 15 -. 10 -. 05, , ,
.05 .10 .15
DISPLACEMENT
(INCHES)

(K-A) -10
Velociy 1<(K+A,)

I 1 1 -20 "
-0
2 2 r

Fig. 1. Wave form, displacement- -30

proportioned (DP) damper Fig. 3. Force-displacement
curve for model of Fig. 4

decreasing displacement, it is K- A2 . The

k, damper spring measured slopes correspond well with calcu-
stiffness lated values, being respectively 280 lb/in, and
Form of 279 lb/in, for (K+ A,), and 168 and 149 lb/in.
damper f= coefficient of
body friction for (K-A2)'

tPosin wt km= stiffness of

springs acting
Returnability
directly on mass Self-centering of the system is assured if
m the spring gradient exceeds the minimum fric-
tion gradient; i.e., if K >A2.

Fig. 2. Single degree of freedom

system with DP friction damper Harmonic Composition
Figure 5 and Table 1 show coefficients for
the Fourier series which represents both the
The elastic stiffness of the system is given square wave and the DP damper force-time
by relation; i.e.,
Kk
= +
+km,
K e k ,(
(3)
)f(t) = Z-F
n sill t ,(5)
where
(4) where n = 1, 3, 5, .... .We are interested pri-
k 4k tan 2 a (4) marily in the first order since it is only this
component which contributes to damping (2).
is an elastic stiffness due to the damper, and k. For the conventional damper, the coefficient is
is the stiffness of springs acting directly on 4F o, 77, compared to 2Fo/7, for the DP damper,
mass (in). and the former might, superficially, be consid-
ered more "efficient." In addition, its third
Figure 3 is the result of tests on the model harmonic is smaller, so it generates less spu-
shown in Fig. 4, and demonstrates clearly the rious vibration.
two-valued nature [Eqs. (1) and (2)] of the fric-
tion gradient. For increasing displacement (x These advantages are more apparent than
and k alike), the slope is K + A, and for real; at high frequencies, where displacements

180
 I are generally small, the conventional damper
continues to transmit large forces and, there-
fore, degrades the isolation system. By con-
trast, the DP damper generates large force
when it is needed and small force when it is not;
this quite outweighs its less favorable harmonic
I .. structure.

SYSTEM CHARACTERISTICS

r ",Equation of Motion and Solutions

VM
From Fig. 2, we derive

m + A
ni lxi + Kx P0 sin t. (6)

The damping term is proportional to displace-

ment by virtue of lxI and has the sign of veloc-
ity due to k1li. The value of A can be taken
as 1/2 (A1 + A2) with very little error. Solutions,
by analog computation, are shown in Fig. 6 for
Fig. 4. DP damper test model the range A/K = 0.1 to 0.5. The transmissibility
at each frequency is the ratio of peak trans-
mitted force to peak excitation force taken from
oscillograph time-force histories, such as in
Figs. 7 and 8. Thus, true trnsmissibility is
AMPLITU E represented, despite nonlinearity of the damper.
OF
FHARMONIC A CONSTANT FORCE D4M Equivalent (same f, and Tres) viscous-damped
_x 4
. A ~~j-E systems are shown for comparison. For DP
B FORCE PRORTIONALTO damping, high-frequency.transmissibility de-
x3 DISPLACEMENT creases at a rate of 12 db/octave, equal to that
F. 4 of an undamped system. The rate for viscous
damp,,,g is 6 db/octave. Conventional friction
2 j Idamping produces constant transmissibility (of
A at least 7/4 in magnitude), since both trans-
Si A.
mitted (friction) force F and excitation force P0
A 8are unchanged with frequency. The high trans-
- A missibility level results from the fact that
I 3 5 7 9 F/P = 77/4 to avoid infinite resonance ampli-
tude (1,2).
Fig. 5. Harmonic composition
! 'of damper force
Wave Form

TABLE 1 The damper is a nonlinear element, and

_______

Damper__ 1_
_

3 5
__T
Coefficients of Harmonic Components
]though
9
-
distorted responses can be expected even
sinusoidal inputs are applied. Figure 9
shows all force components within the system
for a typical condition, A/K = 0.3 and resonance
Square 4Fo 4F 4F 4F 4F operation. The damper force closely matches
,> wave ,7 3I 57 77r 9n the idealized form of Fig. 1. The body motion,
however, having the form of Kx, is nearly
Displacement- 2Fo 2F0 2Fo 2Fo 2Fo sinusoidal; therefore, we justify the assumption
proportioned 7T I 37, 37 57 leading to Eqs. (8) and (9), which follow.

181

pim V-
 100 TRANSMISSIBILITY: DP damping vs. I
viscous dampina
5 (for some (nand Asn wt
[ 1Tres) 15
II9 =1
=.032] K
2\ i=-Oo

,.............................................,."...... ..

. . . (Constant)
. .- - Friction
- -< o- - -,.,- - .
~262

LLO

C.
I'_ (Cntat Fnsananouttanoite F0g
L-I! 5-Damped
a. Z I
"~u z I IO a *d
2 Viscoorefrusiosdmpn rto
D-

03.2 ,., ..
Fig. Fig.re.uenstantaneous.transmitted
na<e X h nycmonn fdme W

__ TIMiEi
the work extaforcebforhevariousfodampinge-ratios

FoW ytmectdb iuodlmto f pyia oe eut nTbe2

Fig..86.nsFrequencys tratiotte

Resonac rasmissibFlityXr(T 2~r s . (7

Sicmfexecutes harmoncsmotiongweacans 182 0
eqat thedin 5-
wor inu byeharnic P force
a 2kK()2kK
/ P .032
0 to.e -
0C 2/fG/'()
thae work, extractedrbyothe odamperoforceitotde-

terhe potthe displacement amplitude aeo-,

2- .16 foc o-aiusdm igrto

fore caablerofx acting

freen e (2)yste _sineetngt
_oeh eato mn
setin'
:m aoo SusttuinginE2 (7) ig.ous. anhstteess transmr-ited mp

noigtaX /,weaanhv n resonance transmissibility:dpnsol

ihrespett nAK
- .s (8) Iti unfece by inu amltd orfeuny
=st 2 A/K Caclae 2a are cae wih"naogan
 I I P sin wt
TRANSMITTED T A

I ,,FORCE l.f .=0lNRT

FORCE K A
INERTIA

APRING
SOC 1.0

ifiwo

II

Fig. 9. Force components at resonance in a single

degree of freedom system with DP friction damping

TABLE 2 7a- -2-

Resonance Transmissibility IO"

:-:=

Eq. By By Model at Inputs 6-

SBy Eq. B (s.a.)
*Analog (~. -_
AA (8)or Con-
Eq. (9)puter 0.005 0.010 0.020 . - - -

ptr In. In. In.

0.1 15.7 11.2 -- - - _j

0.3 5.3 4.1 6.20 6.25 6.45 :E 3

0.5 3.1 2.5

- 2 -

In this, c and c, are, of course, actual and

critical damping for the viscous system; G" and I
G' are damping and elastic components of the
"rubber" modulus.

'.
,0 20 30 40 50
FREQUENCY - CPS
Model Tests Fig. 10. Effect of input
on DP damped .nodel
Static force-displacement curves, from the
~
" model shown in Fig. 4, have already been dis-
cussed. Vibration tests through the resonance
range at three different inputs are shown by vertical forces on the mass (m)with lateral
Fig. 10 and confirm that resonance transmissi- acceleration forces on the damper shoes. The
.J bility depends only on the value of A/K. Trans- latter occurs as the damper passes through its
missibility vs frequency to 500 cps is recorded neutral point, where vibration velocity is max-
in Fig. 11, the input being taken from a typical imum. The shoes, which are approaching each
specification, MIL-E-005272B, Procedure 1. other, suddenly start separating, and the veloc-
Comparison with the computer-derived curves ity change is associated with high horizontal
of Fig. 6 shows some degradation in isolation acceleration. The resulting force depends on
qualities at high frequency. Two factors ap- mass of the shoes, but whatever its magnitude,
pear to be responsible for this: (a) intercoil a vertical component appears on the mounted
resonance of the springs, and (b) coupling of mass because of the damper face. Minimum

183
 > -5-0INPUT
cps .080' D.A.

B 10- I5cps .419

0 _15- 75 cps .036" D.A.
E 75 -500 cps lOg

5 10 20 50 100 200 500 1000 2000 5000

Frequency -CPS

Fig. 11. High-frequency transmissibility of DP damped model

weight elements are, therefore, important in the conventional, constant force damper can be 4
this particular DP damper configuration. relieved, for example, by lost-motion designs
in which the damping is inactive for small dis-
placements. Another technique uses the damper
Other Forms in series with a spring (3). These expedients
are effective if the vibration inputs conform
Displacement-proportioned damping can be closely to that for which they are designed. F

realized in a variety of ways, only one of which Displacement-proportioned damping, on the

has been shown. Figure 12 is another. Here is other hand, is not critical in this respect; it is
a uniaxial or biaxial isolator, in which the sup- predictable in performance and appears to
ported body rests on rollers or balls, spaced merit attention by the designer of vibration
apart by a retainer. Spring restraint on the isolation systems.
retainer cauess it to rub against the balls or
rollers with increasing force as displacement
progresses, obviously a DP damper character- ACKNOWLEDGMENTS
istic.
Sincere thanks are due William T. Polak
In closing, we should note that DP damping and Shankerappa Hulsoor for analog compdter
is not the only form in which dry friction can studies, and to Mrs. William J. Mellow for
be used to advantage. Some of the problems of preparation of the drawings of this report.

REFERENCES
1. G. W. Painter, "Dry-Friction Damped 3. C. M. Harris and C. E. Crede, Shock and
Isolators," Product Engineering (1959) Vibration Handbook, Vol. 2 (McGraw-Hill,
2. J. P. Den Hartog, Mechanical Vibrations, New York), 1961, pp. 30-12 to 30-16
4th ed. (McGraw-Hill, New York), 1956,
~pp. 12-20, 373-377

184

7,0~
 *1

J~

7/ /

Fig. 12. Another form of DP damping -

friction between balls and retainer in-
creases in proportion to displacement

DISCUSSION

Mr. Mahaffey (LTV Astronautics): Were Mr. Wallerstein: We have made a num-
your surfaces metal to metal in this damper? ber of studies, not so much in connection
Mr. Wallerstein: No, they were not. They with this but with another program; we have
had a protective coating. There are a great found a variety of coatings and films which
hvarect icoating tat eranbe
use, de t are pretty good. For example, we have one
variety of coatings that can be used, depending that will operate at 500F for 1000 hours
on the environment in which you are operating. under vibratory conditions and is still pretty

Mr. Mahaffey: Is this friction-damping good. If the conditions are known, a coating
force affected by humidity and various kinds of that will protect the surface can be found.
impurities ?

185

* 2
 'RESPONSE OF RESILIENT MOUNTS UNDER SHOCK LOADING
E. A. Thornton and R. D. Short
Underwater Y-xplosions Research Division
Davi, Taylor Model Basin
Portsmouth, Virginia

Experimental and theoretical studies of the type 5M10,000-H rubber k

resilient mount assembly consisting of a compression and shear mount
and snubber were conducted to determine its performance under shock
loading. In the experimental study, the mount assembly (supporting a
rigid test mass) was installed on the Floating Shock Platform and sub-
jected to a series of five shock tests. In the theoretical study, a sim-
ple mathematical model combining springs and dashpots was used to
represent the response of the mounts under vertical shock loading.
Dynamic force-deflection curves, obtained fromthe experimental and
theoretical data, showed good agreement. The studies demonstrated
that the viscous effects of rubber play a significant role under shock
loading. Thus, the conclusion was that the behavior of resilient mounts
for shipboard installations cannot be based on static tests alone.

INTRODUCTION insight into the performance of the mounts

under dynamic conditions. The information
Resilient mounting systems are used on gained from the evaluation of the test results
modern Naval ships to reduce the structure- on one mount, the 5M10,000-H resilient mount-
borne noise introduced by machinery items. To ing system, showed that the dynamic behavior
meet noise isolation requirements, the mount- of the mounts differed significantly from that'
ing systems are generally made of natural rub- based on static tests. The study of data from
ber, have relatively low static spring constants, these tests indicated the need for controlled
and produce natural frequencies from about 5 to experiments, since the equipment motions ob-
25 cps at rated loads. Since the low mount scured many of the details of the dynamic be-
stiffness permits large excursions during ship havior of the mounts.
maneuvers and shock motions, snubbers are in-
cluded as an integral part of the mounting In this paper, experimental and theoreti-
system. cal investigations of the performance of the
5M10,000-H mounting system under shock load-
The need for understanding the perform- ing will be briefly described. A more compre-
ance of resilient mount behavior under shock hensive coverage of the details of these inves-
loading has become increasingly important as tigations is contained in a recent report (2).
the shock requirements for Naval ships have
increased. Several questions concerning the
performance of resilient mounts under shock EXPERIMENTAL PROCEDURE
loading have arisen. One of the most important
questions concerns the shock mitigation char- Tests of resiliently mounted equipment on
acteristics of the mounts, since the dynamic the FSP had shown the need for a rigid, con-
characteristics of rubber have been found to centrated mass for an accurate study of the
have an important role when subjected to shock mount behavior under shock. To fulfill this re-
loading. quirement, a test'mass was constructed that
consisted of a rectangular steel box filled with
As a part of the Navy's shock-hardening 16,000 lb of lead. The weight was supported on
program, a number of resiliently mounted items the FSP by two 5M10,000-H resilient mounts.
have been shock tested on the Floating Shock Each of the 5M10,000-H mounts consists of:
Platform (FSP) (1). When possible, the David (a) a compression mount to support the major
Taylor Model Basin (DTMB) has studied the portion of the static vertical load, (b) a shear
performance of these mounting systems to gain mount to stabilize the static load with respect

187
 to lateral motion and support part of the verti- constant velocity. These facts are clearly evi-
cal load, and (c) upper and lower snubbers to dent from the typical velocity and acceleration
restrict vertical and lateral excursions under histories of the mass shown in Figs. 2 and 3.
dynamic conditions. Figure 1 is. a view of the A typical relative deflection history of the mass
testfixtureinstalled onthe FSP. The shear is shown in Fig. 4.
_j
. and compression. mounts were installed in a
symmetrical arrangement to maintain static IT
stability; only one of each is visible in Fig. 1. DYNAMIC FORCE-DEFLECTION
Each of the snubbers shown has a clearance CURVES
under static conditions of 5/16 in.
The relationships between the dynamic
The inner bottom of the FSP and.the test forces transmitted by the mounts afid the ac-
mass were instrumented with a number of ve- companying deflections are-of basic interest in
locity meters, accelerometers, and deflection the design of new, more effective mounting sys-
gages to record the shock response. Some of tems. In addition, these results are necessary
these instruments can be seen in Fig. 1. A se- for the dynamic analysis of equipment supported
ries of five shock tests similar to the routine by systems using resilient mounts of the pres-
FSP tests were conducted. ent type. To meet these needs, dynamic force-
deflection curves were derived for the first
compression and expansion of the upper snubber
SHOCK RESPONSE OF TEST MASS of the 5M10,000-H mounts. The force-deflection
curves were found by calculating the instanta-
Instrumentation histories indicated that the neous force transmitted by the mounts from the
shock response of the mass consisted essen- average acceleration histories (Fig. 3) of the
tially of a vertical translation. In each test, the mass. The corresponding mount deflection was
motion of the mass was controlled by the influ- then read from an average deflection history
ence of the snubber engagements. The records (Fig. 4). Force-deflection curves were obtained
show that shear and compression mounts have in this way for each test. Two typical results
negligible influence on dynamic behavior. The are compared with the static force-deflection
engagement of the snubbers is clearly demon- curve in Fig. 5.
strated by periods of sharp acceleration and
deceleration of the mass. During periods of Comparison of the dynamic and static
snubber disengagement, the mass moves with force-deflection curves reveals a number of
I ,

Ip. , Starboard ;
Upper Snubber
'Port Compressia tomporient
Lper Snubber Shear 9A.,R/:

'a!

Fig. I.Mount components in completed test fixture

188
 - 280 -
0. 0 Max. Input Velocity-16fps *

240

. T a v h200 M Input Velocity- 12 fps

0 0 5 7
Fi. Tyiavelocit history 80 "/e /

I0 0
u_
20 as 2TTIC
80
-0 e cc roio. CURVE

-0.
o_0 02 04 06 08
<E1 _IFLECTlON-jnches -

'" Fig. 5. Experimental dynamic F

oforce vs deflection curves for
-20

I0 I5 20 25 30 35 40 45 50 55 60
TIME-msec
' For the range of tests considered, the maximum
Fig. 3. Typical acceleration history of mass force and maximum deflection were found to in-
P 0 crease linearly
0 with
02 increasing
040,0 attaock severity.
This relation is shown in Figs. 6 and 7 where
the maximum forces and maximum deflection
Fwd N experienced by the mounts is plotted versus
08
WFg Typca Aeoe , shock factor.
Fig Exprieticaynmi

Shock factor =--1+sn()

oC J :,:
,o ,/where
-04' '...-.,. W= the charge weight in pounds of TNT,
-0.8
0 20 40 60 8 R = the slant range, and
TIME- msec
Fig. . Tyicalvertcal = the angle between R and the horizontal.
deflection histories The dynamic force-deflection curves also
show a large amount of energy absorption which
indicated that significant damping is present in
significant differences. The static force-
deflection curve is characterized by force in-
creases with increasing deflection as the de-
flection approaches a nearly vertical asymptote.
______

This asymptote is approached at relatively low

___________
the mounts.

THEORETICAL ANALYSIS OF
RESILIENT MOUNT RESPONSE
e

levels of force and essentially establishes that :

the maximum deflection is independent of the The mechanics of deformation of rubber
applied force. For the dynamic force-deflection are significantly different from the linear, elas-
Scurves, on the other hand, the maximum deflec- tic behavior of metallic structures to which
tion and maximum force vary with the input ye- most engineers are accustomed. One of the
locity. Therefore, the force-deflection behavior distinctive characteristics of rubber is the role
is specified by a family of curves which is a of viscous effects in its response to dynamic
function of the loading rate (attack severity), loads. This important role of viscosity has led

189
 300- mounts and one for the snubber. The snubber,
of course, has a clearance 'under static-condi-

C 2oo1-
tions so that it will not engage until the relative
deflection has decreased by, this amount. The
complete mathematical
tical loading is shown inre 0resentation
Fig. 9. for ver-
H
o
0
IL 0
I2 I00.
-= JX
X
I M SNUBBER
.'t CLEARANCED
0 05 I C k C2 SNUBBER
0.05 0.15 0.25 SHEAR a / e kk
SHOCK FACTOR COMPRESSION" 4X6
MOUNTS
Fig.ig6. Maximum force
6.hMaximforce Fig. 9. Mathematical model of
5MI0,000-H resilient mount

Two of the constants, K2 and x4, were found

from static force-deflection curves. The other
0 oconstants were obtained from experimental test
o results. One test was selected for this purpose
and the shock input motion was represented an-
alytically as showp in Fig. 10. The system of
,, Fig. .9 was then programmed on an analog com-
-- o.4 puter and the quantities K , K3, C , and C2 were
" adjusted to obtain the best fit for the measured 5
Xacceleration, velocity, and displacement histo-
ries of the mass. The values found for the con- ,*
0 0 -- stants are Pgs follows: k I = 65 kips/in., k 2 = 10
.o5 015 025 kips/in., k3 = 758 kips/in., k 4= 40 kips/in.,
SHOCK FACTOR cI, = 1.8 kips-sac/in., c2 = 4.71 kips-sac/in.,
Fg7.MxmmM =248 slugs, and D. 5/16 in. The system !
Fig. 7. Maximum '-
deflection vs shock
factor L

LEGEND ,
a number of writers to represent the behavior
of rubber by mathematical models using vari-
ous mechanical systems consisting of combina-
7 Measured Velocity
-v12(I-42t)-2.25 sin 2vt
0.020
I
tions of springs and dashpots. For the present 12.5
study, the model shown in Fig. 8 was selected
to represent the behavior. Other mathemati-
cally equivalent models could have been used to , ,.I-i
give the same results. I Y.ill,,
U 7.5 'I IA '
The 5Ml0,000-H resilient mounts were W 'Wi I
i'll
represented by two parallel systems similar to
Fig. 8, one for the compression and shear 2.51.

I
F

kz CI 250 5 10 15 20 25
k2~CX Fig. 8. Mathe-
of TIME-msec
", redilientmodel
matical
mount
rilet u Fig. 10. Average base
velocity histories

190

,1:
with these constants was then applied to the re- Max. Input Velocity.I6fps
mainder of the tests using analytical represen- 280 ,
tations for the base motion in each case. Theo-
retical-dynamic force-deflection curves were 240 Max.hnput Velocity 12fps
obtained in this way for each test. Two such
curves, together with the experimental data of
Fig. 5, are shown in Fig. 11. In general, the 200 /t
agreement with experimental results was good. /
The maximum error in the worst case was less 0 160 I
,
than 11 percent of the maximum value of the " ,1
variable. W12

CONCLUSIONS 80

Investigations of the behavior of the ,,

40S TC
5M10,000-H resilient mount assembly under
vertical shock loading has provided consider-
able insight into its behavior under dynamic
conditions. Furthermore, it has been demon- 0
strated that the behavior of this mount under 0 02 04 06 08 1.0
shock cannot be based on static tests alone, DEFLECTION-inhes
since the viscous effects of rubber play a sig-
nificant role under shock loading. Because of Fig. 11. Force-deflection
this fact, the dynamic characteristics must be curves for two tests
used in calculation of the response of supported
equipment under shock conditions. The action
of the 5M10,000-H mount under vertical shock by a mathematical model consisting of parallel
load was also found to be adequately described systems of two springs and a dashpot.

REFERENCES

1. J. K. Fleming and R. E. Oliver, "An Explo- 2. Earl A. Thornton, Robert D. Short, Jr., and
sion Shock-Testing Method for Shipboard Ramon R. Walker, "Response of 5M10,000-H
Equipment," Shock, Vibration and Associ- Resilient Mounts Under Shock Loading,"
ated Environments Bull. No. 29, Part I, DTMB Rept. 1899, Jan. 1965
March 1961, pp. 12-18

DISCUSSION

Mr. Levin (Bureau of Ships): I do not think Mr. Short: Right. The point was not that %
that you nor I would want to leave the impres- this is not any good. The response of
sion that the 10,000-lb mount that you showed is these mounts to shock cannot be determined
the only Navy mount that we have. For the sake from the static characteristics, even the very
of illustration, would you refer to one of our small deflection characteristics alone. There
mounts with a lower load rating and show that is no intention to say that this is the only Navy
the purposes of snubbing, compression, and mount, or that all Navy mounts look like this in
shear
mounts,could be taken care of in these other
too ?i' different sizes.
m r. Sho a
Mr. Short: I am not an expert on all types Mr. Hall (Edgerton, Germeshausen & Grier):
.Hvyocnsdrdirmut? ,
of Navy mounts. I do know of a few others which m n -w
are of somewhat different design. There are M.h
some very close to these. I am not familiar
with any if
the gap, that
thishave a continuous
iswhat you haveloading
inmind.without only two mounts.
our program. ThisWeis are
the just
firstgetting
mount.started
There inis
another which is similar to this, except that it
Mr. Levin: There are mounts with suffi- is a conical-shaped rubber mount. We have
cient characteristics to take care of the forces, considered nothing but rubber mounts to date.
the shears, and the compression that you We would be interested in any information you
mentioned. may have on air mounts.
191

4..
 A NEW APPROACH TO PACKAGE CUSHIONING DESIGN
Gordon S. Mustin
Special Projects Consultants, Inc.
Washington, D.C.

Dynamic shock absorption data for 16 different nonlinear cushioning

materials were analyzed. It was found that the shock isolation infor-
mation needed by the designer could be reduced to two curves for each
material, plus a general envelope curve applicable to all of the mate-
rials. This finding, which is consistent with theory, represents an im-
provement over current use of an acceleration vs static stress curve
for each thickness and for each drop height. As a result of this finding,
the remaining challenges of nonlinear isolator technology can now be
attacked. Among the most important of these are variations in isolator
shape, temperature effects, viscoelastic behavior, and configurations
with more than one degree of freedom.

INTRODUCTION 2. When seeking an optimum solution to a

specific problem, the designer must consult a
The problem of using a package cushioning great many curve families. At best, this oper-
material as a shock isolator has generated con- ation is-tedious. At worst, the designer can
siderable interest for a number of years. The never be sure of achieving his goal.
need for logical design techniques has been in-
tensified by the ever-increasing volume of deli- 3. Unless, by chance, the solution falls
cate, costly items which must be shipped with along one of the existing plotted lines, the de-
demonstrable reliability. In a pnrely technical signer must perform at least one interpolation,
sense, package cushioning materials generate the latter largely by intuition.
interest because they are a special case of the
nonlinear isolator with distributed mass and This paper describes a method which will
elasticity. eliminate some of these defects.
None of the existing known theoretical ap-
proaches has proven satisfactory in practice. NOMENCLATURE
On the other hand, the practical expedients
which have been adopted lack clear relation to A Surface area of cushion, sq in.
theory and, worse, do not provide concise en-
gineering design data. a, b Constants defined in Eq. (17)
Figure 1, taken from Humbert and Hanlon c,d Constants defined in Eq. (18)
(1), is typical of the current method of present-
ing cushioning material properties. The popu- D Operator symbol denoting first derivative
larity of using a family of acceleration vs static of function immediately following
stress curves is attested by the appearance of
this approach in the literature (2-5). D"I Operator symbol denoting integral of
function immediately following
It is obvious that Fig. 1 has several draw-
backs for the designer. Among the most perti- G Peak acceleration divided by acceleration
nent are: ofo gravity
ga vido

1. The materiai properties are merely

being sampled. This is true no matter how Go Optimum peak acceleration ratio
many acceleration vs static stress curves may
be drawn. h Height of drop, in.

193
 300

4'

-0j 1\ /

w0wu
o w

0. w

Q01 020 0.507 W 2.0 30 5070100 1

STATIC STRESS (PSI) CL80J 02 03 05Q7 10 20 30 5D070 10
STATIC STRESS (PSI)

70-
THICKNSS 2e 140 :TT
60 - 1 1i 111,11
HCKES1"
3 1
- *..120T f

uj~ 40 8 -.

10-
01 Q2 03 0507 10 2.030 50 70 100 4~~PIH
STATIC STRESS (PSI) 2
06801 0.2 0.3 05Q07 10 20D30 5070 90
STATIC STRESS (PSI)
70 ICKNES 2" 3 4.

60 - 140

4 ~120
w14-- 3
w 6100
30-

20 - 80

24 KI~DROP
HEGHT 111160 -

Q801 02 0.3 05Q7 10 20 30 50 7010.0 0-

STATIC STRESS (PSI) 480~DOPHIH

20
4CosOj 0203 050Q7 I0 20130 507DI00
13 111, DOP HIGHTSTATIC STRESS (PSI)
IN.DRO HEIG
3610 1 11T

0 THICKNESS
2"
0

w 804
0

20.1 02 03 0507 10 20 03D5070100D

STATIC STRESS (PSI)

Fig. 1. Dynamic data on polyethylene foam (1)

194
 j

j Ratio of peak stress developed to energy

r per unit volume absorbed by a cushioning D - f( c,p)] de. (3)
material

J, Optimum value of j Let the cushioning material be cut into a

flat pad with surface area A and thickness T..
T Cushion thickness, in. Support on this pad a weight, w, in such a fash-
ion that the static stress as (as= W/A) is uni-
W Weight of cushioned article, lb form. Drop the assembly flatwise onto a rigid
surface from the height h.
OE Strain, in. deflection/in, thickness
Following the line originally taken by
60 Optimum strain Mindlin (6), while ignoring air resistance and
restricting attention to peak values, we have
SCoefficient of friction
D-
D'lm =a + C) (4)
p Set of parameters affecting dynamic s T(
stress-strain relationship
It is evident that
a Stress, psi
% as G, (5)
o-. Maximum stress developed in any partic-
ular drop where G is peak acceleration divided by the ac-
celeration of gravity. Substituting this relation
so Optimum stress into Eq. (4) and rearranging terms, we have

f a Static stress G0 -1 (6)

as Optimum static stress
where the symbol J, first used by Janssen (7),
is the ratio of peak stress developed to the en-
BACKGROUND ergy per unit volume absorbed by the material
at that stress; i.e., J = O/D'lOm.
Let the dynamic stress-strain character-
istics of a cushioning material be represented Normally, h is considerably larger than T
by a continuous function and e is everywhere less than 1.0 by definition.
IOn this basis, we may write, with good accu-
a = f(e,p) 0 C < 1.0, (1) racy,

where;.G =J .' (7)

a = stress, force per unit area;

Obviously, for fixed G and h, T is mini-
=strain, in. deflection/in. thickness; mized as j is minimized. The set j can be
minimized either when the set has minima
p set of parameters, as yet undefined, when c reaches its maximum set of values or
which affect the stress-strain rela- when there is some other distinct set of minima.
tionship; and These latter minima exist provided there be
real, nontrivial, solutions in F- to
f( )=function of the statements contained
in) Da.(D"%a) -2 0. (8)
Since the function is continuous within lim-.
its, both the first derivative and the integral Equation (8) is the first derivative of J(DJ) set
exist, or equal to zero and simplified. For j to be a
minimum, it is also necessary for the second
Da = f(c,p)] (2) derivative to be positive. These conditions
(2) being met, there is a nonempty set, jT,

195

__________ ____________________________________1-~
 which gives minimum thickness for each value 2. Flat, uniform cushion, laterally confined
assumed by the parameter p. or possessing negligible Poisson effect; and

Yurenka and Giacobine (8) investigated 3. Flat, uniform cushion, and lubricated
.1 tangent, algebraic and logarithmic elasticities surfaces.
with p = 1.0. The value of J. for tangent elas-
ticity duplicated the result previously found by Almost all previous work on cushioning
Mindlin (6). material properties has been concentrated on
plane cushions meeting either criterion 2 or 3.
If 6 exists, Eq. (1) states that there cor- In these cases, the coefficient of friction does
responds a set of optimum dynamic stresses not enter and Eq. (10) may be written
%. Hence, from Eq. (5), there also exists a
set of optimum static stresses. The optimum hh (1
static stress a, is determined from Eq. (4), and (11)
may be expressed with good accuracy as
01 X D1 (9)It is clear, then, that we may consider
o' = - . (9) f(h/T) to be equivalent to f(p).

All problems in package cushion design, or Plots of J vs energy per unit volume for a
the more general problem of nonlinear shock selected number of values of h1/T will result in
isolator design, may be reduced to thB essen- considerable reduction in the total number of
tials of solving Eq. (6) or (7) for minimum curves required. In view of the definition of J
thickness and Eq. (9) for corresponding opti- given in Eq. (7), however, required thickness
mum static stress. In actual practice, however, would appear as a component of the ordinate, of
this is not easy to do. One of the main reasons the abscissa, and of each curve in the family,
for this difficulty is that p is rarely, if ever, As a result, the designer would have to approach
equal to 1.0 or to any other constant. As a re- the solution by successive approximations, and
sult, J. is not a unique number but, rather, a the improvement, from his standpoint, does not
set of numbers. The practical difficulty of de- meet the simplicity criterion, even though re-
termining p, and how the dynamic stress-strain lationship of the plots to theory is better than
relationship is affected thereby, may be consid- heretofore.
ered one of the prime reasons for lack of
progress since the use of G- a, curves was
introduced in 1957 (9). A SOLUTION FOR DESIGNERS

In 1961 and 1962, however, Soper and Dove By definition, the optimum point on a j
(10,11) showed that one could write curve is the minimum value assumed by j.
This point has an optimum energy per unit
volume. Hence, for each value of 1VT, dividing
Jf\ h TT'
h (10) actual J by J0 will result in V.

J G
where u is the coefficient of friction. This T.- G = fs a constant. (12)
special scaling law is considered valid, pro- - o h/T
vided the mass of the cushioning is much less
than th6 mass of the cushioned article, the peak On the basis of Eq. (12), 16 different mate-
acceleration is much greater than the accelera- rials were analyzed. Data were taken from the
tion of gravity, and the local stress states in literature (2-5). Even though different sources
the cushion are dependent only on time, the an- were used, the data were combined unless it
gle at which the stress is applied, and the spe- was apparent that they stemmed from a com-
cific variables indicated in Eq. (10). These mon source. 'the envelope of results is shown
investigators then listed the following nonexclu- in Fig. 2.
sive configurations which appear to satisfy the
last condition: Most of the materials developed, effectively,
a single curve independent of the parameter
1. Material thickness everywhere very h/T, but the rubberized hairs did not. While
small relative to the dimensions of the cush- an interesting topic for investigation, this area
ioned component, and no slippage at one or both was not pursued in the face of the finding that
material surfaces; all curves practically coincide in the stress
19

196
 APPARENT LIMIT OF EMPIRICAL DATA
5.0- LOGARITHMIC ELASTICITY
, TANGENT ELASTICITY I

4.0
ALGEBRAIC ELASTICITY

12 1--
2.0- - -
-,

0 0.2 0.4 0.6 0.8 1.0 1.2 IA 1.6 1.8 20 2.2 2.4 3.0

Fig. Z. Envelope of acceleration stress curves

I ratio region 0.4 to 1.5 and are flat between 0.7

and 1.3.
~ing
Because of Eq. (5), we now have a' on both
sides. Hence, we divide through by os, arriv-
at
In effect, then, Fig. 2 demonstrates that
wide static stress variation from optimum can Go= f(--!-., (14)
occur without significant effect on peak accel-
eration. Some time ago, it was shown (12) that,
for a given thickness and area, wide variation Substitute for J its equivalent taken from Eq.
t in material stiffness was possible. The two (7). Dividing both sides by energy per unit
findings, taken together, indicate that the de- volume, we have
signer has wide latitude both in selecting mate-
2
rials and in detailing the design. As a mini- GoT
mum, Fig. 2 may be taken as establishing - f (15)
reasonably wide limits on the accuracy with a' h2 (T)

igiven
which G. and ' must be determined for any
material which may be further simplified and written as
Figure 2 is only usable ifvalues for G. and
al can be found. We turn, therefore, to solving s= f T (
I this aspect of the problem. Both solutions are
found with the aid of the Soper and Dove rela- Logically then, we proceed to seeking em-
tidon, Eq. (11). pirical relations giving substance to Eqs. (15)
From the definition of j given in Eq. (6), and (16).
we have in Eq. (11) the energy per unit volume Using the principle of least squares, satis-
absorbed by the material as the numerator on factory fits of the data for 16 materials can be
the left-hand side and the energy per unit vol- found by writing Eq. (14) explicitly as
ume imposed on the material on the right-hand
side. It appears reasonable, therefore, that we G 1)
might write T+b. (17)
( ) (13) Values of the constants a and b are given in
0 - T, Table 1. This table also contains computed

197
 TABLE 1
Solutions to the Acceleration Equation

Mate- No. of Standard Com- Max. Corr. Com- Min.

rial Data Dev i puted Y2 puted t
No. Points a,2 (p = 0.99)a Coeff. (t) (p = 0.99)b

1 16 3.65 -2.53 2.50 3.620 5.229 0.9853 22.912 2.977

2 19 2.73 2.91 2.14 4.440 7.012 0.9512 14.318 2.898
3 27 2.85 2.71 2.75 7.321 12.198 0.9650 20.025 2.787
4 22 3.98 3.46 1.50 1.424 8.897 0.9983 76.646 2.845
5 26 3.13 -4.61 2.32 10.069 11.524 0.9923 40.749 2.797

6 16 4.67 -9.43 2.58 2.551 5.229 0.9904 28.105 2.977

7 16 2.79 3.74 1.67 1.464 5.229 0.9887 25.941 2.977
8 16 3.72 6.83 1.18 0.600 5.229 0.9968 47.781 2.977
9 28 3.64 0.44 3.50 9.267 12.879 0.9726 22.910 2.779
10 28 3.95 0.25 0.87 0.826 12.879 0.9992 126.823 2.779

11 16 2.37 4.31 2.18 0.310 5.229 0.9813 19.077 2.977

12 8 3.10 13.14 2.96 2.006 2.167c 0.9456 8.475 3.707
13 12 3.25 -4.02 2.70 2.220 3.053 0.9788 16.434 3.169
14 7 3.12 6.26 1.78 1.279 1.635c 0.9880 15.477 4.032
15 15 4.88 -0.35 1.31 1.123 4.660 0.9965 44.131 3.012

16 21 4.51 1.20 3.52 5.719 8.260 0.9916 34.673 2.861

17 20 4.52 0.68 2.00 2.080 7.633 0.9956 46.067 2.878
18 56 4.54 1.03 2.62 9.536 27.012 0.9934 65.221 2.67
19 78 4.26 2.47 3.11 16.909 43.054 0.9915 68.036 2.64
aFrom R. A. Fisher, "Statistical Mcthods for Research Workers."
bFrom G. W. Snedecor, "Calculation and Interpretation of Analysis of Variance and Co-Variance,"
Collegiate Press, Ames, Iowa, 1934.
CThis value of x 2 is for P = 0.95.

values of the standard deviation of the regres- expression for the variation of the elements of
sion line, the Y-squared statistic, and a com- J. Dividingbothsidesby h/T we have J, equal
puted value of t for use in Student's t-test. tc a constant plus b divided by h/T. Where b is
The fit to the available data is sufficiently reli- small, J. may be considered a constant for all
able for design purposes. practical purposes, even at the low values of
h/T used here. In any case, tendency for J0 to
Note that 19 materials are identified in become constant is confirmed by these results.
Table 1. Materials 15, 16 and 17 are three dif-
ferent densities of polystyrene foam. Forest Turning now to relation Eq. (16), the same
Products Laboratory data (4) concerning the 16 materials provide a satisfactory least squares
same materials were combined as a single ma- fit to the explicit equation
terial and are shown in line 4 of the table. To
satisfy curiosity, data for materials 15, 16 and (G h d
(
17 are combined in 18 and all of 18 and 4 are Sc (18)
combined in 19. Although the various regres-
sion lines do differ, no estimate of the signifi- Values of the constants c and d are shown in
cance of these differences is attempted here. Table 2, which also contains the standard devi-
ation and the y,-squared statistic. Previous
Working primarily with high energy ab- remarks concerning materials 4 and 15 through
sorptive materials, Soper and Dove (10) found 19 apply.
that J. tended to become constant and postu-
lated this tendency as a characteristic of many Equations (17) and (18) are plotted for ma-
cushioning materials. Equation (17) is an terial 10 in Fig. 3.

198
 TABLE 2
Solutions to the Static Stress Equation

Mate- No. of Corn- Max.

rial Data c d Standard puted 2
No. Points_,_ Deviation pte (p= 0.99)a
_ No. Points

1 16 0.80 -0.23 0.025 0.035 5.229

'2 19 0.35 -0.41 0.016 0.065 7.012
3 27 0.21 -0.26 0.017 0.106 12.198
4 22 6.81 -0.40 0.070 0.134 8.897
5 26 0.37 -0.13 0.052 0.297 11.524

6 16 0.41 -0.12 0.038 0.097 5.229

7 16 0.11 -0.18 0.009 0.028 5.229
8 16 0.20 -0.23 0.017 0.072 5.229
9 28 0.72 -0.43 0.011 0.032 12.879
10 28 11.86 -0.48 0.190 0.754 12.879

11 16 0.091 -0.15 0.009 0.030 5.229

12 8 0.069 -0.18 0.004 0.005 1.239
13 12 1.00 -0.23 0.022 0.019 3.053
14 7 0.31 -0.36 0.016 0.023 0.872
15 15 3.54 -0.31 0.101 0.157 4.660

16 21 4.76 -0.32 0.196 1.088 8.260

17 20 3.72 -0.29 0.137 0.417 7.633
18 56 3.68 -0.29 0.164 1.843 27.012
19 78 4.36 -0.32 0.156 2.240 43.054

aFrom Fisher's tables.

DESIGN EXAMPLE IMPLICATIONS

Suppose that a 50-lb object is to be dropped The results obtained are essentially em-
24 in. with a maximum allowable acceleration pirical and, therefore, generalizing could be
of 30 g. Entering the graph with G, = 30, we questionable. Nevertheless, it seems clearly
find h/T is 7.5. Thickness for a 24-in. drop established that three simple curves can be
height is 3.2 in. The corresponding value of substituted for an infinite continuum of curves,
Goh/T is 225 for which the indicated optimum inadequately, and redundantly, sampled by some
static stress is 0.91 psi. From this the re- 28 curves as exemplified by Fig. 1.
quired area is 55 sq in.
As a result of this conclusion, it appears
Because meeting a requirement for exactly that the available handbooks and design data
3.2 in. of thickness may prove awkward in prac- (2-4) should be revised to reflect the results
tice, several other thickness options are 6x- obtained here. Further, the basic test method
amined in Table 3. Note that a surface area of specification (13) should be revised to the ex-
50 sq in. and a thickness of 3 in. results in a tent of. changing the method of reporting the
maximum acceleration of 32 g. Given the usual results obtained.
uncertainty concerning maximum acceleration
permissible in package design situations, and The military specification for cushioning
making due allowance for shock mitigating ef- materials (14) should be revised to abandon an
fects of the outer container, this combination awkward classification of materials based on
provides minimum cushion volume for the par- performance of a limited number of thicknesses
ticular conditions with this material, at a single drop height.

Note that, in Fig. 3, the variable Goh/T With the foregoing suggested changes in
must be formed from the first regression line. existing approaches out of the way, let us turn
Attempts should not be made to work backwards to some of the more obvious limitations of the
in the static stress line. approach taken in this paper. These limitations

199
 TABLE 3
Variations in Cushion Thickness and Areaa
Parameter Calculated J _ _-Value

A. Original computations:

1. Cushion thickness, in. 3.00 3.20 3.25 3.50 7,

2. h/T for 24-in. drop 8.0 7.50 7.38 6.86
3. Indicated G. for step 2 hIT 32 30 29 27
4. Steps 2x3 to give Goh/T 256 225 214 185
5. a, corresponding to step 4 value 0.89 0.91 0.93 1.00
6. Area required for 50-lb wt. sq in. 56.2 55.0 53.7 50.0
B. Effect of setting area at 60 sq in.:

1. Static stress ratio 1.07 1.09 1.12 1.20

2. Probable acceleration, from Fig. 2 32 30 29 27.5

C. Effect of setting area at 50 sq in.:

1. Static stress ratio 0.89 0.91 0.93 1.00

2. Probable acceleration 32 30 29 27

D. Comparative cushion volumes:

1. Volume from A 169 176 174 175

2. Volume from B 180 192 195 210
3. Volume from C 150 160 162 175

aMaterial No. 10, 24-in.drop, 50-lb weight, 3

0g design goal.

are relatively few in numbez but they are sub- restated as.an intuitive disquiet resulting from
stantive: ignoring the theory of models when there is in-
1 e ae
fsufficient experimental evidence that it is safe
1. Secondary data, processed for publica- to do so with these materials.
tion, were used for the curve matches found
here. Future work should be done with original 4. Each sample analyzed was a flat cush-
data now that an approach is indicated. In any ion in the shape of a rectangular parallelepiped.
case, however, reduction in the total quantity of Other shapes should be investigated. Soper and
data required to describe a material',for pre- Dove (10,16) have shown that numerical shape
liminary design purposes should be possible. factors can be found, at least for hemispherical
systems. For one particular material, not ana-
2. Most of the data were obtained using lyzed in this paper, this factor was found to be
constant surface area cushions in which the 0.81.
thickness was varied. While acceptable matches 5. These results are limited to single de-
were found, even with 6-in. thick materials in
multiple
the sample, there is no guarantee that such gree of freedom
degrees of freedom, such asObviously,
systems. are found i.n
rota-
thicknesses do not represent a significant tional drop tests, introducd new problems which
variable. inldotetitouenwpolmwhc
remain to be investigated.

3. Many of the materials are open-celled 6. All of these data are for room tempera-
plastics. Hence, ability of air to escape might ture conditions. It is well known that most
very well affect the results obtained with a elastomeric materials are sensitive to wide
specific design. The previous limitation is temperature changes. Temperature variation
thus partially restated here. Also open to data usable in package design are limited, al-
question is what will result When the sides of though the curves developed by Schuler (17)
the cushion are partially or totally sealed, a have merit. Useful data should be more readily
problem explored in preliminary fashion by attainable by concentrating on the optimum
Gigliotti (15). Objections 2 and 3 can be points of the J curves.

200

- - ~ ~ - - ~ - - - - - - ~~ .. ' ~
 GOhIT
I00.0 20 30 40 506070 90100 200 300 400 600 800 000 2000
10.0- 111
I I I I I 'I I ' I III I --9.0
80.0- 8.0
70.0-
60.0- T- 6.0
50.0- 700-GO= 3.95(-L~)
o-=I11.86 Go+25 4 STDSTD -. 87 .19.
DEVDEV= CORr, C0EFF z.999 7.0
- 5.0

40.0- - 4.0
30.0 Govs h T "3.0

20.0 2.0

Go / o"'

10.0-a vs Go 1.0
9.0 I -0.9
8.0- - 0.8
7.0- 0.7
6.0 - 0.6
5.0- -0.5
4.0- - 0.4

3.0 0
-03

2.0 0.2

jI ,II IIIIII , I II lilIIII I I

2 3 4 5 6 78910 20 30 40 5060 80 100 200
h/T

Fig. 3. Design curves for material 10

7. All of the materials exhibit time sensi- All of the foregoing constitute gaps which
tivity in their dynamic stress-strain relation- remain in our knowledge of the behavior of
ships. While the use of empirical curves is a nonlinear isolators. It is hoped that the sim-
helpful way around this dilemma, such usage plified design technique developed in this paper
does not give a clear picture of total perform- will permit investigators to accept the chal-
ance sufficient to predict behavior beyond the lenges posed by these gaps.
limits of the experimental data. Since these
materials are also temperature sensitive, it
would appear necessary to determine their
viscoelastic behavior. Burgess (18) has shown ACKNOWLEDGMENT
how the complex dynamic modulus can be con-
verted to mechanical impedance and the latter Preliminary hand computations demon-
techniques used for response prediction. At strating feasibility of the approach taken were
present, this would seem the direction that performed by my son, George M. Mustin, dur-
package cushioning research should take. ing a summer vacation.

REFERENCES

1. W. E. Humbert and R. G. Hanlon, Package 3. M. E. Gigliotti, "Design Criteria for Plas-

Engineering, April 1962 tic Package 'Cushioning Materials," Plastic
2. P. E. Franklin and M. T. Hatae, Ch. 41, Rept. No. 4, Dec. 1961
Shock and Vibration Handbook (C. M.
Harris and C. E. Crede, Eds.) (McGraw- 4. MIL-HDBK-304, Package Cushioning De-
Hill, New York), 1961 sign Handbook, 25 Nov. 1964

201
 5. R. K. Stern, Modern Packaging (Dec. 1959) 11. W. G. Soper and R. C. Dove, J. Appl. Mech.,
263 (June 1962)
6. R. D. Mindlin, Bell System Tech. J., 24:
353 (July-Oct. 1945) 12. G. S. Mustin, Shock and Vibration Bull. No.
28, Pt. 4, p. 79, 1960
7. R. R. Janssen, "A Method for the Proper 2
Selection of a Package Cushion Material 13. ASTM Method D 1372-55, 1955 7
and Its Dimensions," North American Avi- 1
ation Rept. NA-51-1004, 1952 14. Military Specification MIL-C-26861

8. S. Yurenka and C. R. Giacobine, "The Se- 15. M. E. Gigliotti, Modern Packaging (Sept.
lection of Package Dimensions Based on 1960)
Optimum Cushion Factors," Douglas Air-
craft Co. Testing Div. Rept. No. Dev-2128, 16. W. G. Soper and R. C. Dove, Shock and
July1956Vibration Bull. No. 28, Pt. 4, p. 65, 1960

9. 0. S. Kerstner, "General Principles of

Package Design; Part I- Cushioning," 17. S. C. Schuler, "Recent British Develop-
Northrop Aircraft Rept. No. NAI-57-187, ments in Package Cushioning and the De-
Feb. 1957 sign of an Impact Testing Machine and a
Dynamic Tester," Royal Radar Establish-
10. W. G. Soper and R. C. Dove, "Data Pres- ment Memo No. 1712, 1960
entation for Cushioning Materials," Paper
No. 27-4, 17th Annual Tech. Conf., Soc. of 18. J. C. Burgess, WADD Tech. Rept. 60-671,
Plastics Engrs., Jan. 1961 Oct. 1960

DISCUSSION
Mr. Zell (Picatinny Arsenal): I have had a to digest the voluminous data admittedly pres-[
peripheral acquaintanceship with the develop- ent, we should give some judicious consideration
ment of some of the problems that Gordon has to finding rationalized parameters that have
indicated. I feel that there was a certain po- immediate application to the problem. For
lemical air in the paper, which I think is unfor- example, the cushion factor or the shape factor
tunate. In some of the approaches used, the could be inverted and characterized as effi-
people who were involved were generally well ciency. This has direct significance to the en-
aware of the many existing limitations, but felt gineer. I have not had time to digest this entire
that these limitations should have been tackled paper, but I am not convinced that the parame-
to provide some order where there was chaos. ters which would seem to provide the best ra-
I do not feel that the work of Forest Products tionalization are necessarily the ones that we
Laboratory in assembling the Packaging Hand- should put foremost into the minds of the people
book should be denigrated. Although the ration- using these materials. I think this bears some
alization of the data presented here is an ex- discussion.
tremely valuable approach toward simplification
of the problem, I think that we should be careful Mr. Mustin: In any investigation one usu- j
in the parameters which we use to rationalize, ally finds himself with a set of numbers. After
The validity of any basic design approach is these data have been run through the computer
strengthened when the parameters are those and the results analyzed, it may appear that it
which bear immediate relationship to the prob- could have been done better in another way.
lem at hand. I think that nondimensional param- Now the first curve, g vs h/t, is actually, and
eters should reflect something that is intrinsic is so identified in the written paper, a plot of
to the nature of the material itself. The factor the variation of the elements of the cushion
j, that was mentioned, which has been known in factor, or Janson's j. In deriving the second
the development of the literature as cushion curve, a.s as a function of optimum acceleration
factor, is actually a shape factor. Thus, it is times hit, I was at liberty, but did not realize
basically a general characterization of the na-
ture of the stress-strain material, whether the
j is static or dynamic. I think that this is one
parameter that should appear in the basic pres-
entation. Before we go off half-cocked and try
it at the time, to have plotted it as a function of
h/gt, which is j - 1, or the efficiency factor that
you mentioned. I still stand on the proposition
that we are able to substitute two straight lines
for a multiplicity of lines.
4
202
 t

OPTIMAL SHOCK ISOLATION SYNTHESIS \-

T. Liber and E. Sevin

Iit Chicago,
ResearchIllinois
Institute

t
The problem of optimum shock isolator synthesis, in the sense of
minimizing the relative isolator displacement for a prescribed level of
input acceleration attenuation, has been formulated and solved for
single degree of freedom systems. The methods of dynamic program-
ming and linear programming are found applicable to a discrete ver-
sion of the problem; the latter technique was more efficient. The opti-
mal isolator is active, possessing the characteristics of "bang-bang"
control. Results of optimal isolator performance are obtained for four
types of input shock wave forms. These results, expressed as a rela-
tionship between limiting'values of relative isolator displacement and
mass acceleration, permit direct determination of the margin'for im-
provement existing between the optimal and any specified candidate
isolator design. The dependence of the optimal isolator performance
on the detail of the isolator function and the input shock wave form is

discussed. t

INTRODUCTION improvement is possible, he almost never

knows whether the margin for improvement I r
Generally speaking, the function of shock makes further consideration practically
isolation is to attenuate the motion environment worthwhile.
transmitted from a base structure to its inte- t
rior components. Often the principal design The path to improved design methodology
requirement is to reduce peak transmitted ac- would appear to lie in the direction of general C
celerations to tolerable levels; this is accom- synthesis techniques, whose power have been
plished by some form of relatively low- amply demonstrated in control theory and cer- t
frequency mechanical system interposed be- tain structural design problems. The practi- f
tween the base structure and the interior cality of such an approach for shock isolator
component. Besides achieving a specified level design is the subject of this paper. While the
of acceleration attenuation, the designer also discussion henceforth could proceed along t
may be required to limit relative velocity and strictly mathematical lines, we prefer to estab-
displacement and to accomplish all this at low lish a setting within the context of a highly
cost and high reliability. Obvious counterparts practical design situation. This will be the de-
to this problem exist in the design of elec- sign of shock isolation systems applicable to ]
tronic, pneumatic, and hydraulic systems protective construction, a problem area of in-
subject to. severe external excitations. creasing importance in the development of
hardened military facilities, such as under-
While the designer may seek to achieve an ground missile launch sites' and command and I
isolator whose performance is optimum within control centers. Our results, of course, are by
a set of imposed constraints (few of which may no means limited to these applications.
be expressed in quantitative terms), practical
considerations often serve to restrict his atten- Shock isolation systems for underground
tion to a very limited number of candidate iso- facility application usually must satisfy two
lator concepts. This being the case, it is un- main operational requirements, aside from cost
likely that his final solution will be optimum in and reliability. Firstly, the interior components
any general sense. Perhaps of greater impor- to be isolated from the effects of the blast-
tance is the fact that the designer remains ig- induced shock environment incident on the con-
norant of the extent to which the performance taining structure are mostly acceleration sensi-
of his system approximates that of the optimum. tive, so high levels of acceleration attenuation
Thus, while he can generally assume that must be achieved at any expense. This inevitably

203

__________71________________I
leads to low-frequency isolation systems, which interposed between the mass (element to be
produce comparatively large displacements of isolated) and the movable base which is subject
the isolated component relative to its containing to the prescribed shock excitation y(t). The
structure. Often the interior dimensions of the displacement of the mass is denoted by z,
structure (e.g., missile silo) are controlled by measured relative to a fixed frame of refer-
the "rattlespace" requirement for shock- ence. The displacement of the mass relative to
isolated components. As the overall facility the base is denoted by x, where x = z - y.
cost generally is a rather sensitive function of
structure volume, the practical desire to reduce The equation of motion for unit mass is
rattlespace is clear. Therefore, reduction of
the rattlespace requirement is the second prin- i + g = 0. (1)
cipal design consideration.
The initial conditions
will be taken as*
In this context the general isolator design
problem can be expressed in the following z(0) = i(o) = 0. (2)
manner: "Out of the totality of possible isolator
concepts, select one which (for a given input) The kinematic conditions are
achieves a specified level of acceleration at-
tenuation and provides the minimum rattle- Z X+ y,
space."
(3)
This would appear to be the most simple
and direct formulation of the design problem z= + j.
and, frankly, it has puzzled us for some time
that the designer does not instinctively think in We will require that the input acceleration ;(t)
such terms. There is, of course, no doubt that and input displacement y(t) be bounded in the
he traditionally does not approach the design time interval of interest. (It is not necessary
problem in this way, and the mere fact that he that
may not be able to discern the solution seems sincey(t)
the be bounded
base beyond the time
could subsequently of interest,
undergo a
a wholly inadequate excuse; if he cannot, there uniform velocity.) While it does not affect the
are many who can. In this paper we develop formulation of the problem, we will subse-
the solution to precisely this problem for single quently restrict attention to inputs which ulti-
degree of freedom systems and compare the mately return the base to rest; i.e., either S
optimum performance with that of conventional y(t) = 0 for all t >T,; T finite, or 9(t) = 0 as
isolator designs. It should come as no surprise t -,Co.
that there can be considerable room for per-
formance improvement over conventional de- The above equations may be considered a
signs. Insofar as the solution techniques are nondimensional formulation, where the dimen-
concerned, they do not appear to be limited to sional quantities are as follows:
the kinematically simple systems considered,
and we would look to numerous developments z =/D o , x =/D, y = /D;
along these lines.
PROBLEM FORMULATION t= /mD
Equation of Motion

The single degree of freedom system con- *The analysis easily can be extended to the
sidered is shown in Fig. la. The "isolator" g is more general case of !(o) constant.

MOVING FIXED
BASE BASE
ISOLATOR MASS ABSORBER MASS Fig. 1. Single degree of free-
dom shock isolator and absorber
9 M m f (t) model

(a) SHOCK' ISOLATOR (b) SHOCK ABSORBER

204
and D yields a g- and min-max I = A, then the
solution to Eq. (4b) with A prescribed and the
g =/ko. same g* will yield min-max lxI = D. There-
fore, we may use A and D to denote either
In these relations, the "barred" quantities are bounds on the prescribed mass acceleration or
of physical dimension; m is the mass, and D. isolator displacement, or their respective min-
and k. are the scale factors for the input shock max values associated with g-.
and the isolator function, respectively. The
dimensional forms of Eqs. (1), (2), and (3) are The desired function g* will be referred to
as the "optimum isolator" for the prescribed
mi" + kog = 0, input since its response as characterized by
maximum mass acceleration and relative dis-
F(O) '(o) 0 placement (rattlespace) is the best possible.
and The relationship thereby established between
.N + ,the lower bounds on relative displacement D
and absolute acceleration A will be referred to
where the primes denote differentiation with as the "optimal (isolator) performance" for the
respect to real time i. The nondimensional prescribed input.
notation will be used throughout the paper.
Characterization of Optimal Isolator
Definition of Optimum Isolator We have not yet indicated a specific func-
We consider that the optimal isolator is tional dependence of the optimal isolator, and
the isolator function g having the property that, there would appear to be several choices avail-

for a prescribed shock input y(t), either (a) able. Itmight seem most natural to seek a
the maximum relative displacement of the mass representation of g in terms of the (relative)
does not exceed a prescribed bound and the state variables x and x; that is, g = g(x,x).
maximum absolute acceleration of the mass is Another choice might be to introduce a set of
minimized, or (b) the maximum absolute accel- control variables u1 , or inspection of Eq. (1)
eration does not exceed a prescribed bound and might suggest seeking g as an explicit function
the maximum relative displacement is mini- of time; i.e., g g(t).

mized. In more precise terms, we desire to It is easily shown that an a priori assump-
find the function g such that either tion of the form g = g(x,x) does not lead to a
max lxi D , a constant well set problem; that is, a unique solution for
tt] the optimum isolator possessing this form can-,
and (4a) not be obtained. This result may be demon-
max =a minimum, strated either through an analytical argument o
mt) within the context of the calculus of variations
(2), or by means of the following heuristic

or argument.

max 121 < A, a constant , Assume that the desired optimum isolator
It) has been found in the form g = g*(x, k). This
and (4b) may be represented by a surface in the (x,i.g)
max lxi a minimum, space as shown in Fig. 2. Associated with g",
fIt) the optimal response is given by x = x*(t) and
= .*(t), which corresponds to some curve c,
In each case, z is the solution of Eqs. (1) and in the (x,) plane (phase plane) of Fig. 2, as
(2), x is given by Eq. (3), and the input function shown. Now g* may be thought of as an explicit
y(t) is prescribed in the interval 0 < t < T. In function of time since
addition, we require all velocities and displace-
ments to be continuous. The formulation is in- g*(x,,) = g* [x*(t), ,*(t)] G*(t).
tended to admit bounded discontinuities in the
accelerations. The function G*(t) represents a'curve in the
(x,k,g) space which is the intersection of a
It can be shown (1,2) that these two formu- right cylinder erected on C with the g* surface.
lations are reciprocal in the sense that either But as nothing has been said concerning the ex-
leads to the same min-max values. In other plicit functional form of g(x, k), there may be
words, if the solution to Eq. (4a) with prescribed many such surfacescontaining G*(t); e.g.,G(t)

205 .

I
-A_
/'

ig is that of the shock absorber depicted in Fig. lb.

Here the absorber h is interposed between the
mass and a fixed base. The mass is subject to
a prescribed force, and the performance re-
GM quirements generally relate to the mass accel-
em
6() eration 1, the transmitted force h and/or the
g (xi) mass displacement z. An analogous formula-
777 *. -tion to that of Eq. (4) then results from finding
the absorber function h, which has the property
that, for a prescribed force input f(t),

max h < H , a constant,

and ax=z
max a minimum,
Fig. 2. Characterization of
optirhum isolator [t]
where z is the solution of mi+ h = f(t) and
z(0) = i(o) = 0 in a prescribed interval of
as shown in Fig. 2. Hence, there cannot be a time; z and i must be continuous and i must
unique expression for g*. It will be noted that be f-bounded. Of course, other formulations
this argument does not really depend on the are possible, such as those imposing bounds on
specific form of the optimization problem as displacement and velocity.
given by Eqs. (4a) or (4b). The situation is dif-
ferent of course as one further constrains the The solution to the problem stated is con-
class of isolator functions under consideration, siderably more straightforward than the iso-
such as by introducing a set of control variables lator problem and is reported in Ref. (2). We
or by selecting an explicit functional form for will not concern ourselves here further with
g. However, it does not suit our purposes to do the optimum shock absorber problem.
so at this time. As a result we must be content
to seek the optimal isolator as an explicit func-
tion of time, which we will continue to call Limiting Situations
g*(t). The implications of this from the point
of view of practical isolator design will be con- The problem formulated in Eqs. (4a) or
sidered in a later section. (4b) admit of two obvious limiting situations.
Consider a prescribed input y(t) and let
The question remains as to the uniqueness
of g*(t). It is clear that for times much beyond max iyI =
Dy
that at which the min-max values are achieved, It]
g*(t) might not be unique, since any form of and
g(t) is acceptable which subsequently does not max ijj = A
violate the constraints of the problem. Indeed, It
it is easily demonstrated that a unique g-(t)
does not exist over a general time interval of Then, with reference to Eq. (4a), if we permit
interest, T. However, all g*(t) which yield the D = D , min-max I I = o which is achieved
same pair of min-max values, A and D, possess with g* = 0, i.e., an "infinitely soft" isolator, or
unique characteristics (e.g., constant force in- a constant force spring if gravity effects are to
crements) over certain subintervals of T. The be included in Eq. (1). On the other hand, with
location and length of these subintervals depend reference to Eq. (4b), if A = Ay,. min-max
on A, or D, the input function y(t), and the initial 0e
conditions on z(t). Specific examples of this lxi = 0, which implies g* =, i.e., a rigid
isolator." Thus, as is perfectly obvious, the
"quasi-unique" character of g*(t) will be given required rattlespace approaches the maximum
in the section on results, input displacement as the level of acceleration
attenuation increases, and, conversely, the
Optimum Shock Absorber, transmitted acceleration approaches the maxi-
Optinu Sgock Aober, mum input acceleiation as the rattlespace re-
An Analogous Problem quirement tends to zero. Our interest lies in
isolator performance between these practical
Another physical problem of essentially bounds. It may be observed, however, that
the same analytical form as the shock isolator nonoptimal systems easily can exceed these

206
 bounds. For example, if the rigid isolator TABLE 1
were thought to be a very stiff, but undamped, Wave Form Characteristicsa

f linearly elastic element, the transmitted ac-

celeration could be twice as great as the max-
imum input acceleration, even in a nonresonant Type Duraion T
Maximum Values
1 r
situation. Similarly, a low-frequency oscilla- Dy I VyI Ay tA
LA tory pulse could cause a very soft elastic sys-
tem to undergo excursions far greater than the I 00 0.5414 0.4612 2.000 0
maximum input displacement. II 4.0 0.6812 0.4620 1.425 0
i1 1I 4.8 0.1614 0.2970 1.C0 2.2
IV 4.8 0.8092 0.4630 1.430 0
Input Considerations
aMax IyI = aD
AThe problem formulation requires only that Max I 'i = vy and
the input displacement and acceleration be Max = - A).
tA).

bounded. Otherwise the solution techniques are

I . not influenced by the particular form of the in-
J* put. Reference Systems

by"The work reported herein was motivaLed For purposes of comparison with the opti-
by application to the shock isolation of equip- mum isolator, we have chosen a damped linear
ment within hardened military installations; oscillator subject to the same inputs. That is,
hence, we were concerned with input wave the reference passive system corresponds to g
forms related to ground shock motions induced in Eq. (1) of the (nondimensional) form
by high explosive and nuclear detonations. Re- . +
~ / sults have been obtained for four specific ve- g(xX) = cx+kx.
locity pulses whose characteristic forms are
shown in Fig. 3. In each case, we have selected the damping and
spring rate coefficients which best satisfy the
performance constraints given by Eq. (4). Hav-
- . ing selected g in this form, we no longer are
dealing with the optimal problem, but merely
the selection of c and k such that max IxI takes
St-- t- .>. on its minimum value for prescribed bounds on
o A.o TImt00-- T- CD. . This
optimal linear be termed the problem of the
couldisolator.

tA-o Time,t T-4.0

TYPE I TYPE I Also, for comparison purposes, we have
determined the response of the undamped linear
VY V" oscillator (c = 0) subject to the same inputs.
\22 Here, there is no consideration of "optmum,"
.> since the response is a unique function of the
T-~4.8 oscillator frequency.
t
> 60T Time -T-4.8
TYPE 1lT TYPE'V"
SOLUTION TECHNIQUES
Fig. 3. Input velocity wave form Available Approaches

During the course of our work, we have

The Type I input is given by identified four separate approaches, each of
different degrees of efficiency. These are
(t) = t(2- t) e-t for t >_0, documented in Refs. (1 and 2) and will be men-
tioned only briefly here. One of the methods
which is derived from y(t) t2 e". It is seen was used to generate most of the results pre-
that y(t) - 0, (t) - 0 as t - and is bounded sented, and this approach will be discussed in
and everywhere continuous. The other three somewhat greater detail.
- - velocity wave forms were prescribed as tabular
functions and in each case the base returns to a The optimal isolator problem would appear
rest position in finite time. The nondimensional to have its natural setting within the framework
wave form parameters are summarized in Table 1. of the calculus of variations. In fact, our

207
problem can be identified as a "Mayer Prob- A computational program using this scheme
lem," and it is possible to show that it is suf- was written for the IBM 7094 computer, and
ficient that the desired solution g*(t) be either results were obtained for a few values of D with
piecewise constant or g*(t) = -5(t) over cer- y(t) = t 2 e-t (Type I input). Results of accept-
tain subintervals of time. While this approach able accuracy were obtained, but only at the
discloses a "bang-bang" type of control, it does expense of dealing with fairly large tables of
not yet yield a convenient method of finding the 0i. It required approximately 12 minutes
'either the bounding (min-max) values or the of computer time to calculate 01(oo; D) for
complete details of the optimum isolator function. each value of the displacement-bound D. While
this program did not provide for calculation of
The fact that the optimum isolator consists g-(t), a somewhat lesser time would be re-
partly of constant force increments suggested quired for the additional calculations.
a graphical procedure for constructing g*(t).
This method is particularly convenient where
the initial force increment is the min-max value. Linear Programming

The method of dynamic programming (3) The method of linear programming was
provides a general solution technique whose found to be the most efficient of the techniques
disadvantage is a relatively major computing investigated. The major achievement, if it may
effort. This approach requires the problem to be called that, lay in recognizing that the prob-
be expressed in discrete, rather than continu- lem possessed a linear programming formula-
ous, form and leads to the following functional tion. While by way of hindsight, such a formu-
equation: lation indeed is seen to be obvious, we must
confess that it took some time for this to occur
4 1 , i~1 ' D)]
4i(ziiD) = min-max [ai, i41 (zi , to us.
(5) As in the method of dynamic programming,

where we first pose a discrete version of the problem.

The total time interval of interest is divided
a. assumed constant value of Yduring into n subintervals of length At. The accelera-
the ith time interval; i.e., (i - 1) At tion i is assumed to be constant over each sub-
< t < iAt for i = 1, 2,..., n, interval; i.e.,

ii = value of i at beginning of (i+ 1)st i(t) = ai , a constant for (i-1) At < t < iAt
interval (end of ith interval) = i=1,2. n. (6)
ii + aiAt,
Denote the value of z at the beginning of the ith
Zi = value of z at beginning of (i + 1)st interval by zi, and similarly for the other quan-
interval = z+ iAt + 1/2 ai (At) 2 , and tities of interest. Thus, zi = z[(i- 1) At]; etc.
The initial conditions of Eq. (2) then become
= minimum value of max Iaj I for
i < j <n+l such that Ixji= [z-yjl <D z = it = 0. (7)
for i<j <n+l.
The motion z(t) now is completely deter-
The calculation is started by observing that mined by the set of numbers a, and the initial
O, I = 0 for permissible values of Zn 1,and conditions of Eq. (7). We seek that particular
application of Eq. (5) then yields a table of val- set of ai having the property that
ues of On. In this fashion Eq. (5) is used as a
recurrence relation until ultimately 0 1 is cal- i-
Yz D; i 1,2... n,
culated. For the stated initial conditions of
interest, Eq. (2), the desired min-max value of and (8)
acceleration is seen to be max ia1I[ a minimum.
A = min -max =I 1(0,O;D).
That indeed this is a problem of linear program-
The function g*(t) is obtained from the table of ming is shown in the following derivations.
values of a, which are found during application
of Eq. (5). This requires a second computational Equation (6) has as its solution
pass, since the appropriate values of ai are not
=
recognizable until the a1 corresponding to i1 +aiAt
01(O,0;D) isdetermined. zz + i1 At + 1/2a(At) 2 for i 1,2..., n

208
 with z1 = o. A general term for zi in- desired optimal isolator function g.(t). All
volving only the a i can be slown by induction results presented herein were obtained by this
to be: means.

A2 [2(i-j) + 1 aj; i- 1,2.... n RESULTS

2(9)
Optimum Isolatbr Performance
As a consequence of Eqs. (8) and (9), the a1
must satisfy the following set of 2n linear The numerical results characterizing the
inequalities: performance of the optimal isolator for each of
the four wave forms considered (Fig. 3) are
2 ]tishown in Figs. 4, 5, 6 and 7 as the curves
2 [2(i-j) + 1] aj Yi+1 - D labeled "Optimum Isolator." Also shown in
j=1 these figures is the performance of the refer-
(10) ence passive isolators, i.e., the optimum
2damped linear oscillator and the undamped
2 [2(i-j) + 1] a > Yi+ + D oscillator. The performance of the latter is
j=1 1+shown only in part, since otherwise the accel-
1,2 neration scale would have been compressed con-
for i 1siderably. The complete performance curve
for the undamped isolator is shown in shock
Introduce the quantity p and require that spectral form for two of the inputs in Figs. 10
a i < 0 and 11.
"
and for i=-1,2,..., n. (11)
ai
10ULI:
As Eq. (11) is equivalent to setting k = max IaI-o6,o.31 at0A0.38
we may take 0 to be a linear objective function " ax,31 at X.
which we seek to minimize, subject to the con-
straints of Eq. (10). This essentially completes -
0.8 Se F 10
see Fi.o _0
the linear programming formulation. To put it cc Undamped Linear
more in terms of standard linear programming g 0.6 Isolator (ULI)
1 language, we may introduce the slack variable
si defined by si = -as.

Then Eqs. (10) and (11) become

I 0.4
U4Optimum

E Damped Linear
2= Isolator
.90.2 O9ptmm
1)S, ih DZ22) - Islao UL,
j
[2(i-j)+ist-i 2
2 2)-D -Yi. (--
(12) 0 0 0.2 0.4 0.6
L
0.8 1.0
Maximum Displacement Ratio, U'-Dy

(2(+1)-i),2\i (2
2
E At-2 > At Fig. 4. Tradeoff limit diagram,
20- s i > 0 Type I input

for i = 1, 2,..., n. Our linear programming

problem, finally, is to minimize 0 subject to The results are portrayed as plots of the
the 3n linear inequalities of Eq. (12). (The min-max values of relative displacement D and
first of Eq. (11) implies s i >0. This need not absolute acceleration A, normalized by the
be stated in Eq. (12), since it may be consid- maximum values of the respective input dis-
ered an implied condition in a linear program- placement Dy and acceleration A . We have
ming problem.) termed these tradeoff limit (TL) curves and
"J diagrams for fairly obvious reasons. It will be
We have adapted a standard linear program noted that TL-ratios are independent of the
routine for the IBM 7094 to the present problem. nondimensional scale factors. As was dis-
The solution time for each value of D and n = 50 cussed earlier, in the section on limiting situa-
is about one minute. Output consists of 0 and tions, the curves necessarily pass through the
the set of a, which, by Eq. (1), provides the points D D/Dy 1, A A/Ay 0, andi = o,

209
 1.0 1.0

* o 0.8
. Undamped Uneor 140.8
K 0.
., - Undamped Umr ,Isosoatoo Undamped Linear
Zsoalo isolator
X
0.6 0.6
_ a Optimum
Optimum Damped Linear
0.4 Dsmped 0.4 Isolator
E IssotlaOpimu E
0.2 -
E 0.2 -Optimum02
Isolators o oo

0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.6 LO

--
Maximum Displacement Ratlo,- Maximum Displacement Ratio,15=y
Dy

Fig. 5. Tradeoff limit diagram, Fig. 7. Tradeoff limit diagram,

Type II input Type IV input

.o0 . force increments, as was discussed in the sec-

2. ULI: Amox=4.1 at D=1.55 lion on available approaches. We also know
0max=2.45 at A=32 that a unique characterization of g*(t) is not
see Fig. II possible once the min-max values have been
.o pachieved. This situation is seen clearly in Figs.
C0.6 Damped Linear 8 and 9 which pertain to the particular solution
. Isolator for Type I input and a displacement bound of
D =0.1.
S0.Figure 8 shows the time variation of the
E relative displacement x. The solid curve up to
"E0.2 Optimum ULI t .5 (point C) is plotted directly from the output

Iolator
0-
0 0.2 0.4 0.6 0.8 1.0
"
Maximum Displacement Ratio, D= 0.I2
Dy 0.10- B
Fig. 6. Tradeoff limit diagram, 0.08
Type Ill input
>1 0.06
0.04*

= 1. Each point on the optimum isolator curve - 0.02

(D,A) corresponds to a particular isolator "de- E 0
sign," having the property that its (nondimen-
sional) maximum transmitted acceleration is .
CL
-0.02 /
AA, and its (nondimensional) maximum rela- !2/ -. 04
tive displacement (= DD) is the least possible, / X .
or, correspondingly, that its maximum relative -0.06 * /
displacement is ND y and its maximum trans- -0.08 ,-3
mitted acceleration(= TA) is the least possible.
The significance of these results to practical -0.1 cX X4;

isolator design is discussed later. -0.12

I I I I

Optimum Isolator Function 0 2 4 6 a 10 12

Time,t
function g*(t) is
The optimum isolator
known to consist in part of piecewise constant Fig. 8. Optimum isolator displacement

210

- -- - - -- r-~- ~- - -
 o0. in Fig. 9. The input acceleration 5(t) is shown
in Fig. 9 for comparison purposes.
0.6 For times t > 1.88, a variety of solutions
-" for g(t) are possible, several of which are
0.4 shown in Fig. 9. The dashed curve labeled g2
is to be viewed as a continuation of g,(t) for
t >s. Thus,

0.9 1 (t) 0O<t <5

o 0r o . . . . g2(t) 0.055 5 < t < 6.35

L0 t > 6.35.
I.
.9 -0.2.
0 This function has the effect of bringing the
mass to rest at t = 6.35 so that at later times,
-0.4 x*(t) = -y(t), as shown by the dashed curve in
-" Fig. 8.
Another possibility, g(t), is an extension
of g (t) for t > 1.88, and is represented by the

' -0.8
-0.6 4 dash-dot curve g* in Fig. 9, where
. (t) g(t) 0 < t < 1.88

0 2 4 6 8 to 12 -5(t) 1.88 <t < 2

I , Tlmejt 93()

-0.075 2 _ t <4.07
Fig. 9. Optimum isolator function

This function has the effect of causing the mass

of the linear programming solution which ter- to execute the same nmotion as the base for
minated at this time. At later times the rela- t > 4.07, so that x- - o beyond this time, as
tive displacement would continue to oscillate shown in Fig. 8. Other possibilities might in-
between the bounds 0.1, despite the fact that elude letting g(t) = -j(t) beyond the time at
the min-max acceleration A occurs initially, which x* assumes a bound value. Such times
this being in the nature of the linear program- are indicated by points A, B, and C in Fig. 8,
ming formulation. where x* = -0.1 for t>2, x, =0.1 for t > 3.6,
x6 = -0.1 for t >5.
The isolator function corresponding to this
solution is shown as the solid curve, labeled
9 in Fig. 9 up to t =5. The min-max accel- DISCUSSION
eration A = 0.59 persists in the interval
0 < t < 0.75, at which time the displacement Three types of questions would appear to
bound is reached. The relative displacement interest the shock isolation system designer
maintains the bound value in the interval during the initial phases of his design:
0.75 < t < 1.88, and then another constant ac-
celeration increment occurs. During the time 1. What is the margin of improvement of-
that x* D = 0.1, we have .= 0 and, hence, fered by the optimal system over some other
g*(t) =- (t). The oscillatory behavior of candidate design?
x*(t) during 0.75 < t < 1.88 (Fig. 8) results
from the inability of the linear programming 2. How dependent is the optimal (or near-
solution to completely represent j(t) by con- optimal) isolator performance on the details of
stant acceleration increments during this in- the optimal isolator function?
terval; the approximation, of course, improves 3. How dependent is the optimal isolator
as time increment, At,
the as the muchasbehvio is
he atua is reduced. Inas- perfoa
now, itwasperformance, ende ited
and the associated isolator
isolator func-
much actual behavior is known, it was tion, on the details of the shock input?
the nu-
not considered necessary to improve
merical solutioa further. The function g*(t) is The present results pertain specifically to
represented as being identically equal to j(t) the first question and permit several inferences
211
 to be drawn concerning the third. The matter undamped system results in an enhancement,
of isolator function sensitivity can only be rather than an attenuation, of the input dis-
guessed at for the present. placement for this input and acceleration (fre-
quency) level. In marked contrast, by means
With regard to margin of improvement, it of the optimum isolator* the rattlespace re-
- is clear that the tradeoff limit curves of Vigs. quirement can be reduced to D = 0.17 (Fig. 10,
4 through 7 provide precisely this informadon. point 2) for the same leiel of acceleration at-
It may be well, however, to show a specific ex- ienuation. This corresponds to a relative dis-
ample. Figure 10 shows a conventional shock placement attenuation of D = 0.31, or a per-
spectrum for an undamped linear isolator sub- centage improvement of [(1.3-0.31)/0.31](100) =
ject to the Type I input. (The numerical values 320, which is a factor of three better than that
pertain to the nondimensional form of the iso- of the undamped system. The only way in which
lator and input; thus, while the values of fre- an undamped isolator could meet the rattlespace
quencies in cycles per second and acceleration requirement of U = 0.31 would be by transmitting
in g units indicated on the figure appear unre- an acceleration about 10 percent in excess of the
alistic, they do pertain to practical situations input level (Fig. 10, K = 1.1, point 3).
of isolator design for protective construction
application. As noted earlier, the normalized The viscously damped linear isolator offers
form of acceleration and relative displacement considerable improvement over the undamped
performance are independent of the nondimen- system, as can be seen from Fig. 4. Thus, for
sional scale factors.) K = 0.2, D = 0.4, which is only about 29 percent

10

Vo ,/ )/Z \ 0

aIsolator

10 1010
Frequency, VA/27r (cps)

Fig. 10. Optimum and undamped linear isolator

shock spectra, Type I input

Assume that it is desired to reduce the less effective than the optimum isolator at this
peak transmitted acceleration to 20 percent of level of acceleration attenuation. The margin
the input value. From Table 1, A = 2, so that of improvement becomes relatively less dra-
in the units of Fig. 10, the desires (nondimen- matic at increasing levels of acceleration
sional) acceleration level, A, is 0.2 x 2/386 =
0.001 g. If the undamped system of Fig. 10 *Neither the frequency nor pseudovelocity scales
were to be used as the isolator, the required of the shock spectrum have any significance
rattlespace would be D = 0.7 (Fig. 10, point 1). for the optimum isolator. Rather, the curve
This corresponds to a relative displacement merely expresses the appropriate relationship
attenuation of U = 0.7/0.54 = 1.3. Thus, the between displacement and acceleration.

212
 attenuation, i.e., for extremely "soft" systems. We consider now the extent to which the
Referring again to Fig. 10, for W= 0.05, A = present results contribute to an understanding
0.00026 g, and the rattlespace requirdments for of the dependence of optimal performance on
the undamped and optimum isolator differ only the input wave forms. Figure 12 is a composite
by about 10 percent. As is evident in Figs. 6 plot of the tradeoff limit curves for the four
and 11, the relative inefficiency of the undamped wave forms considered. (The additional curve
?. isolator can be greater at all acceleration lev- in this figure will be discussed later.) Figure
els for inputs of a predominant oscillatory na- 13 is a similar plot for the undamped and opti-
ture such as Type Ill. mally damped linear systems. The relatively

10 7

0/

I~~~.
10/\ ,..
-II

i p~timum"

01

Fig. 11. Optimum and undamped linear isolator

shock spectra, Type II input

Generally, then, comparison of isolator 1

system performance, either relative to the
optimum or to a more arbitrary standard, is Z
conveniently handled by means of the tradeoff I< 0. Input Type
limit diagrams of Figs. 4 through 7. The dif- o
ferences in abscissae between a candidate sys-Z
tem and the optimal at prescribed levels of 0.6 In
acceleration attenuation is a direct measure of Z
the margin for improvement in rattlespace per- 0.4
formance. It also suggests a meaningful way of , 0ExtLended Optimum
assigning some sort of performance index to E Type I
candidate systems. It should be emphasized 0.2
that combinations of acceleration and relative
displacement attenuation levels, which lie in
the region between the coordinate axes and the -.
optimal curve, are not physically attainable for 0 0.2 O.4 0.6 0.8 1.0

the input shock under consideration. That is to Maximum Displacement RatodI

say, no isolation concept exiss whose perform-,,y
ance falls into this region. Such information
can be of considerable help to the designer if Fig. 12. Influence of input wave
he should be confronted with unachievable per- form on tradeoff limit curves for
formance criteria. optimun isolator

213
 small spread of the four curves in Fig. 12 indi- bounds to the performance sensitivity for the
cates that the optimum isolator performance is associated inputs. Additional study is required
relatively insensitive to major variations in the to determine the actual sensitivities.
input parameters. As these wave forms are
both characteristically different and realistic It should be noted that the distinction be-
.1 for certain applications, this result is promis- tween general system performance and specific
ing from the design point of view. It must. be isolator performance becomes unimportant
emphasized, however, that no claim is made to when the optimal system is intended solely as
the effect that the performance of an optimum the basis of performance rating or margin for
isolator is insensitive to the input, but only that improvement. Here, one would be quite satis-
the optimum performance, for each of the inputs fied to know that the optimal relationship be-
considered, do not differ by much. The dirtinc- tween acceleration and relative displacement
tion can be important, as will be described, attenuation is insensitive to input details, with-
out regard to the sensitivities of particular
designs. The present results suaggest this to be
1.0 ,the case in practical situations, but additional
inputs should be considered in a systematic
manner.
0.8 Type- \ \t
I; It is certain, of course, that inputs do exist
for which the optimum performance curve must
0.6 " differ appreciably from those shown in Fig. 12.
A particular example is discussed below. An
extreme case is one in which the input exhibits
0.4 a small positive velocity for long times after
the maximum acceleration response is achieved.

0.2Here the maximum input displacement, Dy, can

o - Dampedbe arbitrarily large, so that the optimal per-
r _ " mformance curve must approach the coordinate
I\' axes; regardless of the values of A and D. This
0 0.2 0.4 0.6 0.8 1.0 suggests the pragmatic view that input sensi-
6 Dtivity be considered for variations in input wave
M DisplacementloD=
Dximum -Dy forms about some nominal form, rather than for
extreme wave forms unlikely to occur in a given
Fig. 13. Influence of input waire design situation.
form on tradeoff limit curves for
undamped and optimum damped lin- So far the optimal isolator has been as-
ear isolators sumed to start from conditions of rest at the
time of arrival of the input shock (Eq. (2)). It
is of interest to relax this condition and con-
sider the isolator performance, which is opti-
Consider some point on one of the curves mal with respect to Eq. (4) and the totality of
of Fig. 12, e.g., point 1 for the Type I input, admissible initial conditions. Since, on physical
This point, as does every point on the curve, grounds, the isolator must start from rest at
represents the performance of an optimum some time, e.g., t , this extended problem is
isolator designed for the indicated level of ac- identical to the original one, except that the
celeration and displacement attenuation. If 'this solution is obtained over the interval t < t < T.
particular isolator were subject io the Type HI If we retain the original origin of time, t < 0,
input, its performance would plot as some point and Eq. (2) is replaced by z(t 1 ) -(ti) 0.
to the right of curve HI; in general, it would not The desired solution, termed the extended opti-
lie cn the curve. A point on curve III having the mum, is shown in Fig. 12 for the Type I input.
same acceleration attenuation ass the
sameaccleraionattnuaton te first
fist iso-
so- Thimov
The improvement enoertepvouotma
over the previous optimal
lator (point 2) would be optimal for wave form solutions clearly is significant over much of the
1I but would be a different isolator than the rneo ceeainatnain
first. In this example, just where the first iso- range of acceleration attenuation.
lator would plot on Fig. 12 is not known; all that
can be said is that for prescribed acceleration The concept of the extended optimum, that
attenuation, point 2 represents a lower bound is, an isolator which is warned of what is to
on the relative displacement performance of come and able to take appropriate action before
the point 1 isolator when subject to Type I the fact, is not inconsistent with the optimum
input. In this sense, any two of the curves in isolator just considered. That, too, was antici-
Fig. 12 may be viewed reciprocally as lower patory in nature and, hence, an active isolation

214

/ g
 A
system. However, the extended optimum is not of the non-uniqueness of the optimal isolator
1
generally as realistic a standard of isolator function (Fig. 2) seems to make plausible the
performance, since it applies only to pre- contention that many classes of isolators can
actuated systems. be adapted to approximate reasonably well the
desired response function. Clearly, additional
r We have not as yet obtained specific data study of this aspect is reqiired.
concerning the dependency of the optimum per- *

formance on the details of the isolator function.

Such understanding, of course, is mandatory ACKNOWLEDGMENTS
. before serious consideration can be given to
specific hardware designs. It must suffice for This paper is based on work performed
present to suggest that the optimal performance for the Air Force Weapons Laboratory under
may not be critically dependent on the isolator Contract No. AF 29(601)-6487, and supported
function and, in fact, may be rather insensitive by the Defense Atomic Support Agency under
to these details. This guess is based on two Project 5710, Subtask 13.167. A companion
observations. First, the damped linear isolator effort was supported by the Naval Research
was seen to approximate modestly well the op- Laboratory under Contract No. Nonr-4444(00)
timum performance, despite the fact that its (X). S. Cameron and T. Church of IITRI
time response is markedly dissimilar to that of were responsible for the computer formula-
the optimal isolator (Fig. 9). Secondly, the tion of the dynamic and linear programming
graphical argument presented earlier in support solutions.

REFERENCES

1. "Optimal Shock Isolation Synthesis," IT Re- 3. R. Bellman, "Dynamic Programming"

search Inst. Final Rept. M6084, July 1965; (Princeton University Press, Princeton,
to be published as Air Force Weapons Lab. N.J:), 1957
Rept. WL TR 65-82
2. "Optimum Shock Isolator and Absorber De-
sign Techniques," UT Research Inst. Final
Rept. M6078, April 1965

DISCUSSION

Mr. O'Hearne (Martin Co.): Would you - Mr. O'Hearne: I made the comment be-
comment on the adaptability of this method for cause it seemed, in a way, a shame to do such
a much larger class of inputs, say a stochastic elegant analysis for specified'inputs where, in
input? the general situation, one finds they are usually
quite arbitrary.
Dr. Sevin: It would be extremely interest-
ing to look at this approach for an input which Dr. Sevin: Yes, I agree with you. To be
is described as a random function. We have sure we sedom know inputs. There is another
not done that yet, but we plan to. To be sure, potential for analysis which is quite different
we have considered only a limited number of than that which I have mentioned. Consider a
inputs and in many instances, one looks at the system that is optimized with respect to some
response to a superposition of a number of inputs. Use that optimal solution to pose the
inputs. It is interesting, I think, to observe that mathematical problem of specifying a class of
the very nature of the optimal solution permits inputs which varies about the one for which the
one to use superposition. The system may be optimal was found. In a general way, these
subject to four different inputs which may occur should have certain properties that define them
randomly or, at least, in an unspecified way as a family of inputs. Then seek to determine
with time. Generally one would have to consider the bounds to the response and the uncertainty
a superposition of all possibilities. The very in the response of the optimal to a class of in-
nature of the optimal solution implies a trial of puts. I am pretty sure that problem can be
each one independently. The best performance solved.
for any one of the four would be optimal for any
combination of the four. I think that is an inter- Mr. O'Hearne: That is very interesting,
esting observation, thank you.

215

4e-,
 4r

VIBRATION ISOLATION THROUGH

PNEUMATIC SPRING AND DAMPING
Russel L. Hall
Edgerton, Germeshausen & Grier, Inc.
Las Vegas, Nevada

This paper concerns the passive isolation of harmful shock and vibra-
tions as applied to a relatively large package containing fragile data
acquisition equipment. The shock isolator consisted of a lightweight
cylinder containing compressed air and a piston which is fully extended
at steady state. The air cylinder isolator functions as zero deflection
support, spring and damper. Electronic simulation indicated that the
air cylinder isolator will more effectively isolate single shock pulses
than will other conventional systems.

NOMENCLATURE a, 8, y, Solution constants which depend on

C1 ,C 2,c 3 problem initial conditions
My Package mass, Ibm
MIsolator mass,Ibm Forcing frequency, radians/sec
P Pressure, lbf/sq in.

B Linear damping coefficient, lbf/ft/

sec T Temperature, OR

K Spring force, lb f (nonlinear); spring V Volume, cu in.

rate, lbf/ft (linear)
m Air mass, lbm
y Package displacement, ft
Air flow, lbm/sec
ft
x Isolator displacement,
ms Sonic air flow, lbm/sec
ft
z Input displacement.,
At Venturi throat area, sq in.
F Force, lbf
v Specific volume, cu ft/lbm
t Time, sec
a,b Real and imaginary parts of a differ-
A Piston head area, sq in. ential equation root

As Piston stem area, sq in. e Natural log base, 2.718

Ad Damper port area, sq in. Q Heat, BTU

k Ratio of specific heat of air, 1.45 Cp Specific heat at constant pressure

g Gravitational constant, 32.2 lbm ft/lbf Cv Specific heat at constant volume

2
sec
C Coefficient of flow discharge
- p Air density, lbm/CU ft
D Diameter, in.
y8 Isentropic expansion factor
R Air specific gas constant, 53.36 lbf 0 Adiabatic expansion factor
ft/lbm OR
217

--- 4
INTRODUCTION My= Ky-MyG, (I)
This analysis was performed to determine Mx= Bx(- ) + Kx( - x) Ky
a method of shock isolation for excitations orig-
inating through the base of a system. The pack- - Mxg - Ad(Pl -P2) (2)
age of interest resembled a large commercial
trailer (Fig. 1) and was designed to be tempo-
rarily located at ground level in areas subjected Ky was simulated as a nonlinear force. All
to severe jolting earth movements. Efficient throttling processes were assumed isentropic.
isolation through the air spring damper ap- r
peared to require that the nonlinearities associ- rigid with the exception of the pneumatic tires.
ated with the air spring offset the nonlinearities The system was recognized as a nonautonomous
associated with the air damper. system, i.e., a system whose properties are
affected by the independent variable, time.

Air damping control was designed to oper-

ate in open loop control fashion with fixed uni-
directional orifices in the piston head.. Air
friction was imagined as a thermodynamic
irreversibility causing a temperature increase
/I in the throttled air. This damping inefficiency
was not included in the electronic analog prob-
lem simulation due to the estimated negligible
effect on overall system response. Air damping
was adjusted so as to isolate most effectively
" 4shock pulses in a predetermined range.

Fig. 1. High resolution trailer INVESTIGATION

in mobile position
Mathematical techniques were used to study
the response of the configurated single mass
The trailer body, wheels, and air spring linear system. The package acceleration profile,
damper support were imagined as passive i.e- it was postulated, would be a measure of the
ments arranged as shown in Fig. 2. Linear isolator efficiency. Three isolator configura-
nodal force balance equations may be written, tions were investigated using linear differential
describing one degree of freedom and lumping equations having constant coefficients. Figures
all vibration control into one isolator: 3, 4, and 5 represent these systems and their
associated acceleration profiles for perturba-
tions noted.
The given specifications required a 40,000
+F My lbm to be shock isolated from a single strong
earth movement. It was required that the pack-
/age acceleration profile be described by the
minimum absolute value for any reasonable
isolator. The Electronic Associates, Inc.,
KY TR-48 Analog Computer was subsequently used
to describe a more detailed multiple mass sys-
tem employing the pneumatic spring damper.
Linear differential equations having nonlinear
_MX coefficients were programmed for the elec-
tronic computer simulation. Figure 6 illus-
+F +X trates the TR-48 Computer in operation.

Figure 3 illustrates a (air) spring mass

system response and typical package accelera-
F SINE wlttion profile after the system was subjected to
some predetermined free fall height. The (air)
Fig. 2. Isolation through spring was assumed to have zero deflection at
air. spring and damper steady state:

218
1 - .- - o-

00

0j t
Fig. 3. Isolation through a linear spring

2M

0*-
0

Fig. 4. isolation through a viscous damper

a+

F SINE wt0

Fig. 5. Isolation through linear spring

and viscous damper

'M Mg-(F+K'), (3) was assumed that the package would come to a
full stop within the single damper stroke:
. = -_
K C1 sin +a .(4) = M -B(5
Mj Mg - B (5)
The values of C. and a were determined using B B
problem initial conditions. - M Ce exp - - t

Figure 4 illustrates a damper mass system The value of C, was determined using problem
response and typical package acceleration pro- initial conditions.
file after the system was subjected, again, to
some predetermined free fall height. Since The package acceleration profiles derived
there is no restoring force in this system, it from Figs. 3 and 4 were considered inefficient

219

T7I
 Fig. 6. Analog computer

with regard to the type of isolator used. Figure typical force displacement relationship derived
5 illustrates the passive elementa previously from pure air spring isolation. Analysis was
described arranged in parallel to form still an- performed according to the familiar isentropic
other type of isolator. It was assumed that the relationship, for a perfect gas, P =Po(vo/v) exp k.
package base would at no time be separated
from the forcing input function. A sinusoidal For purposes of comparison, time history
perturbation was used: tapes were made from an electronic analog
simulation of the air spring mass system. Fig-
M; = (F - Mg) + B( - ') + K (x sin cot - y), ure 9 illustrates the response of the package
described above when forced by a single posi-
y - - Cleatb 2 sin (bt + a) + 2aCleatb cos (bt + a) tive sinusoidal pulse having a 4-ft amplitude.
The pulse time durations were varied stepwise
+ a2C~e at sin (bt + a) - o sin (cot +,6) from 1.43 to 0.21 sec. Figure 10 illustrates
oC 2 sthe typical (air) spring mass magnification fac-
tor as curve (D.
-co2C3 sin (o 8

The air cylinder piston configuration previ-

The values of C,, C2 , C3 , a, i3, and -/ were de- ously discussed forms an (air) spring damper.
termined using problem initial conditions. Equations (1) and (2) describe a two-node (mass)
system incorporating an (air) spring damper.
The acceleration profile described in Fig. The following relationships were incorporated
5 was set aside in search of a more efficient into the electronic analog model to describe
isolator. (air) spring damper nonlinearities:

Logic gave use to the previously investi- K = AP (Ap- A,) P2 (9)

gated spring and damper elements arranged in
series. It was imagined that an air cylinder MiRTI
piston combination would physically provide the P1 = -- (10)
series configuration desired. Figure 7 illus- I
trates the hardware design of this type of iso-
lator. Functionally, the piston stem provides P2 RT2 (11)
zero static deflection spring support and the 2 -'-(11)
v2
piston head provides bidirectional damping. mI m - in2 , (12)

As a prelude to the above study, the non- t

linearities associated witha (air) spring mass m, = f (dt, (13)
system were studied. Figure 8 illustrates the

220
:1

I. CYUNOER ASS*Y
~.Z2 SCISSORS TOP HALF
". 3. SCISSORS LOWER HALF
2 -*4.
5. AXLE
UPPERASSEMBLY
SHAFT
.- 6. LOWER SHAFT
("7, FLAT WASHER
8. COTTER PON
9. HEX SOCKET SHOLDER SCREW

.0=. 44

Fig. 7. Wheel strut design

LOW VOLUME
MAXIMUM
NALLOWABLE
FORCE

iij
* ...-
LOW PREssiJRE

K- HIGH VOLUME

S0 MAXIMUM
0

y ALLOWABLE
~ STROKE

Fig. 8. Typical air spring rate

V, =v 0 -AP(x-Y). (14) and from Obert (1)

V 2 = (A -As)(x-y) (15) A= 0.525CY,8 kD 2 (p,(PI-P 2 )] (18)

k-i The air flow r was limited at sonic velocity

(p k (16) using the following relationship:
T== pA(kgRT) (19)

p k It was recognized that a pneumatic damp-

2 (17) ing inefficiency does exist for the thermody-
namically isentropic throttling process which

221

I7I
+ 50 flet/oec:
Ic
2

-9N
f 4
' e_ m h 4 10 -.1 4 - fc I -

10 feet/sc m
I

R4141A
A
2,feet//

Fig. 9. Response of air spring mass system

to single positive sine pulse

was assumed. This irreversibility was esti- modeling revealed that the series air spring
mated to be insignificant in the overall problem damper isolation compares favorably with the
and was derived as conventional parallel linear spring damper iso-

Q
rr(T-T)1). F
1- Cv(Ti -
T\l
T+ "(20)
lation when responding to forcing frequencies
below te seomnaturalfrequency. At hghr
frequencies, the air damper appeared to be less
effective because of the compression of the air
The effects of system damping were stud- spring. For the above to be true, the piston
ied by flowing air from one side of the piston head area (air damper) should be designed as
head to the other as the piston moved relative small as possible. Fignre 11 illustrates the
to the encompassing air cylinder. System response of a system isolated from the base ,

222

T4+ -%
I1 1

Fig. 10. Comparison of system

4response magnification factors

forcing through an (air) spring damper as de- System responses of the air spring damper
scribed above, isolator were investigated at higher frequen-
cies. Figure 12 illustrates a sine wave per-
Particular attention was focused on the turbation having a frequency of 8.8 cps and an
package acceleration profile and the amplitude amplitude of *0.5 ft. It was noted that the sys-
displacement for perturbation in the system tern had just approached an unstable condition.
natural frequency range. Figure 10 shows a Figure 13 shows the same system after a
typical air spring damper mass magnification damper had been placed between the package
factor as curve (I. Curve (g)indicates the mass and wheels of 200 lb f/ft/sec.
typical linear spring damper in a parallel iso-
lator configuration previously discussed.

REFERENCE

1. E. F. Obert, Thermodynamics (McGraw-Hill, New York), 1948, pp. 130-153

BIBLIOGRAPHY
; Doebelin, E. 0., "Dynamic Analysis and Feed- Thomson, W. T., "Vibration Theory and Ap-
back Control" (McGraw-Hill, New York), plications" (Prentice Hall), 1965, pp.
1962, pp. 26-57, 76-152 130-153

Olson, F. E., "Vibration Isolation Study H Com-

puter Simulation," EG&G Memo., Aug. 6,
1965.

223
 T .- -- - - -

FI

-224

-- ---- - ----.-- ---- -.-- .--- ~~* p-.- A-

So
2
feet/SeC Ica ~

71I

20 feet/u Ic

2 feet/cu

14 44

~~ocrzmRvo~i~
?AlRSPED25 1/ICCTImRPS55Z
5 oo x3- A . L

AA- Fi.1.Rspneo i srn asdaprsse

fo hihrsiuodatocigfeqece

a4= rS ALMU BONN9.0

2.5 f225
 FIIt

iti

nm flWUmY 1~ s Sawuwmm 0.5 nIT -UWT 9.0 CPSain 0.3 MW

2.5 fet/2

4
P r44 ;f i 4 '' ~ ! ~ .. Mr, 4 n

PIL -4NS

2266
ps/

; T I. 0rw,-~ - ,. ,T 77
.q W!,:,
 DESIGN AND ADVANTAGES OF A TWO-STAGE MOUNTING
SYSTEM FOR MAJOR MACHINES IN SHIP'S ENGINE':ROOMI

Robert M. Gorman
General Dynamics/Electric Boat Diision
Groton, Connecticut j
Approximately 100 tons of low-speed machinery had to be isolated in a
surface ship's engine room. Standard single-stage mounting systems
with sufficient attenuation characteristics would not have the rigidity to
withstand the ship's rolling and slamming. It was decided, therefore,
to use a two-stage mounting system incorporating a massive subbase
on which all the machines would be isolation-mounted. The subbase
would, in turn, be isolation-mounted to the ship's hull.

An analytical comparison of one- and two-stage systems was made on

the basis of attenuation and stability characteristics. A general trans-
missibility equation for the two-stage system was derived. Force
transmissibility was analytically determined for a large range of mount
damping, mount stiffness, and subbase weight. Results include a set of
parametric curves. Hull impedance, obtained from experimentalmeas-
urements on a similar hull, was included in the system's mathematical
model.

Four major machines were on the subbase. These and the subbase it-
self were considered rigid masses to simplify determining ali the
mounting system's natural frequencies. Each element was given six
aegrees of freedom and the resulting 30 natural frequencies were found
using Electric Boat's standard structures program.

INTRODUCTION frequency of 1/2 cps, a static deflection of ap-

proximately 40 in., and a stiffness of only 2000
Effective vibration isolation of heavy ma- lb/in. Clearly this system would be too flimsy
chinery having rotational frequencies less than to expose to the rolling and slamming experi-
30 cps is a difficult problem. One method that enced on surface ships.
may be used is a two-stage mounting system,
consisting of isolation mounting machines on a
rigid massive subbase which, in turn, is mounted NOMENCLATURE
on the ship's hull. The added weight, space, and
cost requirefients are counterbalanced by an A Constant of proportionality, in.
isolation system that is far more effective than
the standard one-stage type. B Constant of proportionality, in.

The particular case investigated here in- c Damping coefficient, lb-in./sec

volved isolating two 80,000-lb and two 17,000-lb
diesel generators. The large diesels run at 12 e Base of natural logarithm system
cps and the smaller diesels at 20 cps. The ship
has an extremely severe radiated noise require- f Frequency, cps
ment which calls for all equipment to meet a sea
state 2 waterborne noise level at 20 ft. This F Force, lb
means the isolation system for the large diesels
has to reduce the fundamental unbalance forces
by a factor of 1/625 or 56 db. To meet this re- J
quirement, the standard sprihgs used in a one-
stage isolation system would have a natural k Stiffness, lb/in.

227
 M Mass, lb-sec 2/in. This transmissibility is plotted versus fre-
quency in Fig. 2. Note that at frequencies above
&0 Radian
t frequency, rad/sec db/decade. the curve has a negative slope of 40
resonance,

\ c1xy k,;i.Wy, rad/sec

t Time, sec Two-Stage Systems

T. Transmissibility Figure 3 shows a schematic of a two-stage

mounting system. As before, damping is ig-
x Displacement (vertical), in. nored and motion is considered in only the ver-
tical direction. Equations of motion for this
Acceleration (vertical), in./sec2 system are

x + kl(xl-x
mrn 2) = F sin wt (5)
ANALYTICAL COMPARISON OF
ONE- AND TWO-STAGE and
MOUNTING SYSTEMS
m2x 2 + kl(X2 -x) + k2x 2 0. (6)
One-Stage System
As before, transmissibility is defined as force
This basic system is shown in Fig. 1. For output divided by input, or
simplicity, damping has been ignored. The basic
equation of motion for the system is k 2x 2 k2x 2

rn:W I +k x, Fsine&t.F sin wIt nx + k 1(x 1 -x 2) .

I
The displacements and accelerations will be of
the form x= Aej i t , i1 = -A,)2 ejt, x 2 = BeJwt,
F sin wt and x2 = -B0)2 eJi&t. Substituting these into Eq.
(6) yields
I " Fig. 1. One-stage kj " k- 2i
X1 mounting system A= 2 2 (8)
KI k ,

If the displacements and accelerations, along

with the ratio of A/B, are substituted into Eq. (7),
the following equation is found: .

Transmissibility is defined as the ratio of

the force output of the system to the force input: T - 1
2)2
1 (2 1k 1 12
F ut k xl kx 1 / k kk
T =F sInctni kxj (2) 12 12
Fin = sin ot mlxl + kl 1
Now, letting k 1 /m 1 k = 21y
The displacement, x,, will be of the form a k 2 E (9) b 2

x 1 = AeJwt; hence, , 2 ejwt. Substituting

-A~ ,a 2/M 2 2 21
q.
these values of x, and into Eq. (2) gives
kAejwO
t
T 21+ 0W
/)602
( 0 (10)
T=(3) 2_____2
_ F 2 ' -2 +-,+)2
-mA0) 2
eij t + kAe j wt 0)11 22 C11 1

Letting k 1/m,I W1 and simplifying Eq. (3) Transmissibility versus frequency is plotted
gives in Fig. 4. There are two resonances in this
curve. Between them the transmissibility falls
T off at 40 db/decade as it did in the single-stage
I - \2 (4) system. Above the second resonance, however,
the transmissibility decreases at 80 db/decade.

228

Y4
 1000 - 60

1001 - 40

10- - 20

TRANSMISSIBILITY 10TRANSMISSIBILITY
FOUT/ FIN IN DECIBELS

SLOPE=0bdc
.01 _40

1.001L 6

.1 .2 .5 1.0 2.0 5.0 10.0

/w llFREQUENCY RATIO

Fig. 2. Transmissibili:." of one-stage mounting system

F sin wot It is here that the great value of the two-stage

.system is felt.

x1 DESIGN OF A TWO-STAGE
Fig. 3. Two-stage KI
mounting system ISOLATION SYSTEM
As mentioned in the introduction, four large
vx2 machines were to be isolated. It was decided to
K2 mount them all on a common subbase. Figure 5
is a schematic'of the machines on the subbase.

1000 - 60

100 - 40

10 - 20

TRANSMISSIBILITY= 1TRANSMISSIBILITY
FOUT/ FIN IN DECIBELS

.1 SLOPE: 40db/dec-2
.I -40

.01
.001 __60
SLOPE :80db/dec
.0001-8
1. .2 .5 1.0 2.0 5.0 10.0

Fig. 4. Transmissibility of two-stage mounting system

229 ~I
 !SSDG
, 3. o22/27 = Vk 27m 2 = 5 cps.
By use of these steps, the denominator of Eq.
PROPULSION
PMAIN II II3 1II MAIN
PROPULSION (10) can be solved for the two-system reso-
DG nances which can be plotted as a function of
m2/m . This is done in Fig. 6.
SUBBASE

SSDG
-80,000 Ib
M I=
CONDITIONS
M S K I=25,0001b/IN.
Mi/2 "=2cps
10 - 22 /2,rz5cps
0w
5
Fig. 5. Proposed subbase and main 2 FIRST
machinery configuration I RESONANT
MASS FREQUENCY
RATIO .5
M2 /M1 SECOND
The ship's service diesel generators (SSDG) .2 RESONANT
are mounted in line, fore and aft; while the .1 REQUENCY
larger, main propulsion diesel generators .05
(MPDG) are mounted outboard of them. Table 1 .02
gives the weights and fundamental rotational
frequencies.
.01 /
TABLE 1
S I I I I I I I I I

No. Fundamental J .2 .5 1 2 5 10 20 500

No. Weight/Unit Rotational RESONANT FREQUENCY (cps)
NameUnits
ts (lb) Frequency Fig. 6. Dependence of resonant
(cps) frequency on mass ratio

SSDG 2 17,000 20

MPDG 2 80,000 12 In choosing the optimum mass ratio for our

application, two major factors had to be consid-
ered. First, the system's total attenuation is
SUBBASE MASS affected most severely by the position of the
second resonance. This resonance should be
For the good of the ship, minimizing the as close to 5 cps as possible. Second, the sta-
weight of the subbase is of.prime importance. bility of the system is largely determined by the
Unfortunately, this is in direct conflict with the lower resonance. The lower this resonance is,
optimization of the attenuation and stability the more unstable the total system becomes.
characteristics of the system. As will be Therefore, the lower resonance should be as
shown, a low subbase weight, or more cor- close to 2 cps as possible. Holding these two
rectly, low ratio of subbase weight to machine considerations in mind and remembering that
weight, raises the second natural frequency to the subbase weight must be minimized, we chose
a prohibitively high value and lowers the first a mass ratio of 1 for the following reasons:
one to too low a value. 1. At mass ratios lower than 1, the second
It was desirable to have the two vertical resonance increases quite rapidly. This lowers
natural frequencies equal to 2 and 5 cps. The the attenuation of the system because it pushes
following can then be stated (Fig. 5): this resonance closer to the machine rotational
speed (12 cps).
1. The machine weight is fixed at 80,000 lb.
2. At mass ratios lower than 1, the first reso-
2. The machine mounts are to be such that nances decrease quite rapidly. This means that
NI1/27, = 1/m = 2 cps, therefore, k, = 25,000 the system is less stable and, therefore, more
lb7in, susceptible to injury due to rolling and slamming.

230
t

* 3., At mass ratios higher than 1, neither of 104

the resonances get much closer to their asymp-
totic values of 2 and 5 cps. Therefore, any gain
from a stability or attenuation standpoint would 103
be marginal and not worth the increased weight.
MECHANICAL 2
Therefore, the subbase will weigh as much IMPEDANCE, 10
as the total weight of all the machines it sup- IN/SEC - ACTUAL IMPEDANCE
ports. This figure is approximately 100 tons. 0 - APPROXIMATE IMPEDANCE

TRANSMISSIBILITY OF ACTUAL I10

SYSTEM FREQUENCY, cps

Earlier in the paper, the equation govern- Fig. 8. Impedance of stiffener in

ing transmissibility of the most fundamental destroyer hull in radial direction
two-stage system was found. The information
that follows shows the transmissibility's de-
pendence on mount damping, hull impedance, Damping was considered in the two sets of
etc. A new, more complex mathematical model mounts. The amount of damping was set at ap-
of the system was made and is shown in Fig. 7. proximately 5 percent of critical. This is a
typical figure for rubber mounts.

A * The model was programmed on Electric

F sin wt Boat Division's Pace TR-48 Analog Computer.
Automatic repetitive operation and a large
amount of on-line signal processing equipment
Mwas utilized to plot automatically the curves
X, displayed in Figs. 9 through 13. This ancillary
K1 C1 equipment proved very useful in making the
Fig. 7. Complex two- complete parametric study presented in these
stage mounting system M? graphs.
X2
K2 C2 The values of the initial system parameters
were as follows:
K3 C 33
tl c 2 =200 lb-sec/in.;

c= 1500 lb-sec/in.;
The hull impedance is represented as ak- =
spring (k 3 ) and dashpot (C) connected in par- k2 500,000 lb/in.;
allel. This is obviously a very crude model of
the actual hull. There is, however, good reason
for this choice. Figure 8 is a plot of the imped-
1 ' ance of a destroyer hull driving into a stiffener. 518 lb-sec 2 /in., or 200,000 lb;
Superimposed on this curve is our simplificd
system. One can see that this is a valid first- m1 = 207 lb-sec 2/in., or 80,000 lb;
order approach to the real impedance; further-
more, since the actual hull impedance would be c3 15,000 lb-sec/in.; and
extremely impractical to model, it may be the k = 20,000,000 lb/in.
best approach. 3

Another point that should be brought out is Of these eight parameters only the first five
1 that the values of k and c obtained from Fig. 8 were varied since the machine weight and hull
have to be increased since this curve only rep- impedance were assumed to be constant.
resents the impedance of one point, while our
model has to have the total impedance into which Figure 9 shows the effect of various values
the subbase mounts look. of c 2 , the damping in the subbase mounts. As

291
ii

+30 r +30

+20 +20
.+N0-

+1 !+10
C/4K2/4
C2/2 K2/2

m
-j0 ~~C
- 1500Kr515
STANDARD STANDARD

"" CURVE
2C -j I CURVE
-K
2

C2o
0-20 4C2 4 -20

zz
!2-20 4
a 20

-30- I I I I- -30

-40 1-40

.2 4 .8 12 2.0 4.0 80 120 200 .2 .4 .8 1.2 2.0 4.0 8.0 Mao 20.0
FREQUENCY CYCLES/SECOND FREQUENCY CYCLES/SECOND

Fig. 9. Transmissibility vs driving Fig. 11. Transmissibility vs driving

frequency for various values r.f c 2 frequency- for various values of k 2

+30 +30[
C/2 2KI 4KI

+20 C1= 200 +20- K =25x10 3

STANDARD STANDARD
CURVE
CURVE
K/
C1 0
K/
u)+10 4C I 0+. K 1/4
w
CO 8CI
Uw
bJ 0

, z -10 >.
in 10-

5 z
In 4
Z-20 -20 -
I-

-30 -30-

-40 -40

.2 A .8 1.2 2.0 4.0 80 12.0 200 .2. .4 .8 12 20 4.0 8.0 12.0 20.0
FREQUENCY CYCLES/SECOND FREQUENCY CYCLES/SECOND

Fig. 10. Transmissibility vs driving Fig. 12. Transmissibility vs driving

frequency for various values of c 1 frequency for various values of k,

232

*.3 .

/ ,JP'~
 +30 d ecoupled because a change in k, shifts the
first resonances but not the second, and vice-
versa. In Fig. 12, the curves for 2k 1 and 4k 1
+20 -4M? are clipped at the first resonant peak due to
4M 2 limitations in the dynamic range of the computer.

+10
+ Figure 13 shows the effect of varying the
U) subbase mass m2 . This, as expected, has the
opposite effect on transmissibility that varying
t5 8 k 2 had. That is, an increase in m2 decreases
o_- , STANDARD
CURVE the second resonance while a decre1se in 2

increases this resonance.

~_j -10 -to /2
U/)
U) Me4 SYSTEM NATURAL FREQUENCIES
z2o In the foregoing section, the force trans-
missibility has been found for two masses
translating in the vertical direction. it is as-
- sumed that other transmitted forces or mo-
ments would be attenuated inthe same manner
regardless of direction, rotation, or any com-
-4o- bination of these motions. The only restriction
is that the rigid body natural frequencies of the
.2 .A .8 12 20 4.0 8.0 IZO 200 system have to be appreciably below the lowest
FREQUENCY CYCLES/SECOND forcing frequency (12 cps). It was, therefore,
decided to liniit the highest natural frequency
Fig. 13. Transmissibility vs driving to 6 cps.
frequency for various values of m2
A mathematical model of the system was
then made. To simplify closing the mount stiff-
expected, c 2 has very little effect on the height nesses for the model, the five components were
of the first resonant peak. It does have a much assumed to be decoupled. The mounts used
more pronounced effect on the second resonance. were two-cycle mounts under the MPDG's,
This resonance would be practically eliminated three-cycle mounts under the SSDG's, and five-
if 20 percent of critical damping (4c 2 ) could be cycle mounts under the subbase. Other assump-
incorporated in these mounts. It is more re- tions made are listed below:
markable that the attenuation above the second
resonance is not impaired by the added damping, 1. The damping in the mounts can be ig-
J- as it would be in a single-stage system. nored in determining natural frequencies;

Figure 10 shows the effect of various val- 2. The stiffness of mounts in the horizon-
ues of c 1 , the damping in the machine mounts. tal plane is equal to the stiffness in the vertical
As expected, the first resonance height is al- planes; and
tered while that of the second resonance re-
r mains fairly constant. Unlike Fig. 9, the slope 3. The machines and subbase act as rigid
of the curve
creased with above the second
increased resonance is de-
damping. This decrease, masses in the frequency range of interest (less
than 10 cps).
however, is not as large as would be expected.
j Therefore, it can be concluded from Figs. 9 and The system was analyzed in two steps.
10 that damping in a two-stage system does not First, the flexibility of each component's mount-
have as adverse an effect on transmissibility ing system was individually determined. This
above resonance as it does in the one-stage enabled us to reduce the system shown in Fig.
system. This is especially true when the sec- 14a to the one shown in 14b. When the five
ond resonance is damped. components were thus treated, a mathematical
model of the whole system was made and is
Figures 11 and 12 show the effect of vary- shown in Fig. 15. Each spring now represents
ing the two-mount stiffnesses. The transmissi- the total flexibility of one machine's mounting
bility at high frequencies is more sensitive to a system and, therefore, has stiffness in all three
change in kI (Fig. 11) than k 2 (Fig. 12). It can directions and three rotations. Eachmasshassix
also be seen that the systems are effectively degrees of freedom. The 30 natural frequencies

233

z _7
 j L- _ -

were determined using Electric Boat's standard SUBBASE DESIGN

structures program. These frequencies, given
in Table 2, show that the highest natural fre- A subbase design is shown in Fig. 16. For
quency is 5.9 cps. This is considered acceptable this subbase to function properly, it must act
for effective transmissibility, as a rigid mass. Since this is physically im-
possible, an effort was made to maximize its
first mode. The highest value we could obtain
for the first natural frequency was 100 cps.
To insure that this or other resonant peaks
would not be too high, damping will be incor-
porated in the design of the subbase. Riveted
construction will be used wherever possible,
and all plating will be coated with a viscoelastic
damping treatment.

(o1 (b)

of component 2'5"
Fig. 14. Mathematical modelj
28-O

SSDG

MAIN MAIN
DIESEL DIESEL
Fig. 14. Subbase
SSDG

Fig. 15. Mathematical model CONCLUSIONS

of system A two-stage system can be far superior to
a one-stage system from the standpoint of force '1
isolation and rigidity, because the transmissi-
TABLE 2 bility of a two-stage system above the second
Natural Frequencies of System resonance is proportional to l/o 4 while in a
one-stage system the maximum transmissibility
Mode Frequencies Frequencies is proportional to 1/W2. Therefore, a rigid two-
(cps) Mode (cps) stage system could be used to replace a flimsy
one-stage system and still give superior isola-
1 1.23 16 2.04 tion effectiveness. It should be remembered,
2 1.28 17 2.51 however, that a two-stage design contributes a
3 1.33 18 2.65 relatively large increase in weight to a system
4 1.37 19 2.71 and, correspondingly, a large increase in cost.
5 1.52 20 2.78
6 1.52 21 2.88
7 1.61 22 3.05 ACKNOWLEDGMENTS
8 1.62 23 3.17
9 1.77 24 3.29
10 1.77 25 3.90 The author gratefully acknowledges the
11 1.79 26 4.27 assistance given him by personnel in Electric
12 1.82 27 5.51 Boat's Sound and Vibration Section, Applied
13 1.86 28 5.72 Mechanics Section, Information Processing
14 1.86 29 5.82 Section, and Creative Design Section. He is
15 2.03 30 5.90 especially grateful to T. A. Champlin, Head of
the Industrial Acoustics Group.

234

I4
 MODELING OF SPACECRAFT FOR
LOW-FREQUENCY NOISE REDUCTION
Richard H. Lyon, Robert E. Apfel, and Charles W. Dietrich
Bolt Beranek and Newman, Inc.
Cambridge, Massachusetts

The acoustic isolation or "noise reduction" (NR) of spacecraft struc-

tures over most of the frequency range can be computed and/or meas-
ured by methods that can be termed "conventional." The increased
interest in noise below 50 cps requires that extensions to these con-
ventional methods be made, Loth in analysis and experiment.
Conventional theory distinguishes between "mass law," or forced re-
sponse, and reradiation and transmission of sound by resonant motion.
The same considerations are appropriated-to low-frequency NR. At
low frequencies, the forced motion is stiffness controlled and derivable
from "pressure vessel" analysis. The mechanical and acoustical res-
onances of the system modify this stiffness-controlled NR.

INTRODUCTION of structural response and sound transmission.

For these and other reasons, it is desirable to
There is an increasing interest in the low- have models available that display the same im-
frequency sound transmission behavior of portant acoustical and mechanical patterns of
spacecraft structures, because the larger behavior that the "real thing" does.
~boosters that are currently under development
create an intense sound field with major com- If the model is to be significantly simpler
ponents in the low-frequency regime below 50 and less expensive than the real item, then it is
to 100 cps (1). In addition, as these vehicles necessary to know what the important features
pass through the transonic and Max Q portions of structural and acoustical behavior are in the
of their trajectory, they will experience addi- actual spacecraft, insofar as low-frequency NR
tional large-amplitude low-frequency load- is concerned. This is so because we want not
ings (2). It is known that an excessive low- only to model the correct dynamics of the real
V frequency acoustic and vibration environment system, but also to avoid introducing by our
can have undesirable psychoacoustic and physi- simplifications spurious effects that will reduce
ological effects on man (3,4). Consequently, the the utility of the experimental results on the
ability of a spacecraft to shield the human pas- model. An adequate theory of the behavior of
senger from these loads is of great concern, structures at very low frequencies and the way
that they transmit sound in this frequency re-
The development of adequate techniques gime is therefore required.
for predicting and measuring the low-frequency

[K
_A
noise reduction (NR) afforded by the spacecraft
structures forms a largely unexplored area of
acoustic technology. The new Low Frequency
We begin by reviewing quickly the more or
less classical approach to NR calculations in
building structures. Although we can make very
Noise Facility at Langley Research Center (5) few direct applications of these calculations,
has been designed to allow the direct generation there are at points certain similarities in the
and measurement of acoustic signals of varying construction of the ideas that we want to point
waveforms in the very low-frequency regime, out. Following this review, we discuss a pre-
and one of the tasks to which it can be put is the vious calculation of the NR provided by a rec-
measurement of low-frequency NR. Spacecraft tangular box over a large range of frequencies.
are expensive, however, and the schedule of We then discuss some features of the NR of axi-
their testing from final construction to their symmetric shells and indicate how these calcu-
deployment as a vehicle is very rigorous and lations" relate to the low-frequency NR of the
does not allow much time for explanatory studies rectangular box. Finally, we discuss an effect

235

~__ -_ _ ____ ____ ____ ___

of
space "double on
the vehicle its acoustical
wall" behavior.
construction of a current ical
fromform
the form of transmission
for the the transmission in theA re-
loss,loss. typ-
gion where the calculation by Eq. (3) is appro-

priate, is shown in Fig. 1 (7). The TL curve

REVIEW OF "CLASSICAL" shown in Fig. 1 is fairly typical of a broad group
NR CALCULATIONS of wall constructions. The most outstanding
features are a notch in the curve at the critical
In architectural acoustics, the NR provided frequency (the frequency at which the bending
by a structural wall depends on the so-called
transmission loss of the wall and the acoustic
absorption of the receiving space. The trans-
wavelength in the wall is equal to an acoustic
wavelength in the medium), and the low-
frequency asymptotic rise in TL at the rate of
p
mission loss of the wall is defined as (6, p. 324): 6 db/octave.

TL = 10 log inc/ltrans , (1) The mass law dependence of the curve

means that if a perfectly limp curtain were
where rlin c is the power incident on the wall hung in place of the wall and if the curtain has
and n is the power transmitted into de the same mass/unit area as the wall, then it
receiving room. The absorption properties of would provide the same NR as the actual wall.
the receiving room are defined through its At first glance, this is rather surprising, be-
"room constant" R, which is given by (6, p. 312): cause most walls have many resonances in this
frequency region and one would expect them to
R Ata( - a)- (2) increase the vibration levels of the wall to some
value higher than that of the limp curtain. In
At is Inendtoafact, if one measures
wall during vibration levels
such an experiment, the
he willonfind
where
wceiro a the asta erae aboptore
ceiving room and oa is its average absorption that the vibration levels are usually higher than
coefficient. In terms of these parameters and "limp wall" nonresonant vibration levels.
the area of the transmitting wall A,, the NR,
defined as the difference in sound pressure The reason that the transmission loss is
level in a reverberant field exterior to the room governed by the lower amplitude "forced wave-
and that within the receiving room, is given by nonresonant" portion of the vibration field is
(6, p.327): that its radiation efficiency is much higher than
that of the resonant vibrations. Thus, the prod-
= TL + 10 log R/AW. (3) ,ct of velocity amplitude and radiation efficiency
is hZigher for the nonresonant transmission.
By putting the equation in this form, an at- iParenthetically,
ge o h it orsnn rnmsin
tputi ngde te separateo thgeoericapro-
s should be noted that this is
not always the case, and in some instances, the
erties of the wall and the receiving space from resonant vibrations may contribute to the sound
the intrinsic properties of the wall. One can transmission. This, in fact, is indicated in the
find the values of the transmission loss for trartsof Th curve, is flucatin
various wall constructions in standard refer- around the mass law valuewre due to modal
ence books (6, p. 329); these values, combined rond
resonances the walue
of the wall.
with geometrical and acoustical information on
the receiving space, allow one to calculate the
expected NR. NR AT LOW FREQUENCIES

This argument assumes that the acoustic If we want to predict the NR of a spacecraft
receiving space and the transmitting wall struc- whose typical dimension is, for example, of the
ture are both several acoustic wavelengths in order of 10 ft, then the traditional approach as
size. Since these calculations are usually ap- described in the previous section will be appro-.
plied to predict NR over the range from a few priate for frequencies above a few hundred cps.
hundred to a few thousand cps, such assump- If we want to know the NR, i.e., the difference in
tions a:-e appropriate. Also, at a few hundred sound pressure level within and outside the
cps, most wall structures have a flexural wave- spacecraft in a frequency range from a few cps
length that is smaller than the acoustic wave- to approximately 50 cps, then clearly we must
length (i.e., they are below their critical fre- devise new techniques for its calculatidn. A
quency), so the wall is also several bending clue to the procedure we might take is afforded
wavelengths in size. by a previous calculation of the NR of a small
box (8). The box was assumed to have a single
Since the room constant changes rather flexible wall and five completely rigid walls.
slowly with frequency, the frequency dependence At higher frequencies, above the first few in-
of the noise reduction can usually be inferred ternal acoustic resonances of the box, one could

236

~ ~ - -

I V
 - REGION I -- *-REGION H REGION I.
EXTENSION OF
MASS LAW
SMALL DAMPING
........ -
a .. MEDIUM DAMPING I

.. z DAPIN ";'-LARE
0 6 -kA, * "

0 RESONANCES t"

STIFFN ESS . .. . MASS

u CONTROLLED ::P" * CONTROLLED I
-'-. :AI

fc

i Fig. 1.Typical form of TL for structural wall

texpect the traditional TL approach, as outlined the high-frequency TL/NR calculations of Fig. 1 '
in the previous section, to be appropriate. Be- will not be of assistance to us. As far as mod-

low the frequency of the first acoustic reso- eling is concerned, this means that we should
nance, the interior volume of the box becomes not have to pay particular attention to the ab-
a compliance, but the flexible panel may con- sorptive properties of the interior space, since
tinue to have resonant modes of vibration. The we are below the acoustic resonances of the
vibration of some of the panel modes produces contained volume, and that volume is behaving
a volume displacement into the box that results as an acoustic compliance.
in increased internal pressure and a loss of NR.
Finally, at frequencies below all structural res-
onances of the flexible panel and acoustic reso- EFFECTS OF AXISYMMETRIC
nances of the contained volume, both the panel GEOMETRY
and the volume act as acoustic stiffnesses, and
the NR achieved can be expressed in terms of Many of the space vehicles that have been
r their relative compliance. The NR curve com- designed have an axisymmetric geometry due
~that
puted for the box is shown in Fig. 2. to thecan
inherent high strength-to-weight
be achieved ratio
with this shape. Let us

As an example, let us consider the axisym- attach a coordinate system s,w to the shell, as
metric spacecraft shown in Fig. 3. This space- -shown in Fig. 3. The azimuthal angle is q5 and,
craft is made of sandwich panels having a core as we have noted, we assume that the geometry
j thickness of 3/5 in. with two facings of 8 mils is invariant to rotations through this angle. The
each. A typical overall dimension of the vehicle coordinate s is the distance along a line of con-
is 10 ft. With these parameters we can esti- stant 0 from the lower pole of the shell to the
mate that the first flexural resonance of the upper pole. Due to the symmetry and periodic-
skin will occur in the neighborhood of 5 cps and ity requirements, we expect that displacements
that the first internal acoustic resonance will of the shell will have the dependence cos me,
occur at about 50 cps. The critical frequency sin mo. If m= 0, then motions of this type will
of the sandwich shell is 500 cps. not have a net volume displacement. For m= 0,
there will be a net volume displacement and a
The ring frequency, i.e., the frequency at consequent contribution to the internal acoustic
which the major part of the shell will lose the pressure. The m= 0 "ring resonance" occurs,
stiffening effects of curvature, is about 7 kc. however, near the ring frequency of the shell
For a cylinder, the circumference becomes a which, as was pointed out above, occurs at about
longitudinal wavelength at this frequency. 7 kc. The lower order modes, therefore, that
Clearly, if we are interested in frequencies up resonate in the 5- to 50-cps range will not have
to about 50 cps, we would be concerned with a net volume pumping contribution. This situa-
the two lower frequency regions of Fig. 2 and tion differs markedly from the rectangular box
237
/I
'2 1-,4
3
 60 1 i I I I

LOW y MEDIUM I HIGH

FREQUENCIES fFREQUENCIES %FREQUENCIES

so I 1-
-A

I
HYPOTHETICAL
*

'z.
-"OSCILLATIONS
l' ti
V

z MULTIMODEN

a 40
0

/ FORCED WAVE NR
-:W__ FREE WAVE NR

SIFFESS __ _ _

CONTROLLED NR

\-SINGLE MODE NR (8% BANDS)

I: j
I- I I -I I II-I I I I I
32 50 so 125 200 320 500 8O 1250 2000 3200 5000 8000 12500
25 40 63 100 160 250 400 630 00 1600 2500 4000 6300 10000
THIRD OCTAVE BAND CENTER FREQUENCY IN CYCLES PER SECOND

Fig. 2. Theoretical NR for supported 6-in. x. 12-in. x 1/16-in.

aluminum panel covering rigid 6-in. deep enclosure

3Q00

t =0.008"
(1) h=0.600"

=
I t 0.008"
(2) h=0,51"
hO1

----4W S h = 2.061"
h

Fig. 3. Major structural sections of outer shell

(stainless steel sandwich)

238
 calculatiQn described earlier. We can expect, whether this model is adequate and what effects
therefore, that the very low-frequency NR cal- the slight structural irregularities always pres-
culation will have its range of validity extended ent might have on NR.
to higher frequencies, at least up to the first
acoustic resonance of the interior volume.
Consequently, in modeling the structure, it will EXPERIMENTAL TESTS OF
be of primary importance to model its quasi- SHELL 'R
static compliance and to test experimentally
whether the model in fact appears to be behav- The steel shell shown in Fig. 3 is similar
ing as our theoretical notions suggest that it to the one used as the outer shell of the Apollo
should. Command Module. The Command Module has
-i an inner shell with an upper hatch cylinder that
We have a situation, therefore, that is rein- is approximately 30 in. in diameter and slightly
iniscent of the acoustic behavior of building longer than this in height. It is of aluminum
wals. In a frequency region where there are sandwich construction with a core thickness of
many resonant modes of vibration, the sound 1/4 in. and two facing sheets of 10 mils each.
transmission is governed by nonresonant mo- As a test of some of the ideas put forward in
tion. The main difference is that the building the previous sections, we have studied the NR
wall NR behavior is mass controlled, whereas of a 1/4-scale model of this hatch cylinder.
the low-frequency NR of the space vehicle is The model is shown in Fig. 4 in its test cham-
stiffness controlled. ber. It is approximately 7 in. in diameter and
7 in. long, and has an aluminum skin thickness
The acoustic compliance of the shell can be of 5 mils. The two ends of the cylinder were
modeled by using a homogeneous sheet of metal fitted with 3/8-in. thick aluminum plates with a
if we know whether the shell is expected to re- hole in one of the plates, so that the microphone
sist deformation primarily by in-plane mem- can be inserted within the cylinder.
brane stresses or whether deformation is con-
trolled by its flexural rigidity. One way of
evaluating this is to assume that the shell de-
formation is completely membrane controlled
and then to apply flexural displacements at sup-
porting edges to satisfy the boundary conditions.
B If the required flexural contribution propagates
itself throughout the area of the shell, then at
the very least,
considered thecompliance.
in the flexural contribution must be
The distancet
tht the flexural effects propagate is given by (9):

) 1/ 2
(atC ,(4)

where a is the radius of curvature of the shell,

and t, is the core thickness. For the shell of
Fig. 3, this length is approximately 6 in. We
conclude that the compliance is membrane con-
trolled when the shell does not have appreciable
structural discontinuities, such as longerons
and other attachments. If the shell has flat seg-
ments, however, we can expect these sections
to have an acoustic compliance that is con-
trolled by their flexural rigidity.

A membrane compliance model of the steel

shell in Fig. 3 would be a shell of the same
overall dimensions but with a skin thickness of
16 mils. If the shell were made of aluminum, it
would be 48 mils thick for the same membrane
rigidity. It is clear that these much thinner '"
shells would have many more flexural modes in -sea
the low-frequency range than the sandwich
shells that they model. It is essential, there- Fig. 4. Scale model in
fore, to carry out experiments that will show test chamber

239
 The theoretical values of the shell compli- acoustic- space between the two shells. In the
ance C. and the compliance of the contained actual vehicle, this space is pretty well filled
volume cv are: with thermal insulation and supporting gear for
the mission, so acoustical resonances may not
Cs = 5 x cu in./psi play a part in its noise transmission properties.
and (5) On the other hand, a model of such a system
Cv = 15 cu in./psi, may display resonances, and if so, we may wish
to know how much sound transmission these
The noise reduction of this combination is contribute and what we should' do to avoid them.
given by
Unusually low-frequency acoustic reson-
NR 20 log (Cv + Cs) Cs 50 db. (6) ances of the interspace between the two shells
is made possible because of the long path length
Thus, the theoretical NR, assuming rigid from one pole of this space to the other, as
end pieces, is 50 db, not a bad NR to be achieved shown in Fig. 6. For the Command Module, this
by a piece of 5-mil aluminum foil. This NR is, distance is approximately 18 ft. Theoretical
in fact, so great that it is "short-circuited" by analysis of the acoustics of such spaces indi-
the 3/8-in. aluminum end plates. If we include cates that we should expect an acoustic reso-
their compliance in the calculation, we predict nance when this path length is of the order of
an NR of approximately 44 db. 1/2 wavelength, about 32 cps. A detailed anal-
ysis, using the Rayleigh-R.itz estimation proce-
If all dimensions of a structure are scaled dure, gives a more precise value of this fre-
by a factor of 1/4, the modeled structure will quency at approximately 40 cps.
have the same dynamics as the full-sized struc-
ture when the frequency is 4 times as great as An experimental study of the existence of
in the full-scale experiment. Thus, a range of such resonances was made in a 1/10-scale
5 to 50 cps in full scale becomes a range of 20 model of this space with nearly rigid bounda-
to 200 cps in a 1/4-scale model study. The NR ries. The experimental value of the resonance
measured for the test cylinder is shown in frequency, scaled to full scale, was about 32 cps.
Fig. 5. We note that the achieved NR does not
exceed approximately 41 db.
CONCLUSIONS
Despite the lack of precise agreement be-
tween the calculation and the experimental re- Theoretical and experimental analyses
sults, the experiment does show the general suggest that the major source of sound trans-
adequacy of our model in two ways. First, the mission at very low frequencies is the quasi-
NR curve over the range of frequencies is rather static volume displacement of the shell. This
smooth and does not display the existence of is particularly true for axisymmetric shells,
strong resonant modal contributions to the NR, since the great majority of modes resonating at
despite the fact that our test cylinder has sev- low frequencies are not expected to have sig-
eral hundred modes that resonate below 200 nificant net volume displacements. The result
cps. The actual hatch structure, of course, of this is the maintenance of relatively large
would not have as many modes, since it is made values of NR up to 50 cps or so.
of sandwich panels. Second, even tknugh the
value of 40 db is less than our predicted NR, it This simple quasi-static picture of shell
is still many times what one would expect if the motion is complicated by the possibility of
shell were flexurally controlled. Nevertheless, acoustical resonances oi the interspace by
other effects that we have not accounted for double-shell structures. Some of these reso-.
must be present and cause the 3-db discrepancy. nances can be driven by the membrane deflec-
tion of the shell and may cause low NR at a few
selected frequencies. Operational space vehi-
EFFECT OF ACOUSTIC cles may avoid these effects, since' the inter-
RESONANCES ON NR space volume is filled with thermal insulation
and other equipment.
As we indicated previously, if the dimen-
sions of the space vehicle are not too great, the It should be emphasized that experimental
major part of the low-frequency behavior of this analyses of NR (or any other mechanical-
space will be controlled by its acoustic compli- acoustical problem) should be looked on as
ance. This is true in general, but the particular more than a test of theory. The intelligent use
double-shell configuration of the Command Mod- of an experimental facility can be very efiec-
ule allows the possibility of resonances in the tive in sorting out competing effects in vibration

240
 4-I

rns I

Sr~.~Iw. - 8-7i & t.'

1:~
7 w v----

~ J 4

W. $P6.

1o so _0 to o _0

OP112Z

5 ~:r4 Kirjuu Or"

ip5 d 5- 251 Ip 0 4

60 No -0 - --

- _ _0
- -2-0.0 .. i
OM_~~--i

H __Sir_

+is
Fi.5 RdtL/-cl oe

L 241
 !-,-

MEDIAN SURFACE r

AXIS OF SYMMETRY

Fig. 6. Sketch of Command Module interspace

and sound transmission. Experiment, used in problems in sound transmission that either one
conjunction with theoretical analysis, can salve alone cannot effectively or efficiently accomplish.

REFERENCES

1. F. M. Wiener, "Rocket Noise of Large 5. P. M. Edge and W. H. Mayes, "Some Re-

Space Vehicles," 4th ICA Congress, Copen- search Features and Capabilities of the
hagen, Denmark, Aug. 1962 Langley Low-Frequency Noise Facility,"
J. Acoust. Soc. Am. 37:1189(A) (1965)
2. H. H. Hubbard, D. A. Hilton, and W. H. 6. L. L. Beranek, Acoustics (McGraw-Hill,
Mayes, "Noise Considerations fGr Manned New York), 1954
Reentry Vehicles," NASA TN D-450, Sept.
1960 7. L. L. Beranek, "The Transmission and
Radiation of Acoustic Waves by Solid Struc-
3. D. E. Goldman and G. E. von Gierke, "Ef- tures," in Noise Reduction (L. L. Beranek,
fects of Shock and Vibration on Man," Ch. Ed.) (McGraw-Hill, New York), 1960, p. 287
44 in Shock and Vibration Handbook, Vol. 8. R. H. Lyon, "Noise Reduction of Rectangu-
I (McGraw-Hill, New York), 1961 lar Enclosures with One Flexible Wall," J.
Acoust. Soc. Am. 35:1791-1797 (1963)
4. N. L. Carter and K. D. Kryter, "Masking
of Pure Tones and Speech," J. Aud. Res. 9. S. Timoshenko, Theory of Plates and Shells
2(1):66.98 (Jan. 1962) (McGraw-Hill, New York) 1954, p. 399 ff.

242
 FLEXIBLE TOROIDAL SPRING CHARACTERISTICS*
C. V. David
General Atomic Division of General Dynamics Corporation
Special Nuclear Effects Laboratory
' San Diego, California

Flexible toroidal springs have been investigated for an application in-

cluding highly impulsive loading both on and off axis, and the results
are reported. The present study was concerned with multitorus stacks
and concentric multistack assemblies that operate as a unit. This
arrangement, therefore, limits the torus mounting to clamping on the
circle of tangency between the upper and lower tori and the plane of
attachment. The torus spring can be submitted to four basic types of
loading: compression-tension, bending, shear, and torsion. The char-
acteristics of the spring are highly nonlinear, and exact solutions are
almost impossible to calculate, especially for large deformations. Ap-
proximate analytical solutions were derived for three modes of load-
4 ing: compression-tension, shear, and bending.

Ition The torus-system characteristics for these three modes of deforma-

have been determined. The cross-coupling effects between the
shear and bending modes and the structural characteristics of the
torus wall are taken into account.

Several small experimental torus spring systems have been constructed

and tested both statically and dynamically. -Also, the equations of mo-
tion, restoring forces, and moments corresponding to the dynamic test-
ing of a plate attached to a three-torus spring system were simulated
on an analog computer. The results obtained have been compared and
are found to be in good agreement.

INTRODUCTION
finally evolved into concentric columns of
The problem of storing energy mechanically stocked tori, pressurized and clamped at both
is growing more and more acute, since weight ends. This system could operate in either ten-
usually becomes more critical with the develop- sion or compression, as required.
-i ment of more advanced technologies. Large
improvements in performance are constantly To determine the dynamic response of such
sought, often in cases where the environmental a system, a considerable amount of effort was
conditions are more severe. For instance, the needed since such a spring system had never
need for a spring system that is light, extremely been studied in detail. The information required
flexible, without moving parts, and requires no was both analytical and technological. What
bulky connections arose several years ago. After were the restoring forces and moments devel-
a thoro gh investigation of various schemes, oped by the system when subjected to large de-
it became obvious that storing energy by com- formations? How irmch energy could be stored
pressing a gas inside a flexible-wall container per unit weight? Could such a system be de-
was the best answer to the problems of weight signed and built? What were its damping
and the severe shock loading conditions experi- characteristics ?
enced at the connection between the loading
mechanism and the spring. Maximum use of Answers to these questions were sought
the volume available and uniform distribution analytically and experimentally almost simul-
of the load were also imperative. The design taneously. The results obtained by both methods

This research was sponsored by the Air Force Weapons Laboratory, Research and Technology Di-
vision, Air Force Systems Command, Kirtland Air Force Base, N. M.

243
agree, and means are now available for cal- displacement AX of the load, AP P(X) -p
culating the response of toroidal springs to where P(x) = P0 x f(AX) is the absolute pressure
static or dynamic loading that are accurate inside the torus in the neutral position and p is
enough that such a spring can be designed con- the external pressure. f(Ax) corresponds to the
fidently, and the response of an existing system compression taking place inside the torus when
to given loads can also be calculated. it is deformed, since its volume then decreases.
The ma;imum theoretical deformation of the
torus system occurs when either the volume has
SYSTEM DESCRIPTION AND become nil and reactions are infinite or a phys-
DISCUSSION ical stop has been reached.

A torus spring can be used in a number of For the three cases of Fig. 1, this occurs
ways according to what characteristics and when
what type of mounting are desired. Figure 1
shows three ways a single torus could be uti-
lized. In all cases, four modes of loading are AX
possible, separately or combined: compression-.d =--\ / - 1 - 1.21,
tension, shear, bending, and torsion. In each
case, the types of reactions developed by the
torus are conditioned Ifby ithe
isaway the torus wall
ssumd,
a frsteither for direction.
c1 reaction istion
n for The reai then infinite.
in
deforms. If it is assumed, as a first approxi- For case 2,
mation, that the wall is infinitely flexible and
does not stretch along the cross section, these
reactions are proportional to the differential AX 1 0.571,
pressure (AP) across the torus wall. For a d (T -

I. SHEAR OF TORUS CROSS SECTION

-
7/91,1

2.ELONGATION AND FLATTENING D72

OF TORUS CROSS SECTION

AX c AXt

3. FLATTENING OR ELONGATION
OF TORUS CROSS SECTION

Fig. 1. Various uses of toroidal gas springs

244
 when the upper section of te torus is com- outside the tori is either constant or nil; in
pletely stretched out, for instance. The vertical other words, p, = P 2 = P 3 at all times. Assum-
reaction of the system is still finite, since the ing that the base is rigid and fixed, the loads
volume has not decreased much. Finally, are applied through a loading plate. The torus
internal volumes are sealed and practically no
_gas can flow in or out.
d d
When the loading plate position changes, the
for case 3 in compression and tension, respec- torus cross sections deform and restoring forces
tively. In both instances, the torus reaction is and moments are developed. Figure 3 presents
infinite. Cases 1 and 2 exhibit a symmetry in the simple case of axial plate displacement. In
o the loading mode, but case 3 does not. In all the compression case, the torus cross section
cases, each single torus section could be re- flattens out and the volume inside the torus de-
placed by a multitorus-multistack assembly if creases, causing P(X) to increase. The reac-
so desired. tion F(x) against the plate is then 277 (P(x) - p] cD.
If tension is applied to the torus cross section,
For the application investigated at General the walls assume a lens shape and the hoop ten-
Atomic, the use of toroidal springs was limited sion developed in the torus walls produces a re-
to case 3, which was the only case that was ex- action opposing F(x), with F(X) increasing with
tensively studied. Also, the torsion mode was ZX. If shear is applied to the torus stack in
not considered, since there was no foreseeable either tension or compression, the lack of sym-
loading in this mode. Figure 2 shows a typical metry of the deformed torus cross section
torus system assembly that can be considered creates a restoring side force. If a tilting mo-
as a model. It will be assumed that the pressure tion is superimposed on either of the other

X z
J 00 0 0

EXTENDED POSITION TYPICAL COMPRESSION,

SHEAR AND BENDING

Fig. 2. Toroidal gas spring operation

245

I'
 (X) I

AX

C--w
C -C

AX r

COMRESION

D
TENSION

Fig. 3. Torus cross-section deformation in

compression and tension

motion modes of the plate, it is easy to see how expressed in terms of displacements for given
the differential flattening or elongation of the applied loads and are as follows:
torus cross section will cause a restoring mo-
ment. F nd'
x = (X)
The object of this paper is to give means
of calculating these restoring forces and mo-
ments when d, and D are known for any
number of tori in series and/or para nyel
that
are inflated with a pressure P0 in the neutral
= _,
M
KF

+,)2
F(z)(nd
(n
)2) [
position and are made of a material of known
3
characteristics. z F(z)(nd')
-+
3
F(z)nd' -___N1()(n
_ _') (3)
(3)
KM Kv 2K I '
RESTORING FORCES AND where
MOMENTS
KF = wr(P(X) -p) d'DKNx, (4)
A study reported in Ref. (1) yielded the
basic equations defining the forces and the mo-
ments developed by a stack of tori. These are KM 8 - P) d'D 3 KNX,

246
 77i AM I

KV -d DKqt
2On
(6)
=
PoD 3 [2 + L
M (d)
(
7KAX(d

KNX
and nd' is the average deformed length of the
torus stack, n is the number of tori in one (8)
stack, and Krx and Kqt are stiffness coeffi-
cients, defined later. An additional contributing nF( z) [ ( .)
term due to the shearing stiffness was neglected, zn - D D j d
since it is negligible compared with Kqt, The
physical significance of most of the parameters
given in these equations is shown in Fig. 3. [3M()- Ad4
Di(z)d 37f6 K
KNX
Equations (1) through (6) are based on the
following simplifying assumptions:

1. The torus structure consists of a heli- +E sin 60 si 1 (9)

cally wound, geodesic fiber pattern.
2. Each individual torus is approximated where %ois the hoop stress in the torus wall
by a ring of short cylindrical sections. and L =nd (torus stack height in the neutral
position), after rearrangement of Kqt as shown
3. The deformation due to fiber stretching below. One must remember that AX > 0 in com-
is negligible compared with those caused by pression and AX < 0 in tension for the signs of
changes in torus cross section. Eqs. (8) and (9) to be correct.

4. Tilting and lateral motions are small Reference (1) gives

compared with axial deformations.
Kq
t
63cos 53
4Asin/(10'
5. Two ideal cases are considered where
the fiber helix angle, /3, is either variable, to
keep a true geodesic pattern, or constant. where A isthe torus cross section perimeter,
t is the length of a fiber between intersections
6. No circumferential constraint at the is the 4),
(Fig. E is
fiber the modulus
cross
elasticity, and A
section. ofThe number of fibers
"
polar attachment circles of the tori is assumed
to exist. per inch of torus circumference is then

As a first approximation, if P >> p and if = (

the torus cross section does not vary consider- -'Sin /3
ably, one can find an equivalent pressure P and
such that the value P(x) - p does not differ ap-
preciably from Po f(AX/d) for any practical
value of AX/d. This pressure Po will be re- 4n'EA sin 2/3 cos /8 (12)
ferred to as Po for simplification from now on, Kqt A (
keeping in mind that Po is actually slightly
higher than the gage pressure inside the torus but A = wd and n'A = t, if t is the equivalent
in the neutral position. In vacuum, p = 0 and wall thickness of the torus structural material.
Po p , of course. One can now express P(x) Then,
as Po f(AX/d), where f(AX/d) defines the poly-
tropic compression of the gas inside the torus 4Et sin 2/6 cos /3
system. Kqt = (13)

Replacing KF, KM, and Kv by their respec-

tive values in Eqs. (1) through (3) and rearrang- t is related to the hoop stress, %, in the torus
ing yields for a stack of n tori of diameter Di , wall in the neutral position by
X nJ FM,7 2tao Pod (14)
Xn PD. x /AX\ (7)
0 f ') KNX Therefore,

247

y z T
 iK

AFILAMENT

I ICONFIGURATION -
WOUND FILAMENT

CROSS SECTION)

2'cos B

II

TYPICAL FORCE DIAGRAM

Fig. 4. Elementary diagram of forces

developed by structural fibers

K 2EPO sofl 3
C ( 15 The torus internal pressure was factored as
qt . . . . shown in Eq. (9). The replacement of K t by
its value in Eqs. (6) and (3) yields Eq.
Kqt is then only a function of the mechanical ..... ..........
properties
-ria and s of the torus-wall structural mate- . t isth..osibe.o.efnea.igdiy .o
the three types of possible motions (X, e, and z)
rl, and in one plane for any torus stack of diameter D,
2E Cif one divides Eq. (7) by F(x), Eq. (8) by M()
Kqt - 2E sin'j cos . (16) and Eq. (9) by F(z). Then, one can finally
W010 write

For practical cases where D is independent x = n

PD X*
of AX/d and does not vary, both n' and t should - xc
F(x) OD
remain constant and

K 2E sin j6, sin 83cos 6 C=n 3

3
(19)
Kqt= s 1o cs . (17) M(0) OpD
P

248
 F- SZZC P nD Z")- ((20) V0 =-- d2 (25)
F(Z) P0 ' VX 4(2rc+77r 2)

where After substitution of Eq. (24) into Eq. (25) and

(21) rearrangement,
nf ) KNX (1
1- = 2 (26)
Vx1--
4 F Z)L I . (22
2 +K M(O) d (22)For a polytropic compression of the gas
inside, one has

"i and
andP 1
4S)[Mo
4 1-
(.)N[()'
2L IM(X
} ad (X) P0 v PO0 1-
A
(27)
ID( ~-. F(Z)D d

E sin A0 sin 6 cos/3 3f( [ ] (28)

For a given structural material, defined by

A, E, and %, and a given aspect ratio L/D, one
sees that for a fixed amount of axial compres- KNX is defined as dNx/d(AX) (1, Eq. (66)), where
sion or tension the torus-system rigidity coef- Nx is the axial loading on the torus per linear
ficients x,, o,, and zc are functions only of the unit of length along its attachment to the loading
torus-system size and configuration and the plane per unit of pressure inside the torus and
pressure inside the torus, since x*, 0, and z* is given by
are dimensionless. It is then possible to cal-
culate XC, 6c, and ZC very easily for any value
of AX/d if L/D and mhe)/F(z) are defined. In Cos ' 1
other words it is straightforward to calculate Nx .o (29)
the deformation of a given toroidal spring under 2" COos
)
_r

known loads. But because of the cross-coupling c os

between 0" and Z*, it is more difficult to de-
termine the reactions of the same system when
the deformations are known. One must define Substituting Eq. (24) into Eq. (29) yields
explicitly both M(,) and F(Z) as functions of the
deformations a and z. Cos
AXd (30)
=
It has been assumed that f(AX/d) and KNX
are independent of any other parameters. This
Nx -
[
d
' - 1) ( -
C(

L can be easily shown for both tension and com-

pression of the torus.
Sohwhere'
Nx is again a function of AX/d.

Compression In the case of a filament-wound torus, which

tis truly isotensoidal in nature, in the neutral
Assuming that the torus-wall stretching is position the value of 80 is well defined and equal
negligible and referring to Fig. 3, one has to 35.3 degrees. But when the torus cross sec-
tion deforms, the angle made by the filaments
77r = d -6X (24) changes, since they are free to "pantograph,"
2c AX ' and /3 becomes a function of AX/d, which makes
the differentiation of Eq. (30) awkward analyt-
and assuming that (D
theremains
torus volume varies as ically but easy to compute
the cross section constant), techniques. Therefore, forwith digital-computer
every value of 6X/d

249
and the corresponding value of /8, a value of Multiple-Torus Stack System
KNX can be obtained.
If concentric stacks of multilayer tori are
63is given in Eq. (51) of Ref. (1) as assembled together between two rigid planes,

,4c one has

Di
tan- + (31) Ft(X) = F(x) -X (37)
71r

Substituting Eq. (24) into Eq. (31) again yields M(9) = M = 0 (38)
n 1

n + d
2
(32)
Ft(Z) =XF(z)
Z*
PnO (39)

The problem is entirely defined for this case.

where 0" and Z* are functions of L/D i and
Tension M o)/F(z)Di as shown by Eqs. (29) and (30). One
should remember that this applies only if all
With the same assumptions as are made tori have the same cross section in the neutral
above (Fig. 3), one sees that position.

sin o = 2a (I + (33) Results

For every value of X/d, Eq. (33) can be solved adKNX in compression and tension for variable
and one gets a single meaningful value of ao. and constant /8 is given as a function of 6X/d in
Fig. 5. The variations of /8, f (6X/d), and a o
The same reasoning about the gas com- (when applicable) with AX/d for the compression
and tension cases are presented in Figs. 6 and
pression inside the torus system applies, and7,rsetvl.Fge8shwXSBcae
7, respectively. Figure 8 shows X*. Because
of the cross-coupling effects between e* and
2a, yz*, these are not given in this paper but are
"X sin 2a (34) available in Ref. 2, where L/Di'M()
, F()Di
2ao
"sd(1 4 and E are used as parameters.

ENERGY STORAGE
between /3 and a o
In this case, the rolationship
is given by Eq. (61) of Ref. (1) and The toroidal spring system described pre-
viously is best suited for axial loading, both in
1
1
si ) compression and in tension. The bending and
/3= tano n 80 (3) shear loads applied are usually incidental and
of much smaller magnitude than the compres-
sion and tension forces. Furthermore, because
Therefore, /3 again becomes a function of of the cross-coupling between the tilting and
6X/d, and a solution is always possible for every lateral motion modes, it would be impossible
value of AX/d between 1 and 77/2 - 1 if the torus to integrate the force and moment equations to
attachment to a flat plate driving the free end calculate the energy stored during the deforma-
of the torus system is assumed rigid. tion of the torus stack. This cross-coupling
effect between the axial motion mode and tilting
The value of Nx in this case (1) is and lateral modes was neglected and, therefore,
the integration of Eq. (7) is straightforward and

Nx
( Cos 80
2a o cos /3
(36)
accurate enough as a first approximation for
any values of 8 and z.
For a torus stack in compression, if one
7 s 81 uses the value KNX 77/2 one gets

250
 16

- CALCULATED (REF I
UNPUBLISHED TEST RESULTS
14 -* REF I TEST RESULTS

KNX
* EXPERIMENTAL POINTS
(FROM BENDING TESTS) j
12

10

8
IO
6

VARIABLE /

2 2
3

0-
0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6

-<- TENSION RELATIVE TORUS ZSX, (/aX/d) COMPRESSION -

Fig. 5. Typical variation of Kjx as a function of 6X/d

251
5

0
3

X1,'
Xi

0010.2 0.3 0.4 0.5 0.6 I

RELATIVE TORUS COMPRESSION,()

Fig. 6. Variation of 3 and f(AX/d) irt compression

252
 3.5 100

3
bJ-- s0ot0 n

U)
W

0, WC 0

_ 2.5 60

2 40

1,5 20

I 0
0 0.1 0.2 0.3 0.4 0.5
RELATIVE TORUS EXTENSION, L---

Fig. 7. Variation of A, ao and f(X/d) in tension

F(x -2nDif
2 o (6X\ X.
Xn) "
(40)
E(s) = "
22
Po D
JI -
~ \1v 1
1T,
(2

For each individual torus, using Eq. (28), one where E(S)is the energy stored. This equation
has can be rewritten as

F(X)
=
:. Pd
.1 o dD
[I -S)
(J()2]V (AX (41
4) E~s P d2 DE*
where Ec is a dimensionless function of AX/d.
since Xn. = nAX. The value of E, is shown in Fig. 9.
The integration of Eq. (41) between the For the tension case the equations express-
neutral position and a compression 6X for each ing KNX and f(AX/d) are too complex for an ana-
torus yields lytrcal integration to be attempted, and Eq. (7)

253

.513FF
 0.30 V-,

0.25 -
7.

0.20
VARIABLE /
x

/
0.15 - R o

0.10 -

0.05

0.6 0.5 0.4 0.3 0.2 0.1 0 0.1 0.2 0.3 0.4 0.5 0.6
TENIONRELATIVE TORUS A X, (A x /d)COPESN

Fig. 8. Variation of X* as a function of X/d

254
 -- 1.0 10 sufficies. F* for Compression and tension is
also plotted in Fig. 9 for convenience if one
0.8 does not wish to use the curves of Fig. 8.
Z These curves are sufficiently accurate for
0.6- 6 most practical engineering uses. Nevertheless,
M one should not forget that they are based on
(Es pod?
d DE, E several assumptions that cannot be grossly
(E___ __o c_ -) 4. violated indiscriminately.
/

/-0.2 2 EXPERIMENTAL RESULTS

TENSION 10:5 Models of toroidal gas springs have been
_____J
______ .0 ___ ___

0.6 0.4 0.2 0 0.2 0.2 AX/0.6 designed, built, and tested under both static and
COMPRESSION - dynamic conditions. Experimental results were
2.2 1 obtained and were compared with the character-
// istics calculated from Eqs. (25) through (39).
-F(xp dDF*) Satisfactory agreement was established between
4 -0.4 the two conditions, although not as good as one
F might wish. This is mainly due to the fact that
U it is impossible to manufacture and install torus
6 -. 6 - systems in the ideal way assumed for the ana-
lytical treatment. For instance, a discussion
0_ of 3 and of its significance might be useful at
8 . this time. As explained in the section on re-
storing forces and moments, 8 is the winding
10- 1.0 angle of the filament. When the torus is in com-
pression or tension, its cross section deforms,
Fig. 9. E* and F* as functions of AX/d but in all cases the free part of the torus wall
(unrestrained) assumes a circular shape, and,
if free to do so, the filaments will rearrange
was integrated numerically. A result identical themselves so that they transmit only pure ten-
in form to Eq. (42) can be obtained: sion, and ,8 will then adjust. At the limit of
compression, then, one has 3 = 45 degrees, and
E P 12 DE* (43) at the limit of tension 3 = 0 (see Eqs. (32) and
(S) o (43) (35)). But two constraints are usually present
in the system:
A plot of Et is also given in Fig. 9. We
can easily see how the problem of designing a 1. Shear reaction between the crossing fila-
torus spring for a given energy storage re- ments and between the filaments and the gas-
quirement can be solved. On the other hand, tight inner shell (for instance, an inner tube);
for a given system configuration and axial de- and
formation it is quite simple to calculate the
energy stored in the gas spring. For a stack 2. Shear reaction between the torus and
of n tori, E(x) should be multiplied by n. its attachments to the end plates used for the
compression and extension of the system.
Incidentally, Eq. (7) can also be written as
If one assumes that these end plates are
rigid (forcing the torus diameter, D, to remain
F(x) 0 dD f jy r'x . (44) constant), one can easily understand that in the
tension case, for example, 8i cannot become
Again, one can express this equation as zero at any time but actually remains almost
constant. Therefore, practically, it would be
F Po dDF* (45) more realistic to assume that 6 does not vary
(x) 0 in tension. In compression the difference be-
tween the two assumptions is less important,
where F* is a function of AX/d. The number of and the actual characteristic is probably in be-
tori in a stack is immaterial in this case, since tween the two limits. Also, other arrangements
they all develop the same reaction, and consid- of the structural filaments should be considered,
eration of an individual torus of the stack such as knitting. In most practical cases, the

255
 structural material must be bonded to a flexible Curve 1 corresponds to the nonstretchable
.shell to stabilize the filament mesh pattern. All wall case with a polytropic compression coeffi-
these-factors tend to make the torus wall over- cient y of 1.3 (see Eqs. (28) and (34)). Curve 2
all stkucturebihave quite differently from the represents experimental data.obtained with the
idealized configuration assumed in the section crudely built three-torus stack. Curve 3 was
'on restoring forces and moments. obtained by Astro Research Corporation (1)
with a single torus. Finally, the values of KNX
The lateral stiffness is also a function of calculated with /3 = /3
o and 3 variable are also '7
the tructural filament characteristics, working given. Two typical values of KNX are also in-
-stress level,, and modulus of elasticity. There- dicated by a point. These two values were corn-
-fore, three typical materials were investigated- puted from a single-torus bending and rolling
nylon, Dacron, and steel wire-to determine test reported in Ref. (1).
whether this effect was significant. Table 1
gives-values of-a 0 and E that were used for the The values of KNX are scattered- appreci-
calculations. ably in the range of 6X/d investigated experi-
mentally but are within the scatterbandwidth
TABLE 1 determined by the two assumptions /3 =.,6 and
o and E Values for Lateral /3 variable, especially for compression. In ten-
Stiffness Calculations sion, all the experimental values of KNx depart
T 10 from the results given in Eq. (36). Increasing
Material o (10 3 psi) 6 psi) vE(10 of AX/d correspond to increasing values
Values
of the tension developed at the torus-system
Nylon 20 1 attachments. Such attachments are quite diffi-
cult to make rigid and nonstretchable without
Dacron 20- 5 affecting the torus behavior. This is not true
for the compression case, which could explain
Steel wire 100 30 why deformation of the attachment in tension
would decrease the torus axial stiffness, as is
observable in Fig. 5.
As shown in Ref. (2), the influence of the
structure is noticeable only for small values of The single value of KNX in compression,
L/D and also if M(O)/F(Z)D is small. This in- obtained from torus bending, also departs ap-
fluence is more pronounced for high values of preciably from the other experimental results. 4
AX/d. A small value of L/D corresponds to a This, too, can be explained by the lack of rigid-
very short torus stack and a small value of ity of the torus attachment. The effect is more
M()/F(z)D to a small amount of bending; there- pronounced for a single-torus system than for
fore, for a long and flexible stack of tori sub- a multitorus system. This would justify the
jected to an appreciable bending deformation lower values of KNX given by curve 3 of Fig. 5
one can ignore the nature of the structural ma- as compared with those given by curve 2.
terial of the torus wall. The curves of KNX in
tension and compression for constant and vari- Experimental results obtained with a single
able 3 given in Fig. 5 also point out that, as a torus in shear (lateral rigidity) are given else-
first approximation, one could ignore the in- where. The lack of rigidity of the torus attach-
fluence of /3. ment is particularly obvious in this case, and the
results presented are practically useless. They
During the investigation, two types of torus yield values of KNX several times lower than
systems were built: a three-torus stack made those theoretically calculated.
of nylon cloth wrapped and sewed around bicycle
inner tubes, and two single tor made of Dacron For practical purposes, itis suggested that
filaments wound and knitted and locked on a the two conditions that are best representative
thin-wall polyurethane inner tube. Both sys- of a practical torus system behavicr are for
tems were tested statically and dynamically to compression, /3 = constant = 3o , and for ten-
determine their response to known loads, sion, /3 variable, unless additional experimental
whether applied slowly or in an impulsive fash- results for larger 6X/d do not confirm the re-
ion. The test specimens, the testing procedures, sults already at hand.
and the test results are described at length
elsewhere. These results were analyzed and Finally, an attempt was made to simulate
reduced to a form that could be presented in the dynamic response of the three-torus system
Fig. 5. model on an analog computer with the use of a

256
 mathematical model embodying the torus re- the platp that could not result from plate-charge
storing force and moment equations derived shock wave interaction. The coefficients of the
earlier. The goal was to check whether the mathematical model equations, the amount of
cross-coupling between the bending and shear damping, and the initial conditions were calcu-
deformation modes was adequately expressed lated from the characteristics of the torus system
by Eqs. (29) and (30). Inthe experiments, the model and also from the test results (static and dy-
torus system was compressed by a plate at- namic). After some adjustment of the lateral stiff-
tached to it and accelerated by the explosion of ness coefficient, the curves of Fig. 10 were ob-
iRHE charge. In some cases the charge had tained. They compare well enough with the two
been placed off-center to study the behavior of experimental curves obtained elsewhere, espe-
this spring concept under off-axis loads. This cially since the latter were not reproducible to the
systematically caused a lateral oscillation of same degree of accuracy from one test to another.

+0.2

-0.1 -

-0.1 -3I.OFC NE
-0.2 . IN.OFF CENTER

+0.1- //-. .. 2IN. OFF CENTER

,A ,\ __ COMPUTER DATA

- 0 //.
NA

-0.1

+0.2

" o ,'

, " -0.2 -

0 10 20 30 40
TIME (MSEC)
Fig. 10. Comparison of experimental
and analytical dynamic characteristics
of three-torus system

257
 Numerous other dynamic tests conducted phenomenon that would exist at early compres-
with a single torus clamped on a plate and sion times of the torus if the plate were flat,
equipped with a counterweight to simulate the as ideally contemplated. This snatching phe-
axial loading by the inertia effect have also nomenon is caused by the fact that the plate
- demonstrated that energy storage capacity and velocity increment v is larger than the ver-
the forces developed in compression and ten- tical component of the wave velocity propagating
sion agreed with the formulas given in this along the torus wall and creates a release of
paper. the wall hoop tension, until this part of the wall t
is later snatched back into place. The shape of
Finally, Fig. i1 presents a comparisonbe- this clamping boss, of course, affects the torus
tween the calculated force characteristics of characteristics and is an additional cause of
the three-torus system and the static test re- errorin. the comparison of experimental results
sults. One sees that the agreement is excellent with analytical calculations.
in tension but is somewhat lacking in compres-
sion. The system had been cycled 1000 times
before the experimental curves were recorded.
The amount of hysteresis is small.

FORCE vs. DISPLACEMENT

(AFTER CYCLING) ;
Po = 55Psig
V
(NYLON WALL) 10

, 500 I~l
C',
PLATE TORUS
ATTACHMENT

-(-THEORETICAL CURVE TYPICAL FLAT CLAMPING GROOVE

2 01 i 2
DISPLACEMENT (INCHES) TORUS
(TENSION)
R
/ -500, i

- 1000

/1500 TORUS ATTACHMENT

/ TYPICAL IDEALLY SHAPED CLAMPING BOSS

Fig. 11. Comparison between ex- Fig. 1Z. Typical torus system
perimental and analytical results attachment scheme
for pure axial deformations

CONCLUSIONS AND
The problems of torus attachment to a plate RECOMMENDATIONS
subjected to repeated high-inten3ity shocks are
very difficult to solve. A typical machined Toroidal, flexible, gas-inflated, low-weight
clamping boss, shown schematically in Fig. 12, structures have proved to be excellent gas
was developed to be an integral part of the plate springs for special applications where no reac-
but also to eliminate the torus wall snatching tion is required in the neutral position and when

258
t both compression and tension strokes are spec- it is believed that the results given in this pap..i
ified. This gas-spring configuration also per- are probably accurate enough for an approxi-
mits rotation and side motion with resulting mate evaluation of the rigidity of a torus stack
restoring forces so that operation is possible when used as a gas spring. It would be desir-
with six degrees of freedom. It is iinportant, able to confirm these results with additional
therefore, to be able to calculate the charac- tests conducted with carefully constructed and
teristics of such systems so that the designer rigidly attached multitorus stacks. Also, the
may evaluate their performance when used influence of the type of construction,. i.e., cloth
within a more complex piece of equipment. or knitted filament, unidirectional or not, should
They are particularly valuable in applications be investigated further.
-where shock loading is expected, since they can
be made very flexible, even at their point of The experimental investigation should also
attachment. Although few applications have include tests at higher values of AX/d than have
been considered so far because of their high yet been performed, especially for tension,
cost due to the manufacturing complexity, torus since the discrepancy between theory and ex-
systems should be investigated as possible gas periment seems greater in this case. The in-
springs whenever special requirements com- fluence of the ratio D/d should be determined,
patible with torus system characteristics have since all experimental data so far have been
to be met. obtained with D/d values of approximately 5
and 7.
Because of the complexity of the analysis
of a system where both bending and side mo- In conclusion, a new analytical tool has been
tions are involved, many simplifying assump- developed, but its limitations and range of ap-
tions have to be made to make the problem plicability remain to be determined more accu-
amenable to analytical treatment. The results rately than has been possible until now. It is
obtained from such a study check relatively hoped that applications for this system will
well with crude experimental data. Therefore, justify continued work in this area.

REFERENCES

1. Astro Research Corporation, "Structural 2. C. V. David, "Axial and Lateral Rigidity of

Analysis of Torus Spring Assembly," Gen- Pressurized Toroidal Filament Structures,"
eral Atomic Informal Rept. GAMD-5148, General Atomic Informal Rept. GAMD-6061,
March 1, 1965 Jan. 13, 1965

259
 VIBRATION QUALIFICATION SPECIFICATION
FOR INERTIAL MEASURING UNITS
S. L. Burgwin
Honeywell, Inc., Aeronautical Division
St. Petersburg, Florida

Lack of understanding of the role of rectified torques in the perform-

ance of inertial measuring units (IMU) has contributed to a somewhat
expensive misuse of military vibration specifications in the vibration
qualification of IMU's. Difficulty arises from the usual addition that
performance shall meet performance specifications at qualification vi-
bration levels. This seemingly logical addition becones illogical when
the nature of rectified torques is considered.
Vibration levels for survival qualifications based on a limited sample
are ordinarily about three times the maximum ambient operating levels
to allow for material and piece part variation. Such levels become un-
realistic and unnecessary for IMU performance qualification when the
nature of rectified torque errors is considered.
Performance vibration errors in IMUs arise from rectified vibration
torques, which are a function of acceleration squared. A factor of 3 in
vibration level becomes a factor of 9 in performance vibration error.
Few inertial systems can afford this margin in performance. The ob-
vious solution is a separate vibration qualification specification for
IMU performance.
The problem of rectified torques is discussed in this paper, and an ap-
proach to arriving at a more realistic specification is suggested. The
discussion deals only with IMU's, but the same problems exist for any
measuring instrument utilizing a moving element exposed to ambient
vibration.

INTRODUCTION (IMU), where the problem has apparently been

most severe and has perhaps had the greatest
It is convenient, in vibration testing, to use attention. A searching investigation may un-
performance degradation as an indication of be- cover other devices and applications where the
ginning failure. Vibration qualification tests, problem is also serious but has not been
as now specified, genekally follow this practice. recognized.
Measuring instruments with moving ele- QUALIFICATION SPECIFICATION
ments, utilizing torque rebalance in the meas- PROGRAM
uring process, generally suffer performance
degradation when subjected to vibration due to Inertial measuring units (for example,
the generation of rectified torques. Such per- Fig. 1) of a particular design are qualified for
formance degradation is not an indication of operation in a vibration environment by sub-
beginning failure, and its use as such an indi- jecting a limited sample, usually consisting of
cator may result in failure of satisfactory one production IMU, to specified levels of
equipment to meet vibration qualification single-axis vibration along each of the princi-
specifications. pal axes in turn and monitoring IMU perform-
ance. Failure to qualify results from failure
Discussion of the problem of rectified of any mechanical or electrical component or
torques and an approach to a more satisfactory from performance degradation to values out-
vibration qualification procedure is confined to side of the specified performance limits for
the particular case of inertial measuring units the IMU.
261
 NV

,4

Fig. 1. Example of Honeywell IMU :

The overall objective of the qualification survival, but are in some measure common to
test is to establish a high level of assurance vibration qualification of any equipment to be
that the IMU will survive and perform satisfac- operated in a vibration environment. Present
torily in the operating vibration environment, practices in vibration qualification appear quite
successful in meeting the survival objective
Difficulties arise in specifying the test and, considering the state-of-the-art, probably
level of vibration for vibration qualification of cannot be greatly criticized,
an IMU. The level of environmental vibration Vibration qualification with respect to per-
to be expected at the IMU mounting point for a formance does, on the other hand, present prob-
specific application usually cannot be precisely lems unique to IMU's, or more generally to any
defined when the specification is written. Some instrument sensitive to spurious torques. Deg-
allowance must be made for lack of precision radation of performance of an IMU in a vibra-
in test simulation of the vibration environment. tion environment is not solely a survival prob-
A further allowance must also be made for IMU lem, but is also dependent on the value of
material and manufacturing variation in quali- rectified error torques generated by the vibra-
fication testing of a limited sample. In the ab- tion. These torques are a function of the square
sence of complete information in these areas, o The ton le anon o the di-
specified test levels of vibration, based on best rfthe vibration level and also depend on the di-
of retosuch o.,vibration. Th ircina eenec
notaeoften are an order
judgment,
jg od o magnitude
t or
ois that rectied torques balance to a zero
more higher than the levels subsequently meas- time average when the environmental vibration
ured under operating conditons.cmltl drcin
u is completely random in direction, in space.
Fo
Rrom the standpoint of survival, a sDegradation
th.tnpitosuv safety i niomn of IMU performance
h . .in an
. operat-
of 3 (a test level of vibration that is three ing environment can then be said to depend on
factor
times operating level) cthe lack of space randomness of the vibration.
times operating level) can e justed, anda in Single-axis testing makes no allowance for
many cases, good arguments can be advanced space randomness and thus does not simulate
.o assafceyofa s ig ae 1..e aonal the operational environment as regards recti-
assurance of survval under operating con- fied vibration torques; it tends to give much
tions may well be wort the prce of some higher values of rectified torque than in the op-
overdesign, erational environment.
The problems mentioned are not unique to The vibration directional dependence of the
vibration qualification of IMU's as regards rectified torques affect the results of the

262
 single-axis vibration tests in, another way. modified. A new and different approach is
There are, in general, three directions of vi- needed for vibration qualification of IMU per-
bration in the IMU which produce no rectified formance.
torque. The location of these directions depends
on cross-coupling in the IMU and may be close At present little or no precedent exists for
* to the principal axes. Qualification vibration a separate vibration qualification sp,'cification
tests now normally specify single-axis vibra- for IMU performance. All present vibration
tion directed along each of the principal axes qualification specifications have been directed
in turn. This procedure has justification for toward survival qualification, with performance
survival qualification but not for performance qualification as more or less a "fallout" from
qualification. Should the zero rectified torque the way of evaluating survival. Rectified
axes coincide with the principal axes and should torques have made this fallout something of a
the direction of vibration be actually along the hazard to successful qualification.
principal axes, there would be no rectified
torque or rectified torque problem, at least as Some discussion of the mechanisms of rec-
concerns meeting the qualification specifica- tified torques are necessary before the require-
tions. There is probably little chance of such ments of a vibration qualification specification
an occurrence, because the normal uncertainty for IMU performance can be fully apprediated
of vibration direction in single-axis testing is and a logical approach to such a specification
about *15 degrees. considered.

The vibration directional dependence of the

rectified vibration torques thus appears in qual- RECTIFIED TORQUE MECHANISMS
ways
;fication test results in two independent
which may tend to nullify each other and mask
/ ' he addition,
In
both effects. shipbeteen
ectiiedvibrtio relation-
the squaredtoruesand In general,
simultaneous
rectified torques result from
vibration at a given frequency in
ship between the rectified vibration torques and two of the six degrees of freedom of the body
1J the test vibration levels makes the rectified
torque problem extremely sensitive to the se- being vibrated.
lection of "safety" factors, relating operating
vibration levels to sr-iified qualification test Rectification arises from:
levels. The factors 3 to 10 justifiable for-sur-
vival qualification become factors of 9 to 100 1. Anisoelasticity- unequal spring con-
between the operating rectified vibration torques stants between the center of gravity and center
and those measured in the qualification tests at of support along the three principal axes.
the higher specified vibration levels. Rectifi-
cation torques are important enough in the per- 2. Coulomb or viscous material and struc-
formance of the IMU that a multiplying factor of tural damping (cylindrical torque) -vibration in
10 cannot be taken lightly, and this factor may two degrees of freedom, time-phased to produce
cause failure to meet performance specifications. a circular displacement of the cg with respect
to the center of support, results .n a constant
Degradation of performance by rectified torque proportional to the material and struc-
vibration torques is not a catastrophic type of tural damping.
failure. Furthermore, variation of the coeffi-
cients of these torques from unit to unit is sub- 3. Anisoinertia - angular oscillation about
I ject to manufacturing control, so that it is dif- two principal axes of unequal moments of in-
ficult to justify safety factors of more than 50 ertia produces an average value of torque about
percent of those suitable for survival qualifica- the third axis 0- -,t of inertia effect).
tion. Even performance qualification vibration
levels of 50 percent of the vibration levels
specified for survival qualification are by no 4. S . i eometry (coning) -angular
. the rectified torque oscillation two principal.axes
aboutmotion also pro-
: means a com.ulet
problem. answer to
Tht_ Llaiectional effects must also be duces rectified about thi third axis re-

considered. gardless of inertia, due to the noncomnmutation

of Euler's angies.
No single vibration qualification test now
available appears adequate to qualify both IMU 5. Static unbalance (vibro-pendulous
survival and performance in one test, even with torque) - several combinations of translational
two separate test vibration levels. Present and angular vibration can give rise to rectified
specifications appear adequate to qualify IMU torque when the cg is displaced from the. center
survival, if the performance requirement is of support.

263
 Almost any combination of vibration in two known types of rectified torque, the so-called
degrees of freedom, coupled with elastic ani- "anisoelastic" torque.
sotrophy, asymmetric mass distribution, or .
plain material damping, can produce rectifica- The mechanism of anisoelastic torque may
tion torques. be appreciated by considering Figs. 2 and 3. In
Fig. 2, a mass is supported within a ring by
Mathematically, any torque resulting from four springs. The springs in the horizontalaxis
the product of sinusoidally varying terms will have greater stiffness than those in the vertical
be a function of sine squared terms, which can axis. Dampers are shown to simulate material
be resolved into constant and double frequency hysteresis damping. This diagram is a reason-
components of torque. able simulation of a gyro spin motor supported
within a cylindrical floated gimbal or almost
The magnitude of rectified torques, in gen- any simple mechanical gimbaled system. Fig-
eral, pose no threat to survival and are usually ure 3 is a simplified view of Fig. 2.
of importance only in the degradation of the
performance of torque-sensitive instruments, In Fig. 3, displacement of cg, relative to
such as gyros and accelerometers used in an the center of support (cs), due to vibration along
IMU. This does not mean that the torques are the 45-degree axis, is represented by the dotted
always of small magnitude. Rectified torques line between the two extreme displacements of
have been observed of sufficient magnitude to the cg. Relative displacement between the cg
override gimbal servomotors, in some in- and cs is not parallel to the vibration direction
stances, of low damped resonant excitation, because of the difference in vertical and hori-
(Torques of as much as 30 in.-oz have been ob- zontal supporting spring constants. This dis-
served in a 19-lb gimbaled platform.) The placement at low frequencies is small com-
magnitude of the rectified torques are, in gen- pared to the displacement of both cg and cs in
eral, a function of mass squared and may be of inertial space. Therefore, the direction of the
interest in some applications outside the scope inertial force (Fm) is essentially parallel to the
of this discussion. A better appreciation of the direction of vibration, and since it is directed
mechanisms and the problems involved can be through the center of mass, will have a lever
gained by a brief discussion of one of the better arm about cs.

K._
(---
Y
I
2 2

Fig. 2. Diagram of mass-spring-damper arrange-

ment to illustrate anisoelastic torque

264
 Yr

o-4

D IRECT ION

ACCELERATION TORQUE

LEVER "IRH

TORQUE

/'--DOI RECT ION OF

CASE ACCELERATION

Fig. 3. Torque due to unequal spring constants

The resulting pulsating torque in this case where

is always clockwise and is a function of time,
as shown in Fig. 4. The direction of this torque 8 = angle between the vibration direc-
is such as to turn the mass and supporting tion and the x axis,
structure until the axis with the lowest spring
constant or maximum compliance is aligned f = vibration frequency,
with the direction of vibration.
fx, fy = natural frequencies along the x and
The example is admittedly not exact, but y axes,
serves to give some "feel" for the mechanism K ,K
involved. Experimental models similar to the axsp and
mechanism have been built and demonstrate the axes, and
principle quite successfully. Bx By = hysteresis damping constants along
x and y axes = twice critical damp-
The relationship between the rectified ing ratio.
torque (Tae), mass (M), and peak vibration ac-
celeration (A) is given by the following equation: When the vibration frequency is low relative to
fx and fy, and B and B are small, Eq. (1) re-
duces essentially to the well-known equation
2
M2 A sin 0 cos 0 M2A - y
Tave 2 sin cos 0 (2)
. .. . 2'
1-- 1

[ __x 2
fx_ f y2 , The anisoelastic torque coefficient (E) is
[K [(I + B] K [ + BE= 'aveM2 sin cos8[.x, .. (3)

2'A6ms

2.65
 DOUBLE FREQUENCY COMPONENT

TORQUE

AVERAGE
TORQUE

TIME

Fig. 4. Time variation of anisoelastic torque of Fig. 3

A factor Z, related to transmissibility 1. The time average of the rectified torque

along the x and y axes, can be defined by the is zero, if the direction of vibration is corn-
relationship pletely random.

Tave = EZ. (4) 2. Rectified torques are zero for direction

of vibration along the principal axes.
Variation of Z with frequency is shown in
Fig. 5. Only the low-frequency portion of this 3. Rectified torques vary 50 percent of
curve is of interest in the problem of vibration the maximum value for variation of vibration
qualification of IMU's. The IMU performance direction of 15 degrees from the principalaxes.
degradation is largely a function of the rectified
torques in gyros that can be designed so that the 4. Rectified torques are maximum positive
lowest resonant frequency is high enough to or negative in the 45-degree directions.
avoid much resonant rise in the test region.
For example, Honeywell gyros are designed so 5. Rectified torques vary only 7 percent
that the lowest resonant frequency is well above for 15 degrees variation from the 45-degree
2,000 cps. directions.

Equation 2 is the equation of most interest Other types of rectified torques mentioned
in the vibration qualification problem. It will earlier differ from the characteristics of the
be noted that a close match of x and y compli- anisoelastic torque described. It is not within
ances (1/K x and I/Kr) reduces the rectified the scope of this discussion to cover each type
.... in detail, but in general they are all functions t,
torque. Practical limits in this direction have
been reached in present gyro design so that ad- of the square of the vibration level and are sen-
ditional factors of safety are difficult, if not sitive to direction of vibration. Variation of
impossible, to achieve, rectified torque with frequency is quite differ-
ent
the for.the other types
characteristics of rectified
of interest torques, qual-
in vibration but I
Attention has already been directed to the ification of performance are essentially similar
squared relationship between the rectified to those listed for anisoelastic rectified torques.
torque and both mass and acceleration. The
dependence of the rectified torque on the direc- A brief bibliography is given at the end of
tion of vibration is of particular interest in the this discussion for those interested in a more
vibation avibrationoofeparticul ariin stin detailed study of the various rectified torque
vibration qualification problem. The variation
of rectified torque versus direction of vibration mechanisms.
is shown in Fig. 6. Attention is directed to five The characteristics discussed furnish a
characteristics important to the vibration qual- basis for a suggested approach to vibration
ification of performance problem: qualification of IMU performance.

266
 30 j

20-

-10-

-20-

-30-
TAVE EZ

-50- Kx-Ky =0. 10

-60

-70-

-80

-901

Fig. 5. Variation of anisoelastic torque with vibration frequency

C.W.
TORQUE

DI.V-RTO

TORQUE

Fig. 6. Anisoelastic torque vs direction of vibration, 0

267
QUALIFICATION SPECIFICATION goal for these levels should be approximately
FOR IMU PERFORMANCE 1.5 times the rms operation vibration level
averaged over the duration of the mission.
it has been pointed out that present speci-
fications for vibration qualification of IMUs 3. Test vibration for performance qualifi-
would be relatively satisfactory from the sur- cation is to be directed along each in turn of
vival standpoint, if the requirement of meeting
performance at the high vibration test levels
were modified. It is suggested that perform-
the six face diagonals of the unit cube formed
with the principal axes along the cube edges
(Fig. 7). Performance is to be measured for
19
ance within specifications limits, after subject- each IMU principal axis for each of the vibra-
ing the test IMU to the specified vibration tion directions, a total of 18 values or six val-
levels, is sufficient to monitor survival. ues per axis. Instead of the six face diagonals
or 45-degree directions, it may be desirable to
The performance degradation due to recti- make a worst-axes search. It is anticipated
fied torques, which are chiefly responsible for that such a search will show the 45-degree axes
suggesting such a change, must be suitably to be close to the worst direction.
measured and limited before complete vibration
qualification is achieved. Performance qualifi- 4. Specify two sets of performance limits:
cation presents some difficult problems because (a) a worst-case limit based on expected compo-
of the two-axis nature of the rectified torques. nent performance and not mission performance;
and (b) superposition limits to approximate
Information gathering and testing, as now IMU performance under three-axis vibration
practiced, are essentially single-axis proc- random in direction in space. A superposition
esses. Simulation of a two-axis environment limit for one IMU axis would be the algebraic
with single-axis tests at best involves some sum of the six test values for that axis.
important approximations.
Justification for the reduced vibration lev-
A suggested approach to the vibration qual- els of step 2 above is relatively simple. Cata-
ification problem for IMU's is: strophic failure is a function of maximum
stress. Survival, therefore, depends on peak
1. Utilize present vibration qualification vibration levels. IMU performance error (an-
procedures to qualify survival only, with the gular rate) depends on acceleration squared.
provision that performance meets specified IMU total navigation angle error depends on the
values before and after vibration, but not nec- integral of the performance rate error and,
essarily during vibration- vibration specified therefore, on the rms vibration lev2l averaged
along principal axes. over the acceleration versus time profile for
the mission. Thus, vibration levels specified
2. Generate separate specification for IMU for qualifying survival may have little or no re-
performance with reduced vibration levels; the lationship to the values required to qualify IMU

Z AXIS

DIRECTION OF
VIBRATION

DIRECTION OF
VIBRATION DIRECTION OF
VIBRATION

AXIS

X
AXIS

Fig. 7. Suggested vibration directions --

performance qualification

268
performance. Furthermore, a factor of safety meager, again due to lack of awareness of its
of 3 suitable for survival testing can be reduced importance.
to 1.5 for performance qualification, because
rectified torque coefficients are subject to It must be recognized that the value and useful-
closer manufacturing control than is ultimate ness of any performance qualification specifi-
strength of materials. It should be noted that cation must necessarily be limited by such lack
the 1.5 safety factor (test vibration level 1.5 of information.
times operating level) covers a variation of 1.5
squared or 2.25 in rectified torque coefficients.
Some assumptions can be made that will
Justification of the 45-degree vibration di- give the suggested approach a degree of value
rection is also relatively straightforward. and usefulness until better environmental and
Ideally, the vibration directions should be such linearity data are available. Until this time,
as to give maximum rectified torques, that is, an assumption of complete space randomness
parallel to the face diagonals of the unit cube of environmental vibration is the only practical
formed with the principal IMU axes along the assumption possible. This is not a safe as-
cube edges. In practice, cross-coupling in the sumption but some evidence indicates that it
IMU shifts the maximum rectified torque axes may be a relatively good assumption. For the
away from the 45-degree directions, so that purposes of performance qualification, it may
vibration in the specified 45-degree directions turn out to be a very good assumption.
will produce less than maximum rectified
torque. It might be argued that a search for IMU performance under this assumption
the maximum rectified torque axes may per- will be represented by the algebraic sum of the
baps be desirable. However, other test errors IMU drift rates about each principal IMU axis
are such that variations of 15 degrees from for the six directions of vibration. If the IMU
the maximum axis are tolerable and to be pre- were mechanically linear, the summation drift
ferred over the greater complexity of a worst- rate ideally would be zero. Actually the sum-
axis search. It is suggested that specification mation values would reflect testing uncertainty.
of the 45-degree orientations of vibration test- The proposed test could turn out to be a good
ing is a practical choice. measure of the IMU nonlinearity within the lim-
its of testing uncertainty. This may be the
Selection and justification of the suggested more important factor in evaluating IMU per-
specification limiting values present some formance, particularly if the environmental vi-
problems. The worst-case component limit, bration is nearly random in space-direction, as
step 4a, must be based on the state-of-the-art some meager evidence seems to indicate.
and is designed merely to assure state-of-the-
art or better components; it does not present a
serious problem. The concept of space randomness of vibra-
tion seems to be a difficult one for most engi-
The superposition limit, step 4b, is more neers to accept. In general it can be said that
difficult to select and justify. It represents an the more complex the structure linking the
attempt to simulate three-axis performance forcing function to the measuring point, i.e.,
with six single-axis test values. The effort is the greater the number of power paths, the
handicapped
p by lack of information in two areas:
bmeasuring more nearly random will be the vibration at the
point. The present practice of spec-
1. Space randomness, or rather lack of ifying one power spectral density curve for all
randomness, in direction of the environ- axes tends to reflect a perhaps involuntary
space radon- in direcin oer practical acceptance of the space randomness
mentalof vibration in existing structures. Perhaps the
spectral density between the six 45-degree axes
is of utmost importance in evaluating perform- assumption of space randomness of vibration
ance. Such data are not now available, probably should be accepted until proven otherwise.
because of a lack of awareness of its impor-
tance. Attention in gathering environmental Completely realistic specification limits
data has been focused on requirements for sur- cannot be achieved without a knowledge of the
vival evaluation where only the maximum power environmental forcing function and the transfer
spectral density, regardless of direction in function between the points of application and
space, has significance. the mounting points of the device to be evalu-
ated. The complexity of these functions inhibits
2. Valid use of the superposition principle progress in this direction. Engineering assump-
requires linearity of the mechanical character- tions can be expected to be necessary for at
istics of the test unit. Such information is also least the immediate future.

269
CONCLUSION of rectified torques described in this discussion,
which include space randomness of the environ-
The problem of unjustified rejection of mental vibration and nonlinear mechanical char-
torque sensitive instruments, such as IMU's, in acteristics of the device under consideration.
present vibration qualification procedures be-
cause of performance degradation by rectified An approach to a performance qualification
torques under unrealistic test conditions can be specification for IMU's is discussed. It involves
alleviated, if not completely solved, by evaluat- practical single-axis measurement of the effects
ing performance and survival under separate of rectified torques and combines the results to
qualification specifications. simulate a three-axis environment. Admittedly,
this approach is not ideal, but it may represent
The performance specification must be de- the most practical approach to the problem at
signed to take into account the characteristics the present state-of-the-art.

BIBLIOGRAPHY
Fernandez, M., and G. Macomber, Inertial Pitman, G. R., Jr., Inertial Guidance (John Wiley
Guidance Engineering (Prentice-Hall, New & Sons, New York), 1962, Ch. 3, pp. 79-91
York), 1962, Appendices G, H, and I Weinstock, H., "Specification for the Permissi-
ble Motions of a Platform for Performance
Goodman, L. E., and A. R. Robinson, J. Appl. Evaluation of Single-Degree-of-Freedom
Mech. (June 1958) Inertial Gyroscopes," M.I.T. Instrument
Laboratory Paper E-1267, Dec. 1962

ttV'

270
 - - - - -
. _ - ,-- ..... . . -.. . - *

EVALUATION OF A DESIGN FACTOR

APPROACH TO SPACE VEHICLE DESIGN
FOR RANDOM VIBRATION ENVIRONMENTS
D. E. Hines and D. A. Stewart
Douglas Aircraft Company, Inc.
Missile and Space Systems Division
Santa Monica, California

Structural fatigue induced by random vibration often sizes bracket

structures in modern aerospace veh.icles. This document reports the
results of an experimental study conducted to evaluate a design factor
approach for application to the fatigue problem and to evaluate the de-
sign factor for cantilever beams.

INTRODUCTION AND SUMMARY than those determined from an analysis based

on Miles' theory (1). The scatter in K values
The study described in this document was was on the order of 20 percent for aluminum
conducted to evaluate a technique for designing specimens and 50 percent for steel specimens
space vehicles to withstand random vibration (within 95 percent confidence limits). In both
environments by means of "design factors." cases, this scatter was deemed reasonable
Using various design factors (K values), 140 considering the specimen materials and surface
cantilever beams with stress risers were de- finishes. Therefore, the general design factor
signed for the test. These were then tested to approach was shown to be feasible and the
failure in a specified random vibration environ- method of defining the design factors was shown
ment. Since the beams were expected to have to be acceptable.
different failure times (depending on the design
factor used), it was expected that a design fac- Further data refinements, based on logical
tor could be isolated which would guarantee a
fatigue life suitable for space vehicle applica- assmt stresc nentrai f
tions. For this sudy
tion. Fr thedesredfatiue
study, ths ifetors
the desired fatigue life at stresses
to give useful random fatigue yield,
approaching were applied
information to
time was 10 min; however, data are included to supplement the limited quantity of data pub-
allow extrapolation to other fatigue lives. lished on this subject. In addition, sources of
structural damping were investigated, and semi-
When tested to the design level, the speci- quantitative agreement between experimental
i mens' fatigue lives ranged from 3 x 103 to data and the theory of Lazan (2) was obtained
5x1l0 5 cycles. However, when the raw data for specimens with and without stress risers
were analyzed, a considerable "scatter" was when subjected to random excitation.
observed. This was attributed to uncertainties
I inherent in random data analysis, variations in
system damping from specimen to specimen, In summary then, the primary objectives
variation of stress concentration factor, and of this study were (a) to attempt to validate the
natural fatigue scatter. design factor approach, and (b) to determine
the constants for use in this design approach
Through the use of statistical techniques, based on random vibration fatigue tests of can-
the scatter due to data reduction was greatly tilever beams. Secondary objectives were (a)
reduced, leaving only that due to variation of to add to the existing body of experimental
the materials. The design factor was then ex- random fatigue data, and (b) to identify (quali-
pressed in statistical form. The design factors tatively and quantitatively) sources of uncer-
obtained experimentally were slightly lower tainty in dynamic response problems.

271
NOMENCLATURE W End weight (lb)

b Width (in.) x Tip deflection (in.)

Ratio of damping to critical damping x Tip acceleration (in./sec 2)

c/c c

=
quality factor p Density (lb/in. 3)
Q 1/2c/cc

E Modulus of elasticity in tension (psi) aeq Equivalent stress (psi)

fo Undamped natural frequency (cps) a. Static stress per unit load (psi/g)

S27fo an Nominal stress at stress riser (psi)

Ftu Tensile ultimate stress (psi) am o%x k f (psi)

Fry Tensile yield stress (psi) a Fatigue curve slope on log-log paper

Ft Working stress (psi) AdB Uncertainty in db at the 95 percent

confidence level for power spectral
i Thickness (in.) analysis

I
4
Second moment of area (in. ) iAf Frequency window utilized in data
reduction
Kdyf Dynamic load factor
1 <> Indicates rms
Kdyn = 1

for sinusoidal loading; DISCUSSION

S7Tfo 1/2 There have been many studies of the fatigue
Kdyn - c- characteristics of materials subjected to ran-
dom vibrations and loads (3-15). However, the
for random loading use of these data in design applications has
been limited. It would be beneficial if structural
K Design factor for random loading design for random vibration environments
could be reduced to a simple procedure, that
K, Design factor for sinusoidal loading is, if design factors could be developed which,
when applied to static response calculations,
a Constant from Miles' fatigue theory, would account for the dynamic characteristics
chosen as 1 for this study of the system and the fatigue characteristics of
the materials. Should such a simple procedure
kt Stress concentration factor for static prove impractical due to the large number of
loads in tension variables and the uncertainty in determination
of each variable, it might be advantageous to
k Stress concentration factor for static attempt a statistical approach to the problem
loads in bending similar to that used in estimation of space ve-
hicle vibration environments.
kf Stress concentration factor for fatigue
In brief then, the purpose of this study was
t Length of beam (in.) to evaluate a specific design technique for ran-
dom vibration environments. This technique
M Bending moment (in.-lb) may be summarized by the equation

PSD(f) Power spectrum of random input ao (load)

2
(g /cps) aeq (K)(Kdyn)

t Averaging time used in data reduction where

T, Time to failure =eq stress compared to Ftw;

272
/,V
 Ftw = working stress which is ultimate bandwidth resonator) being evaluated for a rel-
over 1.4 yield or over 1.1, which- atively short duration; (b) natural scatter in Q
ever is less; for a given beam configuration; (c) finite time
for evaluating the input power spectral density
K = design factor including fatigue (PSD); (d) variation in the stress concentration
properties of the material (Ftw over factor as a function of stress; (e) insufficient
fatigue strength at the desired life); test duration; and (i) the narrow notches in the
input spectrum at specimen resonance. Of
Kdyn = dynamic load factor (71/2) Qf 1/2 these sources of scatter, item (e) requires
for random, Q for sinusoidal ex- further explanation. Assuming that fatigue ac-
citation; cumulates according to Miner's hypothesis, the
predicted fatigue life would be obtained if the
a = static stress for loads, such as proper number of peaks of various levels have
those applied in vibration; and occurred. This would require a test time on
the order of, or greater than, that required to
load = excitation force. get an accurate value for the rms response.
As is shown below, this time might be as long
To evaluate the design approach implied by as 400 sec or 7.5 min for systems with Q's as
this equation, 140 beams were designed based high as those studied here. Any deviations
on a dynamic stress analysis of the beams, an from Miner's hypothesis increases the test
assumed damping, sinusoidal fatigue data and time required to eliminate the uncertainties in
the random sinusoidal equivalence derived by this parameter. Since the average test time is
Miles (1). only slightly greater than the minimum re-
quired (about 10 min), a moderate amount of
It should be emphasized that in 'this study, scatter was introduced.
K and Kdy n are interdependent and cannot be
separated. K can be used to determine a K. The data resulting from this study were
value for sinusoidal input through division by then used to obtain the K values in three steps:
Miles' equivalent stress factor. (a) all uncertainty was incorporated in the K
factor, (b) the effect of stress on the magnitude
The 140 specimen beams were designed of stress concentrations factor was eliminated
and constructed of 2024-T4 aluminum or 4130 from K, and (c) the remaining scatter due to
steel heat-treated to 140 to 160 ksi. The in- K and Kdyn was separated by making logical
tended life of these specimens was about 3 x 10 assumptions concerning the natural scatter in K
cycles, corresponding to a 10-min life for a as discussed in "Test Results and Analysis."
specimen resonant at 50 cps. The design of the
specimens was varied to insure that the 3X 104 The experiment indicates that a K factor
cycle life was included and to provide useful obtained analytically from Miles" theory gives
fatigue information above and below this value, values which correspond to the 80 to 90 percent
The actual test lives varied from 3x 103 to confidence limits. Therefore, values obtained
5x10 5 cycles. from Miles' theory are of the proper magnitude
for design purposes. Table 1 presents K values
The beams were instrumented to yield the using this theory. The proper K value for a
acceleration input level, the acceleration re- specific design purpose will depend on the
sponse, and, in a few cases, the stress levels amount of information available concerning the
to determine the possible effect of modes higher problem at hand.
than the first mode.

Instrumentation of the -tip output accelera- 'SPECIMEN CONFIGURATION

tions fulfilled two purposes: (a) to allow an
estimate of stress for each beam without re- Two basic types of simple fatigue test
lying on strain gage data, and (b) to evaluate specimen might have been used for this study.
the Q's of the individual beams. The first type consists of specimens without
stress concentrations, and the second type in-
The data were plotted in fatigue curve form corporates stress concentrations. Each con-
utilizing the calculated stress per unit of tip figuration had certain advantages and disad-
acceleration and the measured tip acceleration vantages, which are discussed in Appendix A.
levels. An excessive amount of scatter was ob-
served. It was determined that this scatter was To determine which type was to be used
due to the following: (a) the rms tip response for this study, two sets of four specimens each
(which was the response of a very narrow were fabricated and tested to failure. One set

273
 I7

TABLE 1
Derived from Miles' Theory
Stainless
Test IAluminum
2024-T36
Aluminum
7075-T6
Magnesium
AZ 31B-0
Steel
17 7 PH
Beryllium
S 200C
_TH 1050

Tensile ultimate 70 78 32 180 70

stress (psi)
Tensile yield 60 70 18 150 50
stress (psi)
Working stress (psi) 50 55 16 130 45

Fatigue strength in
R -1 bending at

104 cycles 48 49 23 115 24

10 5 cycles 37 37 20 104 23
106 cycles 29 26 17 83 22
K. at
104 cycles 1.04 1.02 0.70 1.13 1.88

105 cycles 1.35 1.35 0.80 1.25 1.96

106 cycles 1.73 1.92 0.94 1.57 2.00

Equivalent stress
factor at
104 cycles 2.2 2.1 2.6 3.0 2.9
105 cycles .... 3.6
106 cycles .... 5.6

K at

104 cycles 2.28 2.14 1.82 3.39 5.45

105 cycles 2.97 2.84 2.08 3.75 7.05
106 cycles 3.80 4.03 2.44 4.71 11.22

had stress risers and the other did not. The from this report. Those tests did produce
specimens without the stress risers exhibited some interesting information, however, which
considerable nonlinear damping and a poorly is discussed in Appendix B. The remainder of
defined failure point. The specimens with the this report is limited to discussion of speci-
stress risers exhibited only nominal nonlinear mens with stress risers.
damping, and there was no difficulty in defining
the failure point of these specimens. There- The general configuration used in this
fore, the specimens containing the stress risers study consisted of an end-weighted rectangular
were chosen for the test. cantilevered bar containing two holes for stress
risers. Such a specimen is shown in Fig. 1;
The design and experiments performed on Figs. 2 and 3 give fatigue curves for aluminum
the specimens without the stress risers were and steel specimens, respectively. The end-
parallel to those for the specimens with stress weighted bar permitted the stress per given in-
concentration factors. Therefore, details on put to be altered without changing the basic bar
these rejected specimens have been omitted configuration. Therefore, many specimens

274
 Specimen Sizing

Before the specimens could be sized, some

basic parameters were to be defined. It was
decided that the stress concentration holes
would be 1/8-in. in diameter. To limit inter-
action between the stress pattern resulting
from one hole and that of the second hole or
from the edge of the bar, minimum separations
were required. A 5-diameter separation be-
tween the hole centers and a 2-1/2-diameter

*
f separation between the hole centers and the
beam edge were sufficient for this purpose. To
meet these requirements, a beam 1-1/4 in. wide
- .f: was required. In addition, it was decided to
S" test four general beam configurations, two
thicknesses of 2024-T4 aluminum and two
* "thicknesses of 4130 steel heat-treated to 140 to
: \160 ksi. The test level was defined as 0.45
. -, gc2/ps, and the average life was to be between
104 and 105 cycles to failure.

A preliminary set of specimens referred

to as the "peg point" specimens was fabricated
Fig. 1. Beam configuration and tested to verify the analysis procedure.
with stress risers These specimens were designed using the fa-
tigue analysis approach suggested by Miles,
estimating damping as described in Appendix B,
with varying expected fatigue lives could be and then (by analysis) relating a K factor for
fabricated of identical material and cut from both sinusoidal and random loading to each
the same sheet. This limited the scatter due to specimen. After the analysis was verified by
r: material composition and specimen thickness. these tests, the final design loads were defined

i' ~SINUS,
IDAL PEAK'TRS

AlABOVECURV
- SIG ,MILES'THER--

101 -

I0 105 I01
LIFE INCYCLES

Fig. 2. Fatig'ne curves for 2024-T4 aluminum under bending

275
 - -,

2L

___ -.... _SINUSOIDAL PEAK

STRESS

RANDOM
RMS
STRESS
DERIVED
FROM
ABOVE USING
CURVE
MILES
THEORY

10'
4 6 8 105 2 4 6 8 106 2 4 6 8 107
LIFE INCYCLES

Fig. 3. Fatigue curves for 4130 steel hardened to 150 ksi

in a manner that would vary the K factor over parameters. The design was then refined using
a suitable range. the more sophisticated approach discussed
below.
The design goals for the peg point speci-
mens were that the specimens fatigue in 10 In the more refined analysis, the Douglas
min when excited by a random loading of 0.45 version of the Mykelstad ',poroach to modal
g 2/cps and that the frequency be on the order properties was used to de..rinine the natural
of 50 cps. The general configuration of the frequencies and mode shapes for deflection and
specimens was chosen as end-loaded cantilever moment (normalized to 1 at the tip).
beams containing stress risers and excited at
the beams' roots. Using this information, the tip acceleration
for a 1-in. tip displacenent in the mode being
The preliminary design was obtained by considered can be obtained from
using the desired life of 3x 10 4 cycles (10 min
=
for a 50-cps system) and utilizing the fatigue IX(g)I { %2(1 in.)/386.4 in./sec2} g
curves indicated in Figs. 2 and 3 to obtain nom-
inal rms stress at the stress r.ser. or

The stress concentration factors were es- JX(g) 1 (,),2/386.4) in g's.

tablished as discussed in Appendices C and D
after
beamsthe desired
were thicknesses
chosen. of 1/8-are
These factors The moment distribution obtained from the
and1.47Thmoetdsrbinotaedfmte
1/4-in.
bams we5fr choe1/8an.These factn. s re- Mykelstad program can be used to obtain the
and 1.57 for the 1/8- and 1/4-in. beamis, re-noiastestthsrssisrfm
spectively, for both the steel and aluminum nominal stress at the stress riser from
specimens. MC 6M
In addition, a frequency of 50 cps was cho- 0 bh 2

sen for the specimens, and Q was assumed to Therefore, the stress per g ef tip acceleration
be 50 (based on material presented in Appendix is given by
B and in Refs. 1 and 3-8). This furnished suffi-
c'ent information to obtain by simplified theory - 6M(386.4)
approximate values for the remaining undefined " ' '

276
If random response is desired, the rms stress Final Specimen Design
can be obtained from
The final specimen designs were arrived
1/2 %at by a slightly different system than was used
Qf PSD(f)J - for the peg point specimens. The first mode
g shape for the beams used in this study is nearly
or independent of the end weight (mb) for both
displacement and moment. Therefore, for a
01 4 P 1/2 aIX 386.4 given beam thickness, the stress at the critical
bh 2o 2 section is dependent only on tip displacement.
or This results in the fact that the stress per
given acceleration is proportional to the in-
1 2
<a>n _ (Q / 3Mx 386.4 verse of the frequency squared.
(g
2
/cps)l/2 ko3] bhi2 The tip acceleration is expressed as

2 I <i> (Q PSD1/2
< Q>n 1.16x13 i/Q
2
,7 1.7 bh f bY

For h=1/8 in., and, as stated above

4.74 x 103 Mn f2
2
(g /cps) 1/2 "b f3
Therefore,
1/4 in.,
and for I, =
1
_ - 1._18x 10 3
<0-> M Q, 1/ 2 <0->
n = 3 2
f /
2 1 2
(g /cps) / ) but

The natural frequency and maximum stress f

including the concentration factors for each mb
beam is given in Table 2, assuming Q = 50 for
the configuration shown in Fig. 4. The test re-
sults from these specimens are included with 3/4
those of the specimens for final analysis. The < n c = b
only result which should be noted here is the
fatigue life for each specimen. These ranged The desired spread in rs and, consequently,
from 1.6x 104 to 4.2x 104 cycles. These levels the K factor can be obtained by altering the end
are sufficiently close to the desired levels to mass.
be used as a basis for the final specimen design.
The end masses for the median test speci-
TABLE 2 men were increased slightly to eliminate dupli-
Frequencies and Nominal Stress for Peg cations, and the K's were established as 66, 77,
Point Specimens for 0.45 g 2/cps Input 88, 100, 111, 125, and 132 percent of the median
K. The mass variations required for these
Beam f Max <a> per
( 2 /p)1 values are included in Table 3.
Description (cps) (g/cps)/ 2
The nomenclature format which is used in

1/8-in, steel 50.0 66 this document to designate the beams is xx-xx.

The first two places are Al, A2, S, or S2, in-
1/4-in, steel 50.6 76 I dicating 1/8- or 1/4-in. aluminum or 1/8- or
1/4-in. steel specimens, respectively. The
1/8-in, aluminum 52.6 26.6 second two places correspond to the end weight
as indicated in Table 3. The order of increas-
1/4-in. aluminum 53.3 30.2 ing weight or stress is S1, R1, S2, R2, S3,
R3, and 84.

277
 3DETAIL II

3.19 - = DIMA
soo
o., 1.500 ~ 64 SEE
TABLE
4 3/8 IA 3/8 14+_-0.62

1.LM ABT

w- GRAIN

0.514 0.0 DIAMETER,

2 HOLES
. DIAMETER.
2HOLES
.0.521

0.191
DIAMETER
HOLE GENERAL NOTES
1. TOLERANCE
2 PLACE DECIMAL
-+0.02
PART NO. MATERIAL THICKNESS
(STK) ON A 3PLACE DECIMAL -0.005
2. AREA MARKED * TO BE
FREE OF SURFACE
Al 2024-T4ALUMINUMSHEET 0.125 5.00 DEFECTS (VISIBLE)
A2 2024T4 ALUMINUM PLATE 0.250 7.05 3. HEAT TREAT SI AND S2
SI 4130STEEL SHEET 0.125 5.60 TO140-160 KS1 BEFORE
S2 4130STEEL PLATE 0.250 7.90 DRILLING.

Fig. 4. Beam configuration

TABLE 3
End Weights and Geometric Data for Final Beam Designs

End Weightb
Material Designation Lengtha (b)
______RI I____R2 IR3 ISi I 2 JS3 S4__

Aluminum K. 1.35 1.8 2.25 0.57 0.76 0.96 1.15

Al (1/8) 5.00 0.127 0.187 0.252 0.108 0.161 0.215 0.262

A2 (1/4) 7.05 0.368 0.527 0.710 0.305 0.455 0.607 0.710

Steel K 1.65 2.2 2.75 0.70 0.93 1.16 1.4

Si (1/8) 5.60 0.272 0.400 0.540 0.232 0.344 0.460 0.560

S2 (1/4) 7.90 0.78 1.12 1.52 0.265 0.97 1.30 1.58

aMeasured from stress concentration holes to end of beam.

bIncludes weight of last inch of beam and of accelerometer.

A more sophisticated analysis using the TEST CONFIGURATIONS

Mykelstad program was then performed, as
previously discussed. The results of these The specimens were tested to failure in
analyses are included in the data reduction, but the Douglas test facility where the test equip-
the changes are less than 2 percent and suffi- ment consisted of (a) the electrodynamic shaker
ciently small so as not to affect the design. It and its support systems; (b) instrumentation,
should be noted that the second mode stress re- including strain gages and accelerometers; (c)
sponse was always less than 5 percent of that recorders, both magnetic tape and oscillograph;
in the fundamental mode. and (d) data analysis equipment, both analog
278
and digital. Figure 5 is a block diagram of the on magnetic tape on an Ampex 14-channel tape
data system. Instrumentation is shown in Fig. 6. recorder.
All test equipment is listed in Table 4.
Power spectral density (PSD) plots (Fig. 8)
The specimens were mounted four at a time and peak distributions were produced by the
(or for final tests, ten at a time) on the C-100 IBM 7094 computer (with Douglas computer
shaker which has a rating of 12,400 force program TD60), utilizing the taped information
pounds. The shaker was driven by a 90-kva obtained from the Ampex recorder. The maxi-
amplifier, and the system was equalized by an mum dynamic range of the analysis is approxi-
automatic equalizer. The original plan called mately 40 db. Resolution and accuracy tradeoff
for the use of peak-notch filters to refine the is in accordance with the relationship'for the 90
equalization, but they were not required. percent confidence intervals as shown below:

200
The beams were instrumented with both Af Xt
accelerometers and strain gages. The accel-
erometers were mounted at the end of the beam
and the strain gages were placed as shown in TEST PROCEDURES
Fig. 7. Unfortunately, little strain gage infor-
mation was obtained due to gage failure, adhe- The tests on each set of beams were di-
sion, and calibration problems. Data from the vided into three parts: equalization, sinusoidal
accelerometers and strain gages were recorded sweep, and random testing.

T TT AMPLIFIERS

LTYPICAL
ACCELEROMETER
CHANNEL

DIGITALDT
REOUCTION TAPE RECORDER

1 14CHANNEL

TIECD TECHNICAL
TCD00
GENERATORJ PRODUCTS
WAVE SL
ANALYZE
R PA

Lo'suA!.
EAmOP

1OSCILLOGRAPH

01 DIR

Fig. 5. Instrumentation block diagram

29
" 279
 I VirtoI
M -0 I

I Truerms VoFig.e 6. Iantumentation

Quantvoity Amlfesrpt ing SJ2-

Moe eiao
1 Viraqenounte
- lttPckr 5-100 634955
1 Oscillatcop-g RM20023
Co-toni
1 Oscillaora - Heletckr 5-24-403
1 Servo ontolAier - Lingve S-2,205H-B G0
1 Truee rsometer - Ballantin 3205 -A0 A

1 Servon DGe Amlf2r2LigS17-

1 Freuecyouter - HelttPckr 5121 63455

1Osillscoe -Tektoni RM2 0020

 Aa

I>:

Fig. 7. Transducer configuration

(0.5 - CYCLE BANDWIDTH)

- -- -- - -- 1C0JItdAL TEST LEVEL

- - - 1.5 Q

1010

Fig. 8. Typical input power spectral density

281
Equalization a 1-octave/min sweep was too fast and did not
allow the beams to build up to full amplitude at
With four dissimilar beams mounted on the rescnance. However, although the sweep rate
shaker, a random vibration input over a fre- was too high, it did allow localization of any
quency range of 20 to 2000 cps was applied major resonances in the beams. To determi-ne
through an 80-channel automatic equalizer, frequencies of resonances accurately, the sweep
The level of thi random input was approxi- was rerun with an input level of 0.1 g peak. The
mately 10 percent of the final rms random test frequency range of the sweep was 30 to 60 cps
level. The output of the control accelerometer, and the sweep rate was 0.1 octave/min.
mounted in the center of the shaker head, was
recorded on magnetic tape. From this tape,
loops were cut and the data ware reduced using Random Tests
a Technical Products wave analyzer. A PSD
plot of the shaker output was obtained over two After the system had been equlized to
frequency ranges, 20 to 2000 cps and 20 to 200 within the limits described for the sinusoidal
cps. The filter bandwidths selected on the wave sweep, the random input was app]ied at the
analyzer for these plots were 20 and 5 cps. 0.45-g 2 /cps design level. Accelerometers
These PSD's were then corrected for the fre- mounted at the tip of each beam, the control
quency response characteristics of the tape re- accelerometer, and the strain gages mounted at
corder, and the corrected curves were checked the stress concentration points on selected
to insure that the input level at the beam's first beams were monitored. The beams were tested
natural frequency was within 1.5 db of the re- to failure.
quired input level. Representative PSD's of the
input signal are shown in Figs. 9 and 10. Several methods were considered for de-
termining beam failure (change in tip response
frequency), but none of these methods were ac-
Sinusoidal Sweep curate enough to be used as an on-line failure
indicator. Consequently, it was decided to use p
Four dissimilar beams were attached to the time when the beam actually separated as
the C-100 shaker in the vertical axis of vibra- the failure point. During this part of the test,
tion (normal-to the specimen) and excited by a the shaker was stopped several times, either
sinusoidal sweep from 20 to 1000 cps at a sweep because of "dumping" of the shaker or because
rate of 1 octave/min. The level was 0.1 g O-P. one of the beams had broken. Each time the
The purpose of this sweep was to determine Q's shaker was stopped, the beams were inspected
and frequencies of resonance. It was found that visually for fatigue cracks around the stress

NOTE: 20 - CPS FILTER BANDWIDTH

20 - 2.000 CPS

1.42
1.12
089

0.71
.. &-, IL -IA& . L

0.51

.. 0.36
_< 0.28

0.14

0.11
0089
0.071 , ,,1' ' ! '

0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000
(CPS)
FREQUENCY

Fig. 9. Typical input PSD (20-2000 cps)

282
 5 - CPSFILTER BANDWIDTH
1dOTE:
(20- 200 CPS)

I.I -
1.12

0.71

1~ ~

I" 0.5
0.36

0.23 0 SFR$ 120'

0.11!

0 20 '.0 60 so 100 120 140 t6 laO 20D

FREQUENtCY
(CPS)

Fig. 10. Typical input PS[ (20-200 cps)

concentration holes. The time interval from in a PSD which was then corrected for the tape
observation of a crack to complete failure never recorder response. The values at each beam
exceeded 20 percent of the total test time, thus resonance are listed in the Table 5 under "PSD
*supporting the method used to determine beam (in.)"
failure.
The grms value for the tip response was
During the random vibration test, tape obtained from PSD's generated by the Technical
loops were cut for the produciion of analog Products analyzer using a 20-cps filter. When
PSD plots. The tape loops were cut at the be- the resonant frequency fell in the 20-cps band-
ginning of each run to avoid any effects caused width of the analyzer, the measured level was
by cumulative fatiWue damage. predominately that of the resonant response.
Therefore, the mean-square value of the tip
response was 20 times the PSD value. The
TEST RESULTS AND ANALYSIS rms value was obtained by the equation
The data resulting from this test are given = 2 1/ 2
grms PSD (g /cps) x 20 cps .
in Tables 5 and 6. The data have been listed
according to beam material and thickness. The
parameters measured during the test were the A value for Q was then obtained using the
individual specimen frequencies, the actual fa- standard equations for a single degree of free-
tigue test input values, and the individual tip dom response to a random input. The moment
response of each specimen. The frequencies at the stress riser per inch of tip displacement
were obtained directly from the sinusoidal (W/in.) was calculated using the Myklestad pro-
sweep data; however, some data reduction was v..'am. These values are also listed in-Tables
required to obtain the input and tip response. 5 and 6. These values, the measured frequen-
The input PSD's were obtained by analyzing a cies, and the rms tip accelerations were used
12-sec tape loop on the Technical Products to calculate the an at the stress riser locatior
analyzer using the 5-cps filter. This resulted for each specimen.

283
 TABLE 5
Aluminum Beam Data
Beam Frequency PSD (in.) grM, IM/in. I ,k1 ' ,X0 4 Q N
(a) 1/8-In. Beam
Al-Si 2 54 0.24 47 237 1.44 2.12 108.6 9X10 4
3 53.5 0.26 45 237 1.41 2.07 96.9 7.4
4 54.5 0.29 103 237 3.10 4.55 427.6 5.3
5 54.5 0.29 74 237 2.23 3.28 220.7 6.3

Al-R1 2 56 0.34 77 .237 1.97 2.90 173.4 5.7

3 57 0.34 104 237 2.84 4.18 302.9 9.4
5 57 0.31 48 237 1.33 1.95 73.0 2.9x 10s
AI-S2 2 52 0.30 57 236 1.89 2.78 151.9 1.9X10 4
3 52 0.30 69 236 2.29 3.36 223.6 8.4
4 52 0.30 43 236 1.42 2.09 86.4 5.5
5 51.5 0.30 67 236 2.18 3.32 213.7 4.8
6 51.5 0.28 70 236 2.36 3.47 248.5 6.7
Al-" % 3 48 0.43 40 236 1.53 2.25 62.5 1.6
5 48 0.37 59 236 2.27 3.34 156.2 2.15
6 48.5 0.44 70 236 2.65 3.89 -191.0 2.4
A1-S3 2 46 0.45 46 235 1.92 2.82 93.1 2.2
3 45 0.45 49 235 2.14 3.14 105.8 2.1
4 45.5 0.45 57 235 2.44 3.59 143.9 2.1
5 46 0.45 57 235 2.38 3.50 142.3 3.7
Al-R3 2 43 0.47 50 235 2.40 3.53 125.1 1.65
3 42.5 0.48 33 235 1.61 2.37 52.5 1.95
4 43 0.47 37 235 1.77 2.60 66.7 2.35
5 42.5 0.48 39 235 1.95 2.87 75.0 2 j
A1-S4 5 41.5 0.36 78 235 1.73 2.55 82.5 2.8
(b) 1/4-In. Beam j
A2-S1 2 59 0.33 44 927 1.10 1.73 47.2 6.3x 104
3 58.5 0.31 38 927 0.96 1.51 35.9 2.4x 105
4 58 0.31 75 927 1.95 3.06 154.3 7.5x 104
5 58.5 0.33 114 927 2.90 4.55 330.0 3.4
A2-R1 3 53.5 0.29 62 924 1.89 2.97 163.0 2.3
4 53.5 0.29 62 924 1.89 2.97 161.3 5.5
5 53.5 0.24 45 924 1.37 2.15 105.0 6.2
A2-S2 2 49.5 0.30 52 922 1.83 2.87 144.3 1.5
4 48.5 0.31 48 922 1.79 2.81 123.5 2.8
5 49.5 0.31 40 922 1.41 2.21 80.4 2.5
6 49 0.29 61 922 2.20 3.45 207.3 1.45
A2-R2 2 45 0.30 28 920 1.22 1.92 54.5 6.1,
3 45.5 0.30 28 920 1.17 1.83 50.8 1.7
4 45 0.30 45 920 1.90 2.99 132.5 1.75
6 45.5 0.30 42 920 1.79 2.81 116.7 1.2
A2-S3 2 42 0.29 36 918 1.75 2.74 105.9 1.45 3
3 42 0.48 45 918 2.23 3.50 106.3 4X10
4 42 0.48 59 918 2.90 4.56 177.7 2.2X10 4
5 42 0.48 38 918 1.85 2.91 72.8 1.5
6 42.5 0.48 31 918 1.49 2.34 47.5 1.4
A2-R3 2 39 0.46 43 917 2.45 3.85 111.4 3.9 3
3 39.5 0.36 35 917 1.96 3.07 94.8 6X10
4 39.5 0.47 42 917 2.32 3.65 103.8 4.4 x10 4
6 39.5 0.36 51 917 2.83 4.44 201.1 4.9
A2-S4 4 40 0.36 28 917 1.48 2.32 57.3 4.5

284
 Beam 6Data
SteelTABLE

Beam Frequency PSD (in.) gris M/in. N xl0 4

ox10 4
(a) 1/8-In. Beam
S1-S1 2 60.5 0.38 48 550 2.71 3.99 63.8 1.8X10
3 60.5 0.38 59 550 3.32 4.88 96.5 910 4
4 60.5 0.38 90 550 5.12 7.53 224.4 2x10 5
5 60.5 0.39 75 550 4.28 6.30 174.3 1.25
Si-Ri 2 58.5 0.38 83 448 4.07 5.98 197.4 7.9x10 4
3 58.5 0.38 99 448 4.87 7.16 280.8 1.2x l0 s
4 58 0.38 99 448 4.93 7.24 284.3 1.1
5 58.5 0.38 120 448 5.99 8.80 412.6 1.0
S1-S2 2 52 0.31 64 546 4.88 7.18 161.9 3.6x10 4
3 52.5 0.28 77 546 5.28 8.49 256.9 6.2
4 52.5 0.31 44 546 3.29 4.84 75.8 8.4
5 52 0.31 55 546 4.19 6.16 119.5 1.08x101
S1-R2 2 48.5 0.45 41 545 3.53 5.19 49.1 5.6x10 4
3 49 0.45 45 545 3.77 5.54 57.2 2.2
I-- 4
6
5,
49
48.5
49
0.45
0.38
0.45
59
55
51
545
545
545
5.03
4.79
4.33
7.39
7.04
6.37
100.6
114.6
71.5
4.9
3.3
4.8
S1-S3 2 44.5 0.47 52 544 5.33 7.84 82.4 3.7
3 43 0.47 42 544 4.61 6.77 55.6 4.2
Y 4 44 0.47 69 544 7.37 10.83 146.7 4
5 42.5 0.47 41 544 4.61 6.78 53.6 4.1
6 ; 44.5 0.47 72 544 7.51 11.04 157.9 4.3
S1-R3 2 42.5 0.47 46 543 5.24 7.70 35.1 4.1
4 42 0.47 33 543 3.82 5.62 136.0 1.15X104 5
5 42.5 0.47 61 543 6.96 10.23 46.7 2.7 X10
6 41 0.48 38 543 4.50 6.62 32.7 9
S1-S4 2 41.5 0.36 39 542 4.57 6.72 64.9 3.1
(b) 1/4-In. Beam
52-S1 2 78.5 0.42 65 2165 2.17 3.41 81.6 6.6x10 5
3 77.5 0.42 92 2165 3.13 4.92 165.6 5.1
S2-R 2 54 0.30 41 2150 2.86 4.49 66.1 1.73
3 53.5 0.30 44 2150 3.11 4.89 76.8 1.31
4 53.5 0.28 55 2150 3.89 6.11 128.6 1.73
5 54 0.28 43 2150 2.98 4.68 77.9 1.9
S2-S2 2 51.5 0.31 37 2140 2.82 4.43 54.6 1.65
3 51.5 0.31 39 2140 2.98 4.68 60.7 1.4
4 52 0.29 52 2140 3.89 6.10 114.2 1.25
5 53 0.29 45 2140 3.24 5.09 83.9 1.4
6 52 0.29 53 2140 3.96 6.22 118.7 1.3
$2-R2 2 44 0.24 25 2135 2.63 4.13 37.7 1.5
3 46.5 0.38 22 2135 2.01 3.15 17.4 1.25 4
4 46.5 0.38 31 2135 2.87 4.50 34.6 9x10
5 45 0.27 26 2135 2.79 4.10 47.5 1.4x10 5
6 46.5 0.38 25 2135 2.30 3.61 23.4 1
S2-S3 2 42.5 0.38 32 2135 3.52 5.53 39.3 6.2x10 4
3 42.5 0.41 36 2135 4.06 6.38 47.4 1.lx l0 s
4 41 0.47 28 2135 3.38 5.30 25.9 7.2x10 4
5 42.5 0.41 29 2135 3.26 5.12 30.7 9.4
6 43 0.41 31 2135 3.38 5.30 34.7 8
52-R3 2 39 0.42 29 2125 2.49 5.91 32.7 9
4 38.5 0.38 32 2125 4.25 6.68 44.6 1.1 > 10'
5 39.5 0.42 31 2125 4.00 6.28 36.9 9.4x 104
6 39 0.42 24 2125 3.20 5.03 22.4 2x10 s
S2-S4 2 39 0.36 23 2125 3.03 4.76 24.0 7X10 4

285
 The crn is the nominal stress multiplied by On the other hand, Q is a parameter which is
the stress concentration factor (k f = 1.47 and difficult to obtain.
1.57 for the 1/8- and 1/4-in. beam, respectively)
(Appendices C and D). The a,was then plotted material, the average of the Q's calculated by
as a function Of N for the steel and aluminum, the expression given above reflects a reason-
The aluminum data are presented in Fig. 11.
The curve indicated considerable scatter. The able value for Q of the specimen. If Q is, in
parameters included in the data were natural fact, a constant and the input level and frequency
freeyr) incluled
frequency (f), calculated momthenat ere ntuwere
m om ent per unit tipus n known
th avprecisely,
r g Q woa lgrmh ve
value
he predicted
c u a y
displacement (m/in.), tip acceleration (gusing the average Q would have the accuracy
stress concentration factor (k ), and fl obtained from reducing a tape loop equal in
failure time to the product of the number of specimens
time (T). Of these parameters, only k f and grms times the average time of each tape loop, or ap--
could contain sufficient scatter to completely
overshadow the natural fatigue scatter (10 to 20 proximately 800 sec, which is sufficient time to
percent) as indicated by the data. obtain an accurate value of the rms value.
Other experimenters have found a great
The grms value resulted from measuring a deal of scatter in Q. Although a portion of that
random signal with an effective bandwidth much scatter is due to experimental difficulties (Ap-
less than 1 cycle due to the high Q resonance of pendix E), the available information indicates
the system. Recalling the equation that the above assumption is not correct.
Therefore, the effects of this assumption must
dB 2 =oo be examined. If Q does display some scatter,
AFx T the above procedure will introduce this uncer-
tainty into the grs data. That is, the above
the accuracy of such levels must be low. To procedure largely eliminates an experimental
correct this, the tip acceleration levels must error and introduces the natural scatter of one
be related to some parameter which is more of the fundamental parameters, which is pre-
consistent than the gr s value. The only possi- cisely the desired result. This correction has
bility in this case is to utilize the expression been applied to the data.
grins (v/2) x Qx f xPSD
/ 2
. The value of f has
been defined, and the value of PSD1/ 2 is known Relaxing the assumption that the stress
to a much higher degree of accuracy than grms. concentration factor is independent of stress

x
S0 0I
OX X
Ox "0' " ,...
044 00 D ..
0

[03 ,,

ID' 10

103 101 10 106

CYCLES TO FAILURE
Fig. I1. Fatigue curve for aluminum including recorder correction only

286
 A-

levels results in values which are functions of The first of these approaches has the advantage
the following parameters: (a) the fatigue char- that it relieves the designer of scatter definition
acteristics of the metals tested, including un- in K and ao. All that is required is nominal
certainties; (b) the natural scatter in 0 and the values for these parameters. The desired K can
effect of stress on this variable; (c) the effect be obtained directly from Figs. 12 and 13 by
of high stress on stress concentration factors; taking the ratio of the working stress to the
and (d) the uncertainties in input PSD, which indicated fatigue curves. The results are
include the notch effect due to shaker-specimen shown in Figs. 14 and 15 as functions of number
interaction and the test duration. Item (d) is of cycles and of natural frequency, assuming a
inherent in this type of test and must be toler- 10-min required life.
ated. The remaining scatter results from un-
certainties in parameters which are included in The sinusoidal K factor (Ks) can be obtained
the product of K and Kdyfl.These data are by dividing K by A, which is 2.04 for both spec-
shown in the form of random fatigue curves imen materials (1). For this reason, a K, scale
(Figs. 12 and 13) and as discussed below: has been included in each figure. The correc-
tion for k f was obtained as described below for
<a>,n = K Kdyn ao load. the second approach.

The factors of this equation can be evaluated The probability of a peak stress occurring
three ways: in each (1/2)7 increment was obtained assum-
ing a Rayleigh distribution, and the mean stress
1. K d yn and a 0 can be assumed to have no in each band was assumed to be at the center of
scatter so that all of the scatter can be com- the band. Miner's accumulation was assumed,
bined into K; and it was also assumed that a = 6.

2. % can be corrected for stress ampli- The sum of the probabilities of occurrence
tude effects, and the scatter can all be attrib- in each band times the mean stress value for
uted to K; or that band raised to the sixth power yields an
Vapproximate damage factor. If the calculated
3. Assumptions can be made as to the value of a given a peak exceeds the yield
amount of scatter in ic, allowing the separation strength of the material, then the actual stress
of scatter in K and Kdy n . for that value must be on the order of or less

V. 10 -

95 PER CENT CONFIDENCE LIMITS

:Z- 00 LEAST SQUARE FIT

Iii I103
1

__F__

10,
FATIGUE CURVE FROM MILES THEORY
RANDOM

__ _6

105

CYCLES TO FAILURE
O

Fig. 12. Aluminum test data assuming Q = 135

287.
 - RANDOM MILES' TH
FATIGUE FROM _____X___9__

_e LEAST SQUARE FIT

I Ii
1041 ' - --

IC4 CYCLES TQ FAILURE

Fig. 13. Steel test data assuming Q=95f

ALIFE OF10MINUTES
ASSUMING
FREQUENCY
56 7 8 10 20 30 40 50 60 70 8090100 200 300 400 500I

6.0 3.0-

40- .- _____ _____ 2.0

0.9

K FIT
SQUARE
A LEAST it.7X
___ ___
__ _ _ ___ __ _- 0.6

0.6_ _ _ _ _ __ _ _ 0.3

0.4 - 0.2

0.2 -- - - 0 - .0.1 [(
LIFE INCYCLES

Fig. 14. Aluminum design factor (YK)

288
 -J

ASSUMING
FREQUENCY ALIFE OF10WNUTES
5 6 7 8 910 20 30 40 50 60 70090100 200 300 500 3.0

4 2.0

_ _ _ 1.0
9 PER CENTCONFIOECE'
21LIMITS_ -1.0

K0.7 f: - DAi KS

0.8 --.- - - - .LEAST -0.4

- I FIT
""- SQUARE
0.6 0.3

0.4 " 0.2

0.2L _I I I I 1 10.1
4 s
10 l0

LIFE INCYCLES

Fig. 15. Steel design factor (K).

than the yield strength. The relationship be- 10 percent value to insure conservatism, the
tween yield and calculated rms stress values combined 95 percent confidence level uncer-
of 1., 1-1/2, 2, and 2-1/2 were chosen, and the tainty would be approximately 18 percent for
probability of occurrence for stresses greater aluminum. Therefore, the scatter in Kdyn is
than yield were added to the band just below overshadowed by the scatter in K and in the
yield. The damage factor was calculated as forcing function measurement. The combined
discussed above, and the ratio of this factor to error for the load and K for steel is on the or-
the factor obtained for low stress was obtained. der of 25 percent. Therefore, the error in Kdyn
This factor times the stress concentration fac- is on the order of 40 percent.
tor of 1.5 is shown in Figs. 16 and 17.
A final word should be added concerning
Since the stress probably exceeds yield amplitude distributions obtained from the test
and the method of including the effect was con- data. Figure 22 shows an amplitude distribu-
servative, the curves resulting from the above tion typical of those obtained from tip acceler-
calculations have been shifted 20 percent to the ometer data. Note the agreement with the
right. This new curve was used to correct the theoretical curve (Gaussian distribution) to
data as shown in Figs. 18 and 19. The K and 2-1/2 or 3cr. Figure 23, which is a distribution
K, factors obtained from these data are given obtained from strain gage data, shows about the
in Figs. 20 and 21. same agreement with the normal distribution.
The magnitude of the multi-mode effect is indi-
The third method mentioned above, i.e., cated in Fig. 24, which is a PSD of the strain
separation of scatter, requires a review of the gage time history.
basic equation orm = KKdyn(load) ao" The scat-
ter in aowas due to k f and is largely removed DISCUSSION OF K AND q0
by the above correction. The scatter in K is dyN
limited to fatigue scatter? only. In this type of The prediction of Kdy n and ao is an integral
test, the fatigue scatter should not exceed 15 part of the design factor approach. The calcu-
percent for the 95 percent confidence levels (in lation of co is straightforward except for the
stress) for smooth specimens like the aluminum variation of the stress concentration factor un-
and 20 to 25 percent for slightly rougher speci- der stress. Assuming that all of the uncertain-
mens similar to the steel specimens. The ties are in K, the stress concentration factor
scatter in load is dependent on the accuracy of must be a dynamic factor (Appendix B). If the
the measured input level. It is felt that the 95 variation of k due to stress is eliminated from
percent confidence level uncertainty for this K, then it must be included in a. This required
variable was from 10 to 15 percent. Using the prior knowledge of the environment and Kdyn.

289
 1.5 X \ i

1.4 ~

1.3

1.2 -

Li -I \~~ ____ __

1.1%

1.0 -- '.
. 10 20 30 40 50 60 10

RMS
STRESS
(KSI)

Fig. 16. Variation of kf with stress above yield for aluminum

1.6 2"

1.4

1.3 _

1.2 I,,

1.0
50 60 70
-

80
-

90 100 11O
I
120 130 140 150
RMS
STRESS ( KSI)

Fig. 17. Variation of kf with stress above yield for steel

290
4' -

2 -A

4 -_ __ _ _ _ _

95 PER CENT CONFIDENCE LIMITS

MILES' FATIGUE CURVE

ARE FIT
IAS SQUI
I
1 0 4~L

2 6 8104 24 6 1 2 4 6
LIFE INCYCLES

Fig. 18. Aluminum test data including f~ correction

FIT
LEAST SQUARE
________CONFIDENCE LIMITS

1291
 -,'
--

FREQUENCY A LIFE OFIO MINUTES

ASSUMING

56 17 3'0 4000 70o8090100 200 3 40 51k

-,95PER CNCOFDENCE "

,,-'-'~
-,. --- -- 0.9

K 0.9 K
08
LEAST SQUARE
FIT 0.6
lop 05--
0.4
8"0.3- -
0.2
0.1-

10,
10 2 4 6 8 l04 2 4 6 8io0 2 4 6 106
LIFE IN CYCLES

Fig. 20. Aluminum design factor K including kf correction i

I
_______ FREQUENCY ALIFE or 10IIlNUTES
ASSUMING
78910 20 30 40 o60708090100 2o 300 40&5

o0.9
0.8
0.7
K ~O.S STSVUR
LEA I 0.6 Ks,:
10 0.4- I

8 _ _ _ __ _ _ 0.2
6 __" -- 95PERCENTCONFIDENCE 0.3

2 0.1

10112
103 2 4 8 101 2 4 6 8 105 2 4 6 8 106

LIFE INCYCLES

Fig. 21. Steel design factor K including k, correction

292
 140

120

3a,-

00
-j

\-GAUSSIAN DISTRIBUTION

I0 -

40 50 60 70 30 90 95 98 99 99.899.9 99.99
PROBABILITY

Fig. 2?,. Amplitude distribution

for 1/4-in, aluminum beam

/0

I-
2,y
Fig. Z3. Amplitude distribution _j
for strain gage measurements 5
U..

00

40 50 60 70 80 90 95 98 99 998 999 9999

PROBABILITY

293
 014

IL;

0.01 0!

0.001

0.0001 10,000
0 10 100 1.000
FREQUENCY (CPS)
Fig. 24. Relative PSD for strain gage measurements

The deviations in Kdy may or may not be The scatter in K- n for both these cases is
included in the value of Kdyn* Regardless of primarily due to the uncertainties in Q/"2.
where the deviation is included the mean value This indicated that the uncertainty in Qfor the
of Kd is obtained as shown below. The Kdy n aluminum specimens is much less than 20 per-
for the random cases is cent and in steel is on the order of 100 percent
for the 95 percent confidence limits. These
(-Il'1/2 values are less than those observed by other
Kdyn = "Qf investigators for aluminum or steel.

The Kdyn for the sinusoidal case requires more

discussion. CONCLUSIONS AND
RECOMMENDATIONS
The equivalent sine level must be derived
from random data, and this derivation requires These tests established the feasibility of
some knowledge of Q. If the Q of the system is utilizing equivalent static design factors in ba-
used for the derivation of the sinusoidal load, sic bracket design by postulating the required
then Kdyn for sine (Kdyfs) equals Q. However, factors and then obtaining these factors through
if the sinusoidal load was derived using a Q tests on simple cantilever beams.
different from that of the specimen (Q'), then
the levels will be higher by the factor of the The factors were obtained by direct meas-
square root of Qdivided by Q'. Correcting urement of the exciting force and the system's
this deviation in KdynS yields life. The required damping information was
obtained by a statistical treatment of measured
Q/2 1/2 tip response. (However, the damping, would be
KdynS (QQ) empirically estimated in an actual design ap-
plication.) The information was then combined
This equation is a general expression and in with the predicted stress information, and fa-
the case where Q Q' yields the first expression tigue data were obtained. From these data, a
for KdynS. fatigue factor (K) was obtained which included

294
 only the scatter associated with the natural re- 2. The effect of various common stress
sponse and fatigue plus the uncertainty in defin- risers on both sinusoidal and random fatigue
ing the input levels. The K factor was then re- and the effect of stress amplitude on these
fined by eliminating the scatter due to various parameters;
parameters. This resulted in three K factors,
each of which would be of use, depending on the 3. The definition (by test) of the static
design application. equivalent design factors for complex bracket
systems; and
This same basic approach is applicable to
actual structures, and the degree of tolerable 4. Combination of the above quantities into
scatter suggests the individual parameters
which must be considered in each design. simple design charts to aid the designer.

Therefore, it is recommended that addi-

tiorai work be focused in the following general ACKNOWLEDGMENTS
areas:
This document was prepared under Contract
1. Better definition of damping, the scatter NAS 8-11617, "Optimum Design Procedures for
in damping, and the effect of stress levels on Handling Space Vehicle Vibration Environments."
this parameter for the damping mechanisms The study was directed by D. A. Stewart, with
encountered in bracket structures; D. E. Hines actirg as principal investigator.

REFERENCES
1. J. W. Miles, "On Structural Fatigue Under Vibration," Shock and Vibration Bull. 29,
Random Loading," J. Appl. Mech. (Nov. Pt. 4, p. 1 (June 1961)
1954)
9. B. M. Hall and L. T. Waterman, Correla-
2. B. J. Lazan, "Material and Interface tion of Sinusoidal and Random Vibrations,
Damping," in Handbook of Shock and Vibra- Douglas Eng. Paper 1048
tion (Crede and Harris, Eds.) 10. R. W. Fralich, Experimental Investigation

3. Research on Techniques of Establishing of Effects of Random Loading on the Fatigue

Random Type Fatigue Curves for Broad Life of Notched Cantilever Beams of 4130
Band Sonic Loading, ASD-TDR-62-501, Normalized Steel, NASA TND-663
Oct. 1962 11. Experimental Investigation of Effects of
4. J. C. McClymonds et al., "Research of Random Loading on the Fatigue Life of
Techniques of Establishing Random Type Notched Cantilever Beam Specimens of
Fatigue Curves," Douglas Aircraft Co. 7075-T6 Aluminum Alloy, NASA Memo-
Rept. SM-30461, Nov. 1960 randum 4-12-59L

5. A. L. Eshleman et al., Structural Design 12. P. W. Smith,

Structure With"Fatigue
Ra~ndomTests of a Resonant
Excitation," J.
Acust h Ran. Ei6 n J
for Acoustic Fatigue, ASD-TDR-63-820,
Oct. 1963, Douglas Aircraft Co. Rept. LB- Acoust. Soc. Am. (Jan. 1963)
31354 13. J. E. Foster, "Random-Sinusoidal Vibra-
6r
6. C. Schjelderup, Avtion
Accumulative Damage p. 6 Correlation
(Aug. 1961) Study," J. Environ. Sci.,
Caused by Random Loading, Douglas Air-
craft Co. Tech. Paper LB-738, Jan. 1959 14. L. W. Root, "Selection of Vibration Test
Levels Using Fatigue Criteria," Shock and
'7. V. C. McIntosh and N. Granick, Experi- Vibration Bull. 34, Pt. 5, p. 55 (Feb. 1965)

t ments in Random Vibration, WADC TN

56-228 15. A. K. Head and F. H. Hooke, Random Noise
Fatigue Testing, Int. Conf. on Fatigue of
8. W. D. Trotter, "An Experimental Evalua- Metals, I.M.E. and A.S.M.E., New York and
4 tion of Sinusoidal Substitutes for Random London, Sept. 1956

295
 BIBLIOGRAPHY
Aerospace Structural Metals Handbook, ASD- 7075-T6 Aluminum Alloys and of SAE 4130
TDR-63-741, Dec. 1963 Steel With Special Consideration of the Life
Range 2 to 10,000 Cycles," NASA TN 3866,
Crandall, S. H., Random Vibrations 1956

Crede, C. E., "Failure Resulting from Vibra- MacDuff, N. J., and R. P. Felgar, "Vibration
tion," in Vol. H of Random Vibration Frequency Charts," Machine Design
(Feb. 7, 1957)
Den Hartog, J. P., Mechanical Vibrations
McClymonds, J. C., "Sonic Fatigue Design
Eshleman, A. L., J. D. Van Dyke, and P. M. Analyses," presented at Symposium on Fa-
Belcher, "A Procedure for' Designing and tigue of Aircraft Structures, Dayton, Ohio,
Testing Aircraft Structure Loaded by Jet Aug. 1959
Engine Noise," Douglas Aircraft Co. Eng.
Paper 693, March 1959 Metallic Materials and Elements for Flight Ve-
hicle Structures, MIL-HDBK-5, Aug. 1962
Fisher, E. G., "Theory of Equipment Design,"
in Handbook of Shock and Vibration (Crede Norris, C. H., et al., Structural Design for
and Harris, Eds.) Dynamic Loads

Granick, N., and J. E. Stern, "Material Damp- Peterson, R. E., WADC Tech. Rept. 59-507,
ing of Aluminum by Resonance-Dwell Aug. 1959
Technique," Shock and Vibration Bull. 34,
Pt. 5, p. 177 (Feb. 1965) Roark, R. J., Formulas for Stress and Strain

Grover, H. J., S. M. Bishop, and L. R. Jackson, Timoshenko, S., Vibration Problems in Engi-
"Fatigue Strengths of Aircraft Materials," neering
NASA TN 3866, 1956
Vreeland, R. H., Douglas'Aircraft Co. Strength
ig, W., "Fatigue Tests on Notched and Un- Bull. No. 12 - Fatigue Manual, rev. Feb. 13,
notched Sheet Specimens of 2024-T3 and 1961

Appendix A

COMPARISON OF PLAIN AND NOTCHED SPECIMENS

It was originally decided that plain speci- order to fail the specimens with the available
mens (without stress risers) would be used in shaker, the specimens would have to be fabri-
this test program. The two advantages of these cated of "weak" materials. Aluminum in the
specimens were that a variable k f would be T-O condition and normalized steel were used,
eliminated from the tests and since there would but it was recognized that these are not com-
be local yielding near the failure point in any monly used engineering materials, that their
specimen (with an attendant modification of the properties cannot be closely controlled, and that
distribution of peak stresses), use of a plain due to their ductility, they have the unusual
specimen would allow measurement of this ef- property of a yield strength less than fatigue
fect with a strain gage at the critical section. strength for the life range being studied.

Prototype plain specimens were fabricated New peg point specimens with two round
and tested with the results discussed in Appen- holes for stress risers were then fabricated
dix B. Based on the results of these tests, it with the following advantages:
was decided that the extreme nonlinearities ob-
served with the plain specimens would compli- 1. The stress concentration factors tended
cate analysis of the test data, particularly since to reduce scatter in fatigue test results;
it was planned to correlate test results with a
linear response and fatigue analysis. A further 2. The double concentration factor doubled
disadvantage of the piain specimens was that in the number of possible crack initiation points,

296
 iii
III

Fig. A-2. Specimen with

Fig. A-1. Plain s.pecimens stress concentrations

which tended to reduce scatter in the fatigue The irregularity of the failure process for the
data; plain specimens is illustrated in Fig. A-1. Note
that there are at least two out-of-plane crack
3. nucleation points, thus causing a very irregular
of actual bracket design; and failure surface and considerable difficulty in
defining failure. In Fig. A-2, which shows the
failure surface for the specimen with the stress
4. The stress concentration factor pre- risers, note the regular crack growth from
vented the drastic reduction in Q observed in each of the crack nucleation points. In these
the former specimens at the fatigue stress specimens, general failure quickly followed
point, formation of the first crack.

Appendix B

LINEARITY AND DAMPING CHARACTERISTICS OF SPECIMENS

Before initiating design and fabrication of specimen. Peg point specimens were originally
the final specimens, preliminary tests were fabricated without stress risers. Later it was
run on prototype specimens referred to as peg decided that the final specimens should have
point specimens. It was hoped that these speci- stress risers. However, the results of the peg
mens would have life characteristics similar to point specimen tests are still of considerable
the median of the final production specimens interest because of the insight they provided
and that their dynamic and structural character- regarding structural damping and fatigue phe-
istics would be similar to a typical production nomena. The configuration of the first peg point

297
 An analysis of the damping characteristics
of the beams was then conducted to determine
if nonlinear damping was one of the causes of
the large discrepancy between the predicted
and measured failure levels. Originally, damp-
ing had been estimated from Trotter's data (8).
Prior to testing, damping of the specimens was
* ~- I experimentally determined from resonant de-
cay tests. Finally, damping was again measured
in low-level sinusoidal sweeps. In every case,
Qwas established to be between 65 and 130.
The design test level of approximately 0.2
g 2/cps was determined based on those esti-
mates of Q (the stress in the beam under ran-
dom excitation is proportional to Q). When in
actual test the failure level was established to
be closer to 2 g 2/cps than to the design level,
nonlinear damping was investigated as a possible
cause. Lazan (2) has shown that in the stress
range above the endurance limit, the damping
.9 ratio is proportional to a high power of stress.

With this in mind, damping was again esti-

mated using the procedure suggested by Lazan
and available narrow band (f = 0.2 cps) power
spectral analyses. Tip-to-root amplification
ratios were used to determine damping during
the high-level (1.2 g 2/cps) tests.

Theoretical damping was calculated using

Fig. B-1. First peg point specimens
the following steps:
1. Estimate a "fatigue steength at 108 cy-
cles," o' for each material- 1.3x 104 psi for
specimens is illustrated in Fig. B-1. When 2024 aluminum and 4.7X 104 for the 4031 steel.
these beams were vibrated to their design level,
several things were observed: 2. Determine the stress (during the test, a
stress two times the rms" stress was used).
1. There was a sizable reduction in beam
response as fatigue was accumulated. This is 3. Determine the specific damping energy,
illustrated by Fig. B-2. This plot shows the D, from Lazan [(2), Fig. 36.17].
relative responses of a typical beam as failure
is approached, obtained from third-octave band 4. Determine the loss factor for a uniformly
analyses of the response data. It represents loaded specimen from Eq. 36.14 of Ref. 2.
data from the beam end accelerometers which
have been plotted as a function of time before As a result of this analysis, it was decided to
failure. use specimens with stress risers. There were
many advantages in this type of specimen (Ap-
2. Increases of input levels did not cause pendix A). However, prior to fabricating the
the expected increases in tip response levels, new specimens, an analysis was conducted to
For example, a doubling of the input PSD value estimate the damping of the new specimens.
at fatigue levels resulted in a 7 percent increase To eliminate any nonlinearities and also to pro-
for the 6-in. aluminum specimen. This indi- duce sufficient response to fail the beams in a
cates a highly nonlinear Q in the test region. reasonable time, Q's equal to at least 50 were
This also indicates that the peak distribution desired for the second set of peg point speci-
may be drastically altered. mens. An analysis confirmed that the Q's
would be in this range.
Because of these effects, lives of the first peg
point specimens were far above design life. Figure B-3 shows the stress distribution
Actually, levels four times higher than the de- in the vicinity of a round hole in the tension side
sign level were required to fail the beams. of a plate and an analytical expression for this

298
 PER CENT OF TEST DURATION

j0 20. 40 60 80 100

dB

Ii ~ ~~~~-15
___ ______

Fig. B-2. IReduction in rms acceleration

vspercentage of test duration

iid
d_

Fig. B-3. Stress distribution nearPSCPRSIN

round hole in plate in tension 0

0
R ..0 (2 d

299
 rL

DAMPING
SPECIFIC
ENERGY
FORPLAIN
~~~SPECIMEN iCURVE C'

08 18
0.6 -. 6 NOTE: CURVES
UNDER
AREAS B ANDCARE

Uj TOTAL DAUPING WITH

FORSPECIMENS
SAND WITHOUT CONCENTRATIONS
STRESS

SPECIFIC
DAMPING
ENERGYFOR -
WITH
SPECIMEN STRESS
0/% (CURVE
CONCENTRATION Bi

(CURVE
( A)FORSPECIMEN
Uj
0.4
N/ 1.4
WITH
STRESS
CONCENTRATION

0.2 1.2

00 1.0

0 02 03 04
CONCENTRATION (IN)
DISTANCE FROM CENTER OFSTRESS

Fig. B-4. Comparison of damping energies of

specimens with and without stress concentrations

distribution. Based on these data, the stress power of stress. Damping energy is shown in
distribution was determined near each of the Fig. B-4 (normalized to the values resulting
stress concentrations for the second peg point from the str3ss at the failure point). The area l
specimens. Note that the holes are sufficiently under curve B is proportional to the total damp-
separated from each other and from the edge of ing energy for the specimen with stress con-
the beam to minimize interactions (confirmed centrations, and the entire area is proportional
by the photoelastic analysis described in Ap- to the total damping energy for a plain specimen.
pendix D). As the area ratio is at least 10:1, only one-
tenth of the material damping in the plain speci-
With this stress distribution, the specific mens is expected in the specimens with stress
damping eriergy for each small volume of ma- concentrations. Q's should, therefore, be near
terial in the vicinity of the holes could be deter- the desired value of 50 when both material
mined from Ref. 2. At high stresses, this damping values and damping values from other
damping energy is proportional to the eighth sources are considered.

300
 Appendix C

STRESS CONCENTRATION FACTORS

There is a sizable difference between static k = 2.23,

and fatigue stress concentration factors, k. and
k f, respectively. An expression for this differ-
ence, presented by Peterson (C-i) and used by k tb = 2.43
Smith (12) is
These values can be corrected to the bending
kf = (k t + a/r)/( I+ a/r) , case by finding the given values of k , on the
a/h = 0 curve of Fig. 86 of Ref. C-I and then
where a is a material constant and r is the dropping down to the proper a/h curve and
radius of the cut in the material. The a for reading the values. This results in values for-
both the steel and aluminum used in these tests plate thickness of 1/8 in. of
was given by Peterson as 0.05. Therefore, the
=
remaining task was to establish the static stress k ta 1.67,
concentration in bending for the specimens.

The static stress concentration factors k =

were obtained from Peterson in the following
way. Since a good deal of the required infor- and for plate thickness of 1/4 in. of
mation was to be obtained from tension curves
and since the value in bending for a sheet with 1.92,
a/h = 0 corresponds to the tension case, the
a/h = 0 was used as a standard where a = 2r
=
and h = thickness. The value of kt for two ktb 2.04 .
holes in an infinite plate under tension normal
to the line drawn between the hole centers is The maximum k t is ktb; this is the only factor
given by of interest, therefore, for correcting fatigue
yield:
yed
'm.. B ( 1 - a/b)
k (1-a/b) kf = (1.84+0.05/0.0625)/(1+0.05/0.0625) = 1.47

max A =for the 1/8-in. specimen and

t3.02.k f = (2.04 + 0.05/0.0625)/(1 + 0.05/0.0625) = 1.57

The ktb factor should be correct; however, the for the 1/4-in. specimen. These are the k
k ta factor obviously must be changed to account values used in the design of the new peg point
for the edge condition. beams; they are reasonably close to the value

The k t. factor for a hole the proper dis- determined experimentally by Smith (12).
tance from the edge of an infinite plate is ob-
tained as k t, = 2.23, which should be nearly REFERENCE
correct for this case. Therefore, the following
stress concentration values in tension were C-1. R. F. Peterson, Stress Concentration
determined: Design Factors

Appendix D

PHOTOELASTIC TEST RESULTS

A quantitative analysis of the stress distri- The experimental analysis utilized the bi-
bution around the two stress concentration holes refringent coating technique. One side of the
in one 6-in. aluminum beam under various aluminum beam was coated with a layer
static loads was performed both analytically (0.118-in. thick) of photoelastic plastic (Budd
and experimentally. Instrument Company Type X.C.) (Fig. D-1).

301
 -- Fig. D-1. Specimen for! ." -_
~~photoelastic analysis -

Fig. D-Z. Test setup for

photoelastic analysis

kg = stress concentration factor based

on gross section;
The beam was clamped at one end and illumi- a , . ==maiustesthoeede
m x m m sr s thl d e
nated with white polarized light. A known static
load was applied to the other end, and the =
strained bcam was viewed through a polariscope anet net section stress;
lens (Fig. D-2). Under this condition, the dis-
tribution of the actual strain in the beam was a oss,= gross section stress;
seen as a series of black and colored fringe =
patterns. Figure D-3 indicates the changes in net net section moment of inertia;
the strain for different static loads. These = }
photos also show that there was very little in- Igos, gross section moment of inertia; i
teraction between the stress concentration
holes, even for large static loads. AverageM=moet i
various loads fringe
isochromatic are plotted
ordersin determined
Fig. D-4. for c = distance from neutral axis to ex-

treme outer fiber;

For more quantitative results, the magni- FOre, = -.verage slope at hole edge; and
tude and direction of the principal strains were
=
determined, using the Budd Instrument Com- FOnom interpolated nominal slope be-
pany large field polariscope (Type LF/Z-U). tween holes.
Data from this test follow:
2. Equations:

1. Symbols:
k t =stress concentration factor based k a-tion factormbas
on net section; snet

302
 3L 6 LB 9 LB

12LB 15LB 18LB

Fig. D-3. Fringe patterns showing stress distribution around holes

303
 RUN 2

5.01 1

- L
U. 3.0
FOmax 0.1943 FRINGE/0OUND
CME

- 2.0

1.0 I

5 10 15 2e 25 30
LOAD (LB)

Fig. D-4. Average isochromatic fringe

order (FO) vs load at hole edge

kg =ross= M-c = Mc 3. Calculations:

k aV C'max Oax igross
g Wc
Igross First run kt 0'188 x 0.8 1.98;
0.076
Mc kg
-max 'gross 0. 1943
Mc kg Ine FOma Ine t Second run kt - 0.2 x 0.8
0.082 1.90; "
kt = 9g n e~g
t =kg x net - mx x net
Igross Igross FOnom Igross
Mc kt = 1.94 - 0.04

Appendix E

DIFFICULTIES IN MEASURING STRUCTURAL DAMPING

As the intent of this study was to evaluate where material damping predominates over
techniques for designing for vibration environ- other sources of damping.
ments, estimation and measurement of struc-
tural damping were of concern, Several tech- w h e tuied icd e fomoing
niques (listed below) were utilized at various
times during the study, but all these techniques 1. Resonance decay tests;
have fairly serious shortcomings - at least
when dealing with lightly damped structures 2. Sinusoidal sweep tests;

304
 3. Random excitation- (a) Narrow band slow to allow full response to be achieved in a
power spectral analysis, and (b) broadband lighily damped system. The slowest sweep rate
power spectral analysis. commonly available in automated equipment is
about 10 min/octave. A Q of 30 is the maximum
Decay tests at resonance and sinusoidal that can accurately be measured with this sweep
sweep tests present difficulties when dealing rate.
with material damping which is highly nonlinear,
particularly when the stresses are above the Even random excitation poses some diffi-
endurance limit of the material. Obviously, to culties. At high stresses, damping may increase
determine the damping in an actual application, as fatigue damage is accumulated. On the first
comparable stresses must be achieved in test- peg point tests, response in one case decreased
ing. As an extreme example, on the first (plain) by nearly 20 db prior to failure. This agrees
peg point specimens, Q's from low-level tests with Lazan's results (2). Even without this
ranged from 65 to 130; under high-level random time dependence, however, there are two prob-
excitation, Q's ranged from 6 to 15. On the lems when dealing with high-Q systems:
second peg point specimens, Q was determined
as a function of input level for high-level sinus- 1. Even with a massive shaker armature
oidal sweeps. The results are shown in Fig. and relatively light specimen such as used in
E-1. An additional difficulty with sinusoidal this study, a high-Q system may introduce a
sweeps is that the sweep rate must be extremely significant notch in the input spectra (Fig. 8).

80

70

60

63---

'.<
0 50

I-

U-

C-C
40
U

30

20

023 45
INPUT LEVEL Wg

Fig. E-1. Measured Q as function of input level

305
 - -

2. If an attempt is made to correct this by thus eliminating any possible time savings
comparing both input and response PSD's, an
exceptionally long averaging time is required
to obtain an accurate PSD. (This averaging
achieved by using wide filters.) For a system
with f0 = 50 Q = 100, averagingtime for 10 per-
cent (1db) accuracy with 90 percent confidence is
V
time is controlled by the ,filter bandwidth or the 400 sec - a significant portion of the life of a
resonant bandwidth, whichever is narrower, specimen such as was tested during this study.

DISCUSSION

Mr. Gertel (Allied Research Assoc.): Did such as a stress concentration factor which is
you make any attempt from your data to see if not as well defined as it could be. We were con-
you could obtain a constant for summation of cerned with short-term life, such as is encoun-
cycle ratios that are used in the cumulative fa- tered in space missions. We were up fairly
tigue damage calculations, or was that the fac- close to the knee in the fatigue curve. But
tor k? Miles' approach using 1 for the k factor in the
k alpha for the slope did yield a conservative
Mr. Hines: We included all of the fatigue value. It corresponded closely to the 90 per-
information that we obtained experimentally in cent confidence level of the data.
the k factor, and went no further than to try to
define a k factor and the error or scatter in it. Mr. Root (Collins Radio Co.): I believe
that Miles' analysis is based on the Palmgren-
Miner summation. You have already shown
Mr. Gertel: In the past work that has been this in one of the figures in comparison with
done on cumulative fatigue using various dam- the Miner summation.
age concepts, the most popular being the linear
or Miner's cumulative damage theory, the sum- Mr. Hines: I think he was more interested
mation of cycle ratios is usually set equal to in the k alpha term where alpha is the recipro-
some constant factor for convenience. It would cal of the slope, in what k value should be used,
seem, perhaps, that this constant might be and if it could be obtained from the data. The
found from your experimental data for the par.- comparison I made was based on a k of 1, but
ticular sets that you did. we were at a bad place on the fatigue curve to
make any real comparison. We did have other
Mr. Hines: Possibly it could. One may be factors in it. I would not want to use that and
able to get some basic fatigue information from say that you should use 2-1/2 or something
data ouch as this, but one has things introduced else.

* * *

306
 AN APPLICATION OF DECISION THEORY
TO A VIBRATION-FATIGUE PROBLEM
D. M. Aspinwall and R. E. Blake
Lockheed Missiles and Space Company
Sunnyvale, California
I
Statistical decision theory provides a model for the orderly considera-
tion and weighing of a number of diverse factors and pieces oi evidence
which must be brought together in making a decision. As a realistic
example of such a procedure, we discuss the problem of whether a cer-
tain Polaris missile structure will have acceptable fatigue strength
throughout several years of exposure to submarine vibration.

Available choices are: (a) to increase the design strength, with a pen-
alty in weight and cost; (b) to accept the design as is and risk the de-
velopment of service failures in the future; or (c) to spend more money
on improving the data. The choice will depend on the relative values
(utilities) which the project leader and/or customer place on these al-
ternatives and oa the risk of failures.

Since the discussion is aimed at a vibration engineer who specializes

in a technical field rather than in setting project goals, we emphasize
the problem of communicating his technical judgments in a form best
used by project management for making decisions. The vibration spe-
cialist would like to predict the probability of failure in service. But
this cannot be done with precision because the available data and theory
lack precision. This lack of precision is too large to ignore by using
nominal values or to envelope by use of conservative "worst" assump-
tions. Decision theory provides for estimating and keeping tabs on the
various sources of error.

The format required for this decision theory analysis is a probability

density of reliability. The essential elements in defining this curve
are the parameters of the probability distributions for load and
strength. A procedure for combining engineering judgments with ex-
perimental data on these important factors was used to develop the
probability density for reliability.

INTRODUCTION Such a dilemma was presented by the fail-

ure in fatigue of a major structural part of the
Standardized specifications and tests for P3laris reentry system during a MIL-STD-167
resistance to shock and vibration are familiar vibration test. The test was known to be rather
features of most military and space design conservative for application to a complete mis-
projects. They expedite the procurement, sile system, but preliminary analysis had sug-
screening, and passing of hundreds of hardware gested that no problem would occur. Because
designs with production-line efficiency. De- of schedule in the Polaris program, the part
signs which fail to passcan usually be changed had already been developed and used in several
without too much trouble, or the test can be re- ground and flight tests, A redesign would in-
vised to be more realistic for the specific hard- volve scrapping much development time and
ware and its anticipated environment. But oc- flight reliability data.
casionally a problem comes up which is too
complicated, too important, and too expensive Thus the problem was to examine critically
to be disposed of in any easy fashion. all the environment and strength information
3I0
 1-

related to "thispart to decide whether redesign apparent after the development program was
or acceptance would be the better cource of ac- well along. The structure was found to resonate
tion. The decision is, of course, the responsi- at a frequency generated by the submarine pro-
\! bility of the project management, since it de- peller, and the fatigue strength of the part was
pends on cost, effectiveness, and schedule not quite adequate to pass the standard vibra-
tradeoffs of the missile system. It is the task tion tests. These facts would have been impos-
of the shock and vibration engineer to provide sible to foresee by analyses; late discoveries
the project with a clear and understandable pic- of this type are one of the risks of a parallel-
ture of the technical side of the problem. This effort development schedule. The usual solu-
usually involves trying to convey the results of tion of increasing the strength of the part was
years of experience and study in a short inter- made difficult by flight design and schedule re-
view. Theory is difficult enough to explain, but quirements. To satisfy these requirements with
engineering judgment, intuition, and uncertain- a redesigned part would require a repetition of
ties are seldom communicated satisfactorily, ground development and flight tests, or a de-
crease in confidence in flight reliability.
In the problem to be discussed here, a rel-
atively new approach is used for presenting and Another hope for a quick solution was to
assessing the shock and vibration picture. This work out a substitute for MIL-STD-167 that
is based on the premise that the project's real would be more realistic for a Polaris missile
concern is with the risk of failure in service in a Polaris submarine. Review of actual data
and the dependability of our information and indicates that this probably would have elimi-
predictions. They listen to our troubles about nated the difficulty if it were not for the reso-
data, analysis, test simulation, and statistical nant amplification. So the situation remained
variations only to get a feel for the dependabil- one of considerable uncertainty, with the uncer-
ity of our information and methods of handling tainties difficult and expensive to resolve.
it. Thus the aim of this study will be a predic-
tion of the reliability that the part would have Not only were there many technical uncer-
in service, and a description of the accuracy of tainties as to the vibration environment and
the predictions. fatigue strength, but also these quantities were
highly variable statistically. Samples of data
The analysis follows the model provided by would not pin down the details of these statisti-
statistical decision theory. The theory simply cal populations. They would only allow esti-
shows how various facets of a problem can be mates to be made. It would not be meaningful
handled separately and reduced to a common to ask whether all or none of the parts would
denominator, so that they can be fitted together fail in service; one should ask for the probabil-
logically and consistently. Decision theory is ity of surviving, i.e., for the reliability.
much like systems analysis and some aspects The reliability of a missile system is one
of game theory; it has been explained and dis- of the key parameters related to its military
cussed in several references
fromsysemsanalsisin (1-6).
hatit It differs
tkesaccunt efetiees
effectiveness. Inreera te costfeiveness
In general the cost/effectiveness
from systems analysis in that it takes account ofasteimrvssthpyldan,
of a very important but elusive factor in making a reiblt inres as the ost ande-
a good decision: uncertainty. Uncertainty is a and reliability increase, and as the cost and de-
rather dirty word; it is generally taken for velopment time decrease. Since any decision
granted that no engineer with real talent is con- concerning this part will have its small effect
taminated with it, and no engineer with real on several of these factors, the decision is a
courage will admit to it in public. But it is al- good one if its net result is to improve the cost/
ways with us to some degree, and it usually effectiveness. Evaluation of the relative worth
costs money and effort to reduce that degree. of increments of cost, reliability, weight, etc.,
Dealing intelligently with uncertainty consists is the objective of systems analysis (5,7); judg-
of determining whether a prospective reduction ment of whether the consequences of one course
in uncertainty is worth the effort. This too is a of action will be better than another is a func-
problem handled within the framework of deci- tion of project management. Decision theory
sion theory. Decision theory is not discussed provides a way to avoid putting th? shock and
directly here; it is used. The purpose is to vibration specialist on this unfamiliar ground
show by example how it differs from and im- by asking him whether he recommends that a
proves on the aernative approachesn part be redesigned. He should only be asked
what is the risk of failure in service. The proj-
ect should be asked how high a risk of failure
PROJECT DESIGN PROBLEM would justify a redesign. Both should be asked
how accurate their answer is, and what would
The fatigue sensitivity of the reentry sys- be needed to improve it. If these answers do
tem was due to some coincidences that became not make the choice an obvious one between

308
 (a) acceptance of the design, (b) changing the the component distributions and then combining
design, or (c) gathering more- information, de- them. There are a number of approaches which
cision theory provides a means for closer can be applied, each with its advantages and
analysis. each with its own degree of accuracy.

The project engineer has the task of evalu- The principal stumbling block at this point
ating the alternative courses of action that is usually the uncertainty as to the form of a
might be taken and their influence on system distribution (such as normal, log-normal, or
goals. This can be very difficult, and system Weibull). Occasionally, as in the case of fatigue
tradeoffs can be hard to discern after being strengths, there has been enough research to
converted into the artificial black-and-white predict the mathematical form, or one may un-
language of contracts, requirements, and spec- derstand the causes ef variation well enough to
ifications. But the fact remains that the accu- make a shrewd guess. In other cases, one may
racy required of the shock and vibration engi- have little information and may follow the herd
neer is limited by the accuracy with which the in assuming a normal distribution. In any case,
project knows what it should have. Both groups there is a great danger of getting tangled in the
have to improve together, and it is not efficient intricate structure of mathematical statistics.
, for one to throw the burden of extreme accuracy Exquisite formulas and 8-place tables are often
onto the other. The remainder of this discus- used, but they are no better than the basic as-
" sion concerns the task of the shock and vibra- sumptions behind them. It is more reasonable
tion engineer: to assemble his information into to use mathematical approximations in line with
a prediction of reliability and an assessment of the accuracy of the rest of the decision analysis.
its accuracy.
The mathematical procedures and computer
program described in the Appendix were used in
RELIABILITY this study as the primary means of combining
the distributions and formulating the reliability
The information related to the reliability prediction. The program is based on the distri-
(probability that the part will not fail in service) butions being normal. In some instances, when
falls naturally into two categories: (a) the load - it was known that the distribution was not nor-
factors in the external environment tending to mal, an equivalent normal distribution was sub-
produce failure; and (b) the strength - proper- stituted which best matched the actual distribu-
ties inherent in the structure tending to resist tion in the range of most significance to failure.
failure. The load L and the strength S can be Actually the natural logs of load and strengths
expressed in the same units, with L as a meas- were taken as normal. This avoids the embar-
ure of t' ..pplied ehvironment and S as the rassing fact that normal distributions can in-
magnitude of the environment required to cause clude negative values, but strengths cannot. A
failure of the structure. With these definitions, check indicated no great difference between the
the reliability is simply the probability that a results of assuming normal or log-normal
missile selected at random will have a value of distributions.
S greater than the values of L which it experi-
ences; that is, S > L, which is equivalent to Some other approaches to calculating reli-
ability have appeared in the literature. In one
S- L > 0, study, the mathematical distribution of S/L was
(1) worked out for the assumption that s and L are
S/L > 1, each normally distributed. In another report, a
or computer program was described which gives
or the distribution of any mathematical function of
(log S - log L) > 0. several parameters, when the distributions of
the parameters are defined in any way.
The load and strength are independent statisti-
cal variables. The functions (S-L), (S/t), and The simplest approach to defining reliabil-
(log S - log L) each have statistical distribu- ity, appropriate to a preliminary study, can be
tions of a different form, but of course they based on two theorems of statistics: (a) the
each contain the same reliability prediction. mean of the sum (or difference) of two variables
is the sum (or difference) of the means, and
Just as the distributions above are com- (b) the variance of the sum of two independent
posed of load and strength distributions, the variables is the sum of the variances. One thus
loads and strengths in turn may be the resultant concerns himself with the means and variances
of several effects. Thus, the general problem of the distributions without regard for their dis-
of predicting reliability consists of predicting tribution. On finding the mean and variance of

309
 the combination by simple addition, one can find Thus the acceleration could be expressed

what the reliability would be if the combined as the product of several factors:
distribution were normal. Combinations of
prior knowledge and statistical data can also be x = P1 T 1 Tn2T 23 R, (2)
made with approximate hand computations.
Vj These procedures were used in the early stages where P I is the resultant hydrodynamic force,
of the present study. Tin is the hull transmission factor to the loca-
tion of the nth launcher, Tn2 is the transmission
from the hull to the launcher, T2 3 is the trans-
FACTORS AFFECTING mission from launcher to basic missile struc-
LOAD PREDICTION ture, and R is the amplification factor from
basic missile structure to R. To describe i
The vibratory stresses and loads on theal one hte
must define the statistical distributions of
sonhergtidofE.(.
structural member of interest are proportional all the terms on the right
g side ofEq (2.
to
in the
the center-of-gravity acceleration of the body
missile. This acceleration Data concerning submarine vibration is
9 is the basic understandably scarce. One measurement has
load variable factrs
and is affected by several other
Fig.1).been reported of a basic missile structure vi-
bration, that is, of the product Pi 1 I Tn 2 T 23 .
The amplification factor R has been the subject
The source of the vibration is in the of nine measurements, and two measurements
propeller-blade interaction with the hull; these have been made of x.
hydrodynamic forces will vary with speed,
number of blades, etc. The vibration response During the several years of operation which
of the hull, a combination of bending and tor- the missile is required to survive, the subma-
sional modes, will differ somewhat even for rine will operate at a variety of propeller
submarines of the same class. The response speeds. The only one of concern here is the
of the missile launch tube to the hull vibration one causing a frequency of 16.5 cps.
will also be somewhat variable because of un-
intentional structural variations. The location The remaining information needed is not in I
of the launcher in the hull is also a random the form of "objective" numerical measure-
variable from the point of view of predicting ments. It is available only as "educated
where a given missile will be stowed. There guesses" based on experience and/or theoreti-
will then be a transmissibility factor for the cal analyses, and extrapolations based on as-
vibration response of the basic missile struc- sumptions (which are also educated guesses).
ture to the launcher. Finally there is a near-
resonant amplification of I from the basic mis- In using "subjective" assumptions and es-
sile structure. timates, one has the problem of defining the

/7

I_
kiliI

Fig. 1. Sketch of submarine and equipment

310
 dependability and completeness of the source of The problem of expressing the "subjective
information. This is analogous to the problem probabilities" about dependability of information
of using samples of data from a statistical pop- is considered first. These subjective state-
I ulation
ture as a basis
sampling. forstatistician
The predicting makes
resultsaof"1rea-
fu- ments
sity are made
functions forinthe
themeans
form and
of probability
variances den-
of
assumption of the type of statistical the important load and strength variables. The
distribution, and then calculates "confidence means and variances of load and strength vari-
levels" for various predictions. A confidence ables are the descriptive parameters being es-
level is the probability that one has not been timated, rather than the loads and strengths
misled by a very small or unusual sample into themselves; subjective probabilities must,
making a false prediction. therefore, apply to the mbans and variances.
Some probability density functions are defined
The confidence method contains the implicit by equations which contain several parameters,
assumption that no relevant information is known but only the mean and variance are needed to
A-,except
V the numerical data and the type of dis- define a normal distribution, which is one rea-
tribution. This is an assumption more easily son for its wide application. For example, if
satisfied by a statistician than an engineering the variable w has a normal distribution, then
specialist with several years' experience. In the probability density is:
engineering work, it is often necessary to com-
en2 statistical data, theoretical analyses, and
bine '
subjective knowledge. To combine these quan- (4)
titatively, the subjective knowledge will be ex- f ... e
pressed in probabilistic language suited to a ' a
combination, which will formally resemble and the required parameters are and aw.
}Bayes' Rule of inverse probability (2). This
does not require that subjective opinions "really Attention will now be focused on the prob-
are" statistical variables, although one 'could ability distributions for the variables r and t,
argue with considerable force that numerical where t = P 1 + t In + + t23. Subjective
+tn2
predictions made by a population of people form probabilities will first be derived for the means
a statistical population. and variances of r and t; then these functions
may be combined to give a subjective probabil-
' . Although probability will be used to de- ity density for u and o,,. In this work, the
scribe both the statistical fluctuation of uncon- probability functions for I At I"r , and <
trolled variables and the dependability of infor- are described by an equation similar to the
mation, the two will be kept separated and normal probability distribution, and all param-,
termed "probability" and "subjective probabil- eters are independent. Then the probability
ity," respectively. This distinction is impor- functions for u-, and o-2 are also normal and
tant for deciding whether more information
should be obtained. I --Pt
+
It and ax2 ar2 + 2
(5)
x r (5

The normal density function used to describe

ESTIMATES OF PARAMETERS these parameters is not a theoretical necessity,
For convenience in predicting the statisti- but
sumsit of
is variables
a convenient
aremathematical
concerned. form where
cal properties of X, we note that if a quantity is
a statistical variable, its logarithm is also. Yn Considerable data were available on the
Eq. (1), R is the product of several factors; so response factor, r, because this variable could
log R is the sum of the logs of the factors. So, be measured in ground tests of different re-
representing the logs by lower-case symbols, entry systems. Some typical amplification
curves are shown in Fig. 2. The sample aver-
log X =x p +tn 1 +tn2+t 23 +r. (3) age and variance for r are F = 1.26 and S. 2 =
0.0458. Because these data do not define R, but
Equation (3) contains several variables for just one of the factors, the results are used to
which no separate information is available; make judgments about the subjec tive probability
therefore, probabilities will not be estimated densities for Ar and ar. The subjective prob-
for these variables individually, but only for ability densities can be used to describe past
their combined effect. The equation does indi- experience, analysis, or data interpretation.
cate how complicated the expression for x is When data are directly applicable to the vari-
and warns the engineer that wide variations may able of interest, they can be incorporated di-
be expected in data for the base vibrations and rectly into the subjective densities in a mathe-
the body cg acceleration. matical operation (see Appendix). When data

311
 IIT

U) .

/ZI 1-
0

/0 f
S/ 16 7 18
13/ /R 20

ZEXCITING FRE-QUENCY (cps)

Fig. 2. Typical transmissibility curves for

the region of ;.nterest

must be related by judgment to the variables of /

interest, a more intuitive approach mustbe used. f- e _0'
2 - (6).s .

Tables of sampling distribution for nine

pieces of data were used to aid in this process. 2
Tbese tables showed that F should vary within [ -oos
1.26 0.07 two-thirds of the tin~e, and s2(a e 7 ]()
should vary within 0.0458 0.03 two-thirds of =
e t2<
the imeeperenc
l~evios wit amlifca- Only one piece of data was available on the var-
tion factors indicates that the mean should be fb et h a ea cl rto o h yt mi
somewhat xigher, although with nine pieces of the submarine. The subjective probability den-
data, the subjective probability should be tai- sities for the mean and variance of this vari-
lored to correspond closely to the sampling able were based primarily on engineering judg-
distributio)ns. ment. The functions chosen were
The Chi-square tables fb ,sample var -I
ances, s 2 , from a normal population give a t '
s
curve whibh plots like a log-normal distribu- . \ ]J (8)
tion for higher values of s . The higher valuesf(/t e,
Of S2 are of more interest because they tend to
lower the reliability; so the standard deviation, (%0. 1 2
o.' 0.03 for fI(s 2 ) was estimated from this Chi- - \ -. /(9)
)
square plot on log-normal paper. f 4(O e 2 0

After taking all the information into account, Note that variances for the estimates of
the following subjective probability density func- b t a a ee sa el r e . so eti d o
tions* were adopted to describe a, and ar 1.59, the' best estimate of pt, and 0.2 is twice as
large as 0.1, the best estimate of ar . These
*The coefficient 111Y,, a has been omitted; this large numbers are reflections of the complicated
has no effect on later computations. origin of t and of the lack of data concerning t.
312
 Our objective in deriving these subjective FACTORS AFFECTING STRENGTH
probabilities for 'ar, ar2' At , and a2 is to PREDICTION
write similar equations for u. and a. 2 . From
Eq. (5) and the fact that the probability density Although there were tests to failure on
for the sum of two normal variables is again twelve full-scale pieces of hardware, the fatigue
normal, the following equations may be written: strengths remain subject to several uncertain-
ties. To obtain results in a reasonable time, the
p_-1. 3+1. )2 tests were run at a higher-than-expected stress
- o (10) level. The test points shown on Fig. 3 must be
fs(A) e 0. 51 , extrapolated out to the vicinity of the 2.5 mil-
lion cycles estimated for a full service life.
-o (11) A further complication is that static
f 6 (,a) .e2 / stresses are added to the vibratory stresses.
The static stresses are highly unpredictable,
With these subjective probabilities defined, ac- since they are caused by bolting together two
tual data on X may be considered. parts which cannot be made to fit perfectly.
Strain gages were used to measure stresses in
Two pieces of data on X were available, a critical fillet region, but the experiments
The values for tn X were -0.94 and -0.41. The were handicapped by the shape and a compli-
sample average is cated stress distribution.
-0.675, (12) To extrapolate the data of Fig. 3, the best
method is to use the shape of a fatigue curve

and the sample variance is established by tests of laboratory specimens.

Figure 4, copied from Ref. 8, shows that no sin-
2 = gle curve has been established. Even in tests
sx 1406. of carefully made small specimens, fatigue
strengths exhibit a large statistical scatter.
All the information concerning the load varia- One can anticipate even greater scatter in large
bles is sunmmarized in Eqs. (10) through (13) in production parts.
a form which can be used in a computer pro-
gram which will eventually be used to compute A study of some other fatigue curves show-
a probability density of reliability. ing the influence of static stress and stress

0 T(ST DATA
- 0 EXTRAP0I1ATiD DATA

(3

0- TOLERANCE

9AI

0 /0 /0 CYCLES /0

Fig. 3. Fatigue strength test data

313
 II

6"0000 V"1 \
-\000

300\

1420000 ..

/00o0-, t "+ ""

3 + " a
/0 /40 /0 /0 /0 /0 /0 /09

CYCLES

Fig. 4. Fatigue data on notched aluminum specimens

concentration factors produced a very fortunate perfection for a insignificant difference. Each
result. Although these factors had consider- missile costs millions of dollars to build and
able influence on the magnitude of the fatigue deploy; a 1 percent chance of loss is equivalent
strength, they had small effect on the shape of to a very large sum of money.
the extrapolation curve. Thus, the remaining
problem was to select a nominal method of ex-
trapolation and estimate its probable error. MEANING OF THE RESULT
This was done by judgment based on the avail-
able curves. Figure 3 shows the extrapolation The output of Fig. 5 could be explored by
of only one of the points to its nominal value at integrating the curve and noting the percentage
2.5 million cycles. The one-sigma tolerance C of the area under the curve above various
band on the extrapolation is also shown. values of R. The value of c then is interpreted
as a generalization of the statistical confidence
The data were treated as if there were only idea; it is the probability that the data and ex-
7 readings instead of 12. The reason for this is perience used have led us to be correct when
that the computer program as written did not we say "the reliability exceeds R." Since the
provide for uncertainty as to the exact value of word probability has had various controversial
a test result. This uncertainty arose from the uses and definitions, its meaning here can be
extrapolation problem; in other problems, it summed up by the advice: "If our usage seems
might be due to an inaccurate meter, etc. From to stretch the meaning beyond your favorite
judgment as to the extrapolation error and by definition, make your decision as if it did not."
consulting appropriate statistical tables, it ap- A fuller explanation is given in Refs. (1), (2),
peared that using 7 instead of 12 readings would and (4).
have the equivalent effect on the computer out-
put. This subterfuge is, of course, not the most The probability density of reliability sums
direct approach, but it was the shortest path up how much we know and do not know about the
around this unforeseen obstacle, prospects of fatigue failures in service. If it
seems too vague and inaccurate, it is because
the information available is vague and inaccu-
COMPUTED RESULT rate. The method of handling the information
is intended to increase the vagueness to a lesser
The output of the computer is plotted on extent than any alternative. Many of the alter-
Fig. 5. The probability density of reliability is native methods give an appearance of precision
a highly skewed curve indicating that the actual by using a single number to i.eplace a probabil-
reliability is numerically close to. 1. One should ity density, calling the average, the best esti-
not mistake the small numerical difference from mate, the unbiased estimator, the maximum

314
 /40

120

"T/00

t 80

' ~60
4J

40

01
.98 0.99 /00
RELIABILITr

Fig. 5. Probability density of reliability

likelihood, a conservative value, the mode, etc., of winning, by the judgment that the prospects
"the" value. In effect, one is being advised to are of greater value than the other uses that
make his decision as if he believed the single one might make of the dollar. The probability
value to be correct. Decision theory puts the density curve is the best statement we can make
computation of an "equivalent-single-number" from the available evidence of the projects
on a rational basis. chances of winning the various possible prizes
or penalties, if it chooses to acce'pt the design
Since the various possible results (values as is. The alternatives are to . - the cost of a
of reliability) of using the design in service redesign or pay the cost of better information.
have such a wide range of importance to the
system effectiveness, the most satisfactory If the above forms of presentation of re-
evaluation consists of stating a numerical sults fail to interest the customer, one might
measure of worth (utility) as a function of the abandon some of the subtleties and state the
reliability. This is the procedure called for by risk of failure of a single missile. If a missile
decision theory. It enables one to compare al- fails in service because we failed to beef it up,
ternative designs and to set a value on getting it makes little difference whether the cause was
more information or data. It also permits one an exceptional statistical aberration in the en-
to replace the probability density of reliability vironment, the hardwaie, the data on which the
by an equivalent reliability, that reliability design was based, or a mental aberration of an
which would have the same worth to the system engineer. The first two relate to reliability;
as the "expected worth" of th-3 set of uncertain the latter two to confidence.
reliabilities. This is much lke the judgment
that the value of a ticket to a certain raffle ex- The average, or expected value, E(R), of
ceeds one dollar. One replaces the various the reliability is obtained from the probability
prizes, their value to the bidder, and chances density of reliability p,(R) by the integral

315
 .1 confidence value, 99.6 percent (the median).
E(R) F
f Rp(R) dR. (14) The difference expresses the price that we are
0o paying for insufficient data. If the p(R) curve
gives us a 70 percent confidence that the relia-
For Fig. 5, E(R) is 99.3 percent. The risk of bility is in fact satisfactory, this means that
failure is thus 100 - 99.3 = 0.7 percent, or 1 there is a 70 percent chance that enough infor-
chance in 143. It must be repeated that this is mation to dispel our uncertainties would prove
a perfectly good statement of the risk that any the design to be satisfactory as is. Lacking I '
single missile faces, but this risk is not quite this proof, we must act as if the equivalent re-
what one should be interested in. If 1430 mis- liability is the true value.
siles were deployed, the E(R) above would be
the actual reliability, if we had perfact infor- As mentioned earlier, if one is dissatisfied
mation, and we should expect 10 failures. Or with the uncertainty due to lack of precise and
there could also be a 0.7 percent chance that all voluminous information, he is free to try to ob-
would fail. These are not equivalent, tain mo:e information, at a price. Dezision
theory provides a technique (1) for comparing
The expected reliability is the same as the the price of more information to its expected
equivalent reliability mentioned oariLtc, if one worth. When further investigation is not con-
assumes the worth of the system is proportional sidered worth the price, one should then choose
to its reliability. Such an assumption is not a course of action, accepting the residual risks
likely to be valid over a wide range of values of and hedging against them if possible.
reliability. But if one has narrowed the range
of p(R) sufficiently, E(R) may be a useful ex- Although this problem originated some time
pression of the result. ago in the Polaris Missile System and the set-
ting has been preserved to make the discussion
It is significant that the p(R) curve (Fig. 5) seem more realistic, the numerical values have
is not symmetrical and the expected value of re- been arbitrarily chosen and do not apply to any
liability (the mean) is lower than the 50 percent real missile system.

REFERENCES
1. H. Raiffa and R. Schlaifer, Applied Statis- Decision Theory," LGckheed Missiles and
tical Decision Theory (Harvard U. Press, Space Co., Rept. SD/R-9, Aug. 24, 1964
Cambridge, Mass.) 1961
6. D. M. Aspinwall, "Bayesian Statistics and
2. J. W. Pratt, H. Raiffa, and R. Schlaifer, Decision Theory in Engineering," Lock-
J. Am. Stat. Assoc. (June 1964) heed Missiles and Space Co., Rept. 801071,
Nov. 13, 1961
3. A. M. Mood and F. A. Graybill, Introduc-
tion to the Theory of Statistics (McGraw- 7. C. J. Hitch and R. N. McKean, The Eco-
Hill, New York) 1963 nomics of Defense in theNuclearAge (Har-
vard U. Press, Cambridge, Mass.), 1960
4. R. E. Blake, Shock and Vibration Bull. 31,
Part 2, pp. 88-97, Oct. 1962 8. Metallic Materials and Elements for Flight
Vehicle Structures, Dept. of Defense,Wash-
5. R. E. Blake, "Optimum Meteoroid Shielding ington, D. C., M1L-HDBK-5, p. 3.3.1 (g),
of Spacecraft- An Application of Statistical Aug. 1962

Appendix
MATHEMATICAL PROCEDURES
The computations which one might make to Company. This program provides the capability
predict reliability will vary in cost and accu- of performing the many computations, although
racy, depending on how efficiently one chooses requiring that all probability distributions be
his simplifying assumptions. In the example of normal. One can evade this restriction to some
this paper, it was decided to take advantage of extent by vari.ous subterfuges as discussed in
a digital computer program prepared in 1963 the paper, such as letting the log of the varia-
for such problems by D. M. Aspinwall and J. E. bles be normally distributed.
McFeely of Lockheed Missiles and Space
316
 - -- - - -

Reliability can easily be computed for nor- very similar, and no subjective knowledge is
mal (or log-normal) distributions. If L is load required or provided for.
and S is strength, then reliability is probability Since Bayes' Theorem can be used even
of survival
With and probability
each normal distribution (S - L) by
thatdefined > 0.
its when one's subjective knowledge is very slight,
cand variance
ithmean 2
, the difference (S-L) its result should then agree with that of fiducial
will beannormalncith2mean
will be normal with m ean s -diffandcvariance)
L and variance probability.
z r uj ciThe
e kmost
o l plausible
d e o way
a a toe express
e s b
2
as + The reliability is the area of the
02. zero subjective knowledge of a parameter is by
probability density of S - L above zero; that is, a constant probability density over the entire
parameter range. This yields agreement with
2
E .L fiducial probability for 1 but .notfor o2 . This
-S-L-,s- L apparent disagreement can be resolved, if one
R 0 d(S- L) (2( s22) /2J (A-i) considers that the choice of a linear scale f1'r
R V (2 is completely arbitrary. One could just as
0 / well have chosen some other function f(a 2 ) to
express the spread of sample measurements.
The values of the four parameters would It turns out that f(a 2 ) = 1/a 2 will satisfy both
thus fix the value of reliability. But the param- approaches. A flat subjective distribution of
eters are not accurately known; their values 11 2 corresponds to an odd distribution of a 2 .
can be predicted only by appropriate probability But if one objects to the odd distribution of a2,
statements. The computer program calls for he is claiming to have subjective knowledge of
the subjective prediction of each parameter in what a2 could reasonably be expected to be.
the form of a normal distribution. Note that we Hence, he should use this knowledge to con-
are involved in normal distributions at two lev- struct a better prediction.
els: (a) the statistical sampling distributions The combination of subjective and statisti-
of loads and strengths defined by four parame- cal information is effected by Bayes' Theorem:
ters, and (b) the distributions of these parame-
ters used to express the chance that a predic-
tion that a parameter is in a certain numerical pM p(X) P(datX) (A-2)
interval will be correct. There are, thus, eight N
subjective parameters to be selected; the means where pa(X) is the subjective probability den-
express a best estimate, and the variances ex- sity of a parameter X, P (data IX) is the proba-
press the probable error of the estimate. bility of observing the estimator if X is given,
In addition to one's subjective N is a constant to satisfy the normalizing re-
Nowledge,
kee nlodge quirement that p(X) integrate to 1, and p(X) is
there are experimental data on loads and the resulting probability density of x. Then
strengths. These data can be summarized with- tersligpoaiiydniyo .Te
out loss of information by the estimators of the ,p(X) is the product of two curves, as is illus-
trated in Fig. A-1 for the parameter ar2. It is
mean and variance of the data, R and s 2 . The
relation of such estimators to the mean and easy to see that if one of the peaks is broad
elementarytstatisticspobooks;on is given in compared to the other, the narrower peak
variancewill dominate the result. In such a case, the
elementary statistics books; R- 1 is normal, exact shape of the broad peak is not of much
and s 2 /l 2 is proportional to the Chi-square ctnshaence
distribution. These distributions state the prob- consequence.
ability of observing a value of the estimator if With the eight parameters of the subjective
the parameters of the population are known, probabilities and the four estimators of the load
whereas our concern is to predict the parame- d.nd strength parameters, the computer then
ters from knowledge of the estimators. This calculates the probability density of reliability.
problem of "inverse probability" has been The primary need for the computer is due to
solved by two different approaches: by Bayes' the fact that the four variables which define re-
Theorem and by Fiducial Probability. liability must be considered simultaneously.
Furthermore, the probabilities of the estima-
Although Bayes' Theorem has been ac- tors 5 and s 2 are not independent. The proba-
cepted for certain classes of problems since it bility distribution of 5? is normal with a vari-
first appeared in 1763, it has been only recently ance, a 2 /n, where n is the number'of samples.
that its use with subjective probabilities has Since a 2 is not known, the computation involves
been shown to be, justified (2). Fiducial proba- the joint probability distribution of R and S2 for
bility is substantially the same as the method given pairs oi I and a2. This is best handled
of confidence intervals, in that the proofs are by the computer.

317

V%
 a)UB5 TIVE PROBABILITY

0A I U8 / ' R

0 0.025 0.05 0075" o.10

LOCATIOEN OF-
20 [ ) ,, PRO8AS/DITYF 'NC TION O
4 L

20

0 0.025 0.05 0075 0.10

IL

DIS CUSSION O

Mr. O'Hearne (Martin Co.): You said that Mr. O'Hearne: It seems to me that when
you could be very arbitrary in selecting your you mix what you actually experience with your J.
a priori density function. It seems to me that preconceived notion, you are doing something a-
you have taken this preselected density function, little subtle. I would not be arbitrary about it .
weighed the density actually experienced with at all.
that density function, and used the derived den- / s0 :
sity. I think you would have to be prepared to Mr. Aspinwall: I do not mean that it should
defend your a priori judgment very strongly, be taken lightly. One has to do the best he can
whereas you seem to indicate that that would and use all of his experience and judgment. But
not be necessary. one must plot a curve, and it must be stated in
Mr. Oe
spiwal:
ha todefnd hatnumericalo form.
L. In that sense, it is arbitrary.
judgment just as one has to defend any engi- Dr. Jaenke (AF Missile Development Cen-
neering judgment. A good engineer will have a ter): Did you try to specify confidence levels
more accurate estimate of the parameter that I for your estimates in this work? This might
was discussing or any other parameter. He has be an answer to the question which was just
to defend it against criticism, and I think it is asked.
everyone's duty to make the best estimate he
can. What I meant to imply was that it is an Mr. Aspinwall: No, we did not specify the
individual matter. Your a priori density func- standard confidence levels for each of these
tion may be different from mine, and no one parameters because there were so many in-
can say I am right and you are wrong, or vice volved. The final response was the sum of the [
versa. The better engineer should have the logs of these factors. Then with the load and
better density function, but we can not tell who strength with which we are concerned, the dif-
was right until we have flown the missiles to ference gives the reliability. We could specify
see whether or not they fail. confidence intervals for each variable, but I
318
think it is meaningless to try to specify stand- are using a similar approach in an entirely dif-
ard confidences for the reliability. It is too ferent area. It is again an area where statisti-
complicated. The advantage of the decision cal evaluation and estimation is involved, and it
theory approach is that the uncertainty in each is quite surprising how much this combination
variable can be included in each step, and the of a priori and experimental investigation really
computer program yields the final probability helps in solving a problem.
density curve which combines all of them. Mr. Aspinwall: I agree. This is the heart
of the technical presentation of data. I did not
Dr. Jaenke: I was interested in your tech- go into the utility function, but as far as esti-
nique of combining a priori estimates with the mating is concerned, I think this is the heart of
experimental observations. In our activity we the decision theory approach.

319
 GENERAL SOILS MODEL FOR
SHOCK PROPAGATION STUDIES
C. X. C. F. Miranda
Department of Civil Engineering
University of Detroit
Detroit, Michigan
and
David C. Kraft
Department of Civil Engineering and Engineering Mechanics
- University of Dayton
Dayton, Ohio

A review is made of the currently accepted method for describing the

one-dimensional stress-strain relationship for soils under shock load-
ing. The implications of the currently accepted model for describing
soil behavior are discussed with regard to its application to granular
and cohesive soils. Based on experimental evidence presented, a new
and more general model is proposed to represent the one-dimensional
stress-strain relationship of all soils during locking behavior. The
proposed model is a general one and can be used for shock propagation
studies in soil media.

INTRODUCTION the shock wave propagation phenomena, it is

necessary to define mathematically only that
The study of one-dimensional shock propa- portion of the curve which is concave to the
gation phenomena in a nonlinear medium has, stress axis.
as its starting point, the one-dimensional stress-
strain relationship for the soil. It is, therefore,
necessary to formulate thir, relationship in such NOMENCLATURE
a manner that it reflects as nearly as possible
the observed behavior of the soil, and yet not be A0 , A1 , Coefficients
so complicated that it hampers analytical evalu- A2 ...
ation of the shock propagation phenomena.
AL, Au Constant characteristics for stress-
Recognizing this, several investigators strain curve of the current model for
(1, 2, 3) set out to develop such a relationship, loading and unloading phases
In this context, two well-defined problems are
at once evident. The first concerns the defini- HL Area under loading curve
tion of what should be the correct procedure of
obtaining experimentally the stress-strain curve; Hu Area under the unloading curve
the currently accepted thinking on this subject
seems to be the use of an oedometer (4). The L Subscript used to indicate loading
second problem, with which this paper is pri-
marily concerned, involves the selection of a Ms Secant modulus
mathematical model that would approximately
describe the experimentally obtained curve. Mr Tangent modulus
Results of both static and dynamic oedometer
tests indicate that the stress-strain relation- u Subscript used to indicate unloading
ship is generally concave to the stress axis for
both cohesive and cohesionless soils. Fre- aL, bL, Constant characteristics (nondimen-
quently, the initial portion of the curve exhibits au , bu sional form equivalents) for stress-
an S-shaped behavior which is concave to the strain curve of proposed model
strain axis at comparatively low stress levels
(2,4). From a practical standpoint of studying f Nondimensioralized stress

321
 n Constant characteristic for stress- 77 Relative size of hysteresis loop
strain curve of current model po D
Density o medium
of i

,n2 Constant characteristics of stress- a Stress

strain curves for loading and unload-
ing (current model) L Stress during loading cycle
07%ax Maximum stress, also denoted by po
p(t) Impulse loading function
s Normalized strain a Stress during unloading cycle

so Normalized residual strain

MODEL CURRENTLY USED
c Strain
The nonlinear hysteretic model first sug-
emax Maximum strain gested by Hendron (5) has been used by other
investigators (6,7). The details of this model
6 Residual strain are shown in Fig. 1. Also the mathematical

oax or Po

O L = ALE ni

n
6 TU = AU (E-) 2

o max

Axial Strain, a

3.0

MT/MS = n, for loading

0MT/Ms = n2 for unloading

S1.0
1.
Also: ni n2 = n
0
01

Dynamic Axial Stress,0

Axial Strain,Et

Fig. 1. Model currently used

322
 implicatiofis of the model are shown in the
appendix. It is shown in this appendix that for 250o
the model in question the following relationships
should hold: 2000

1. The ratio MT/Ms is independent of the

stress level, where MT is the tangent modulus , 1500 --

at the stress a, and ms is the secant modulus

at the same stress o-.

is 2. The relative size of the hysteresis loop,

77, is given by 5of/,
Z LOADING RATE--O001 IN
>- 'fI MIN.
6%/eax, (I0o o eo=O.54
0 0.004 0.008 0012 0016 0.020-

x (a) AXIAL STRAIN, E, IN/IN

where e.ax is the strain occurring at a -

and E0 is the residual strain on unloading at 0.030

or
As is indicated later, the implication that
MT/Ms is independent of the 6tress level obtains 0.o1.
only in the case of sands. The second implica- < oll
tion regarding the relative size of the hystere- -I

Ito
sis loop appears to be a rather restrictive re-
lationship for all soils, and it may be difficult
find a theoretical basis for justifying such a
relationship.
-
< 0.003 _

76.0
_ -

(b) DYNAMIC AXIAL STRESS, o-, PSI

& UNLOAD

EXPERIMENTAL EVIDENCE FOR .-

SANDS AND OTHER SOILS1
64.-
Figure 2 presents results for a Minnesota P A
sand (5). Figure 2a of this figure shows a 206 * i &Ia A
typical stress-strain curve for this sand, while 0 LOAD
Fig. 2b suggests that the curve can be adequately A U

described by an exponential form of the type of 000.20.40.6 0.8 1.0

Eq. (A-1). Further, a plot of the variation of (c) NC2:DIMENSIONAL DYNAMIC AXIAL STRESS, o'/P0
t) based on an analysis of the stress-strain
relationship of Fig. 2a, and as shown in Fig. 2c,
Ii suggests that for thisMparticular sand, the ratio
of~g Hendron
E for Minnesota sand
xpenfrinetaodta sand
Of MT/MS, at least during the loading cycle, is
very nearly independent of the stress level.
Similar conclusions, for both the loading and
the unloading cycles, could also be drawn for Figs. 3c, 4c, and 5c, which show that the varia-
sands studied by other investigators (4,8). It tion of MT/Ms is indeed a function of stress.
could thus be said that the currently used model
is a fairly good representation of the stress-
strain curires for sands. In fact, the model was PROPOSED NEW MODEL
developed from data on sands.
Since the variation of MT/Ms is not independ-
However, a suggestion has been made (7) ent of the stress level, one may in the most gen-
regarding tue use of this model for studying eral form assume
wave propagation in all soils. Figures 3 to 5
present results of one-dimensional tests for MT =A + Aa + A
three different'soils, other than sands (9, 10 M 0 1 2 (2)
11). Figures 3b, 4b and 5b indicate that plots
of stress versus strain on logarithmic paper
do not yield straight lines, thereby indicating But going back to Figs. 3 through 5, it can
that an exponential curve is not an adequate be seen that this expression can be, without much
representation. This is further confirmed by loss of accuracy, simplified to

323
LC
, 750

80 /0
b

60
./
450
U .100200 .4 .50)

aSrl /79.%"

00 0 1 002 0.03 () AILSTRAIN, E, IN/IN

( AXIAL STRAIN, E, IN/IN- -

V I Ao
171 II

0. 300
1.00 I0.00 00.0 .o4 -il &.N ,IAI
o.00--I- IIll .. I
I
0 ( S, E.- I

(ba)DYNALYM
AXIAL STRIS W ,
-;j- -I

_- _, IT.
2.C OLOAD
a )a UNLOAD 0 LOAD
0.00 ITT 0.0 o A UNLOAD

0.0
0.0AUNLOA 0.4 0.60. 1.00.0.2 04 0. .8 LOAD
10
( DYNAMIC AXIAL STRESS,
(bN)IMNIA 0 ( (b)NMN/OA DYNAMIC AXIAL STRESS,
', P 0
(c NNIMNIA DYNAMIC Ms
AXIAL
0-0 STRESS, (3A.-d 00.2I00.4I0.60. 1 1.0 0.
-A A -AD.6
fo Utlbrgclyfo Gos ak.ca

I= zr Ao A
In_som se t beg dsab i tosde- 1.0 AA I-0 Ao 6

! fine the variation of MT/MS by two,0 Aor NOD0LA

at best Further defining
A UNLOA
three, linear segments.
,; f = omax = nondlmensiona stress,
Now, recalling that
"? S = /Emax = normalized strain, and
MT = do-/dc '(4)
and -? aL, bL = nondimenslonal equivalents of
A and
o A1 for the loading cycle,

Ms = a-Ic, (5) one bbtains

324
 6000 ~~1a
LOADING RATE STATIC (auibu)if (9)
t 4800
8 -1-0 w11.7%
Sri22.3%
-sS0 [f0Ubf
L - (
3 where
El U)
-j

x 2400 =
E0 o
/i So 6 max
XC

I200 ---
Equations (8) and (9), then, define the pro-
posed general model for all soils.. The model
0 Q5o.I) .i .2, h2g is developed essentially from the experimental
(a) AXIAL STRAIN, E, IN/N evidence that MT/m s is not generally independent
0.30 of the stress level, but is, in most cases,of the
form I + Azo. The model can be used for ex-
tension of wave propagation and other studies
from sands to other soils.
2' 1I
-- It is interesting to note that the following
"- -models can be generated from the proposed
model by appropriate selection of the coeffi-
t ~~~ OAD cients a and b and of the residual strain so:
(02[-"
Q UNLOAD I

I, I I 1. Elastic nonhysteretic model;

(b) DYNAMIC AXIAL STRESS, 0, PSI 2. Linear model
8.0 - - -na hysteretic (12);
f 3. Model as used by Weidlinger and
6.0- -- h Matthews (6);
A.0-- 4. Linear form of the Weidlinger and
S, - - -Matthews (6) mode; and
A
0 c,,
A A0 -A 5. Ideal locking model.
a OLOAD
*
O - A UNLOAD

0.2 0.4 0.6 0.8 1.0 APPLICATION OF MODEL FOR SHOCK

(c)NONDIMENSIONAL DYNAMIC AXIAL STRESS, AoPP ATIONOFMDEL
PROPAGATION STUDIES
Fig. 5. Experimental data from Davisson
for Playa silt The stress-strain relationship, as ex-
pressed by Eqs. (8) and (9), can be conveniently
4 used for shock propagation studies. The model
jVhas been used by the authors (11) to study shock
~df/ds phenomena in a nonlinear medium using the
f/s a + bLf, same assumptions as were used by Weidlinger
a su and Matthews (6). Taking into account these as-
sumptions, Eqs. (8) and (9) become:
s = 1.00, yields

S Lb~
+aY
(aL + bL) f
7

l/aj
/I/a

(8)
L
=~~a L
aL + bLf
b
(aL+bL)fl/j
J (10)

s= So . (11)
For the unloading portion of the curve" the
derivation remains essentially the same, with The details of the development of the equa-
the only exceptions being that s in Eq. (6) is to tions governing shock propagation phenomena
be replaced by (c - i 0 ), and the integration is are given elsewhere (11), and only the resultant
carried out subject to, the f = 1.00 when s 1- so , equations are summarized below to demonstrate
leading eventually to the applicability of the proposed model:

325
z

1. Shock velocity iihen-reduced for -the condition bL = 0.00 (i.e.

nr/ms = constant), can be shown&to be idential to
i = r Po'I 2
/. f \ 1/2 the equations of Weidlinger andt Matthews (6).

_L +/bbj~ f
(L+ CONCLUSIONS

The limitations of the currently used model

2. Shock location have been discussed in the context of their pos-
sible use as a general soils model. Based on
Streported experimental evidence, it has been
pt shown that the model is not adequate for this
/f ) 1/2 a
opo)-
(Po [( L + bL)f] I/ 2 L , (13) purpose.

aL + bLf The new model now proposed is in agree-

ment with experimental. evidence and lends it-
where po is the density of medium and p(t) is self to incorporation in shock wave propaga-
the loading Junction. Equations (12) and (13) tion analysis in soils.

REFERENCES

1. R. V. Whitman, "Stress-Strain-Time Be- 7. L. Seaman and R. V. Whitman, "Stress

havior of Soil In:One Dimensional Coin- Propagation In Soils," Stanford Research
pression," M.I.T., Dept. of Civil Eng., Inst,, Final Rept., Part IV, DASA 1266-4,
Res. Rept. R63-25, May 1963 June 1964

2. M. T. Davisson, "Static and Dynamic Be- 8. D. E. Calhoun, "Interim Report on Con-

havior of a Playa Silt in One-Dimension strained Modulus Testing," unpub. rept.,
Compression," Kirkland AFB, New Mexico, U. of New Mexico, Air Force Shock Tube
TDR-63-3078, AFWL-TDR-63-3078, Sept. Facility, 1964
1963
9. T. Ginsburg,
in the "Propagation
Ground," of Div.,
J. Structural ShockASCE
Waves
3. A. J. Hendion, Jr., "The Behavior of Sand (Feb. 1964) J
in One-Dimensional Compression," Ph.D.
Thesis, U. of Ilinois, Urbana, 1963 10. H. Kane et al., "A Study of the Behavior of
a Clay Under Rapid and Dynamic Loading
4. R. V. Whitman, E. T. Miller, and P. J. In the One-Dimensional and Triaxial Tests,"
Moore, J. Soil Mech. and Found. Div., Kirkland AFB, New Mexico, AFSWC-TDR-
ASCE (1964) 63-3116, Jan. 1963

5. A. J. Hendron, Jr., "The Behavior of Sand , 11. D. C. Kraft, C. Miranda, and Delmar E.
In One-Dimensional Compression," Kirk- Calhoun, "Shock Propagation Phenomena in
land AFB, New Mexico, AFSWC-TDR-63- Nonlinear Media," submitted to Soil Mech.
3089, Oct. 1963 and Found. Div., ASCE, 1965

6. P. Weidlinger and A. T. Matthews, "Shock 12. M. G. Salvadori, R. Skalok, and P. Weid-

and Reflection Phenomena In a Nonlinear linger, "Waves and Shocks in Locking and
Medium," presented at Ann. Structural Dissipative Media," J. Eng. Mech. Div.,
Eng. Conf., ASCE, Oct. 1964 ASCE (April 1960)

326
 I :JAppendix.
IMPLICATIONS OF CURRENTLY ACCEPTED MODEL
Referring to the loading cycle of Fig. 1, nl=n =n. (A-9)
Swhich Isdescribed by
Now, defining HL as the area under the loading
7 % ale' , (-l) curve,
curve, we Hu as the area under the unloading
andhave:
one oa= the As-1o
Sone obtin the expressions for MT and ms as [%,=OLd =! A nd= AL n+1
# HL -f ALe de = n-+ ,,.,1 (A-10)
' M T = da AL n j t'-
das (A -2 ) 0.

and

us - - L '" , (A-3) ' sx

HU = cde= - 6) de
leading to + e o (A-11)
MT j Au ,"+ i+
,,
sL (A-4) n+1 (emax - ,6o)

This indicates that the ratio MT/Ms is independ- but

ent of the stress level. The same conclusion
can be drawn from the unloading cycle described kmax
AL Emax =AA( - e0 ) (A-12)
by
Hence
" "HL u =AU(6 6 0 )2 (A-5)
= + 6max x 0 ax ' (A%-13)
provided the secant modulus is defined with
reference to the residual strain, %,inwhich and
case

(A -6) HUHuAu ( co) x

,max- ,max
. (A -14)
f (E O)n2 1
M6 Thus, the relative size of the hysteresis loop,
++- 7, becomes ML Mo
2 (A -7)
MS 5 AU( E C o)
77 HL - HU
HL

and
Mmax max ( E0)
0max-max
MT r +1 n+
j S = (A-B) %ax (A-15)

n+1
An assumption often made that the size of
the hysteresis loop is independent of the maxi-
mum stress has been reported as coinciding
with experimental evidence (5). This assump-
tion leads to the conclusion that - __.
max

327/
 ~DISCUSSION

SMr. Hall (Edgerton, Germeshausen & Dr. Kraft: I will answer your questionin
\ J Grier): oktyesilad
Has any testing:"een
x accomplished
o i hreayassumed three parts. First of to
the medium all,beitais true that we
continuum. have
This is
reason
ona why your studies
ock-type could not"be
soil n:,:ir:o,:s applied
th:e an the assumption generally made, at least up to
to the same? this point, in the studies that have been conducted.
Secondly, with regard to your. questioning the
Dr. Kraft: No tests have been conducted variations in density and moisture content which
on rock-type soil. The tests that we conducted may occur with depth, this has been dii'cussed,
while at the Air Force Shock Tube Facility although not in great detail, in the other paper
were on a silt material. Everything presented to which I referred. It was found that the vari-
in the paper wal. on either a sand, a silt, or a ations in moisture content and density did not
clay. There is ho reason, however, why it can make a considerable amount of difference on
not be extended to a rock-type material. Gen- the stress attenuation with depth. A range of
erally speaking, the behavior for rocks might moisture contents and densities were analyzed
be very close to elastic behavior. There with respect-to depth. Finally, you mentioned
would be no question of going into greater detail the problem of loading and unloading. A nuclear
as we have here for extreme variations in the blast wave is a one-shot type of loading, not a
stress-strain curve. continuing loading and unloadihg phenomenon.
In truth, we are trying to study wave propaga-
Dr. Sevin (lIT Research Institute): I tion with regard to hardened structures, in
respectfully suggest that your suggestion that other words, the transmission of one blast wave
this could possibly be a stress-strain relation- to that hardened structure. So, I do not think
ship, applicable universally to all soils, is not that the loading and unloading phenomenon is
really a valid one. I think there is much data particularly pertinent unless You have to go
which have not been quoted, which on one extreme through a number of loading and unloading cycles
suggest that perhaps the continuum model is to obtain soil properties which would reflect
not adequate for the representation of all soils, wave propagation. Does this answer your ques-
At the other extreme there is much data by Whit- tion?
man, Selig and others which suggest we do not
really know how to characterize soils in terms Dr. Sevin: Partly. I would refer you to the
of such properties as densities and moisture Wilson-Sibley paper in an ASCE Journal of about
content. These data indicate that a simple re- 2 years ago. This reports a good deal of explosion
lationship, no matter how complex it might ap- data and laboratory tests reflecting the properties
pear in equation form, could not treat the shock of real soils of various types from silts to a rock-
propagation problem in an adequate sense. First like medium as a function of depth. These data
of all, the one-dimensional treatment is in ques- show at least an order of magnitude change in the
tion. In the real world where the one dimension confined modulus with depth. Secondly, my ref-
goes down, we have a very prominant variation erence to load, unload, reload was more in ref-
of the confined modulus. Even on a one-dimen- erence to strain rate reversals due to various
sional model with depth, the might
fariation
be possibilities such as reflections, loading, and un-
a factor of 10 in modulus within, say, the first loading. I am not necessarily implying that you
40 feet - or perhaps the next 100 to 200 feet, have to unload to the zero condition, reload up
The work that Selig has published suggests that and so on. It was with reference to the blast
the loading, unloading and reloading-unloading phenomenon, of course,
behavior Is not reproducible. If one attempts,
'however, to treat the stress-strain curves
analytically on a load, unload, load, unload, re- Dr. Miranda: I want to stress one particular
peated cycle, one can get reasonably close to comment which Dr. 'Kraft has made, referring
the predominant characteristics. In summary, particularly to the variation of a and b with
I would think that a model could not possibly be depth. The changes with depth can be reflected
adequate for all soils in one dimension if it did by the variation in the values of a and b with
not include the variation of the modulus, i.ot depth. Now a study.of this type has been made
only as a function of stress intensity, but as a and we have carried out an analysis to verify
function of position. It should also treat the that, although there is variation in a and b,
load, unload, reload-unload characteristics, quantitatively it does not make very much differ-
Again, the one-dimensional model itself has very ence in the variation of stress with depth. We
serious limitations for real world applications, have verified this by an actual calculation.

328