Isdp.

336853 AQARD-U-800

T A \ $D
ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT
7 RUE ANCELLE, 92200 NEUILLY-SUR-SEINE, FRANCE

AQARD REPORT 800

The Design, Qualification and Maintenance

of Vibration-Free Landing Gear .I.
(la Conception, la qualification et la maintenance Iy ' 2"
.,
* des trains d'atterrissage sans vibration)

-
Papers presented at the 81s Meeting of the AGARD Structum d Materials Panet
held in BanB Canada Octobw lH5.

..........signed.
by DIMS

NOT kQh DESTRUCTION

NORTH ATLANTIC TREATY ORGANIZATION

Published March 1996

Distribution and Availability on Back Cover

 AGARD-R-800

ADVISORYGROUPFORAEROSPACE RESEARCH &DEVELOPMENT

7 RUE ANCELLE, 92200 NEUILLY-SUR-SEINE, FRANCE

AGARD REPORT 800

The Design, Qualification and Maintenance

of Vibration-Free Landing Gear
c
(la Conception, la qualification et la maintenance
des trains datterrissage sans vibration)

Papers presented at the 81st Meeting of the AGARD Structures and Materials Panel,
held in Banff, Canada 4-5 October 1995.

- North Atlantic Treaty Organization

Organisation du Traite de IAtlantique Nord
I I
 The Mission of AGARD

According to its Charter, the mission of AGARD is to bring together the leading personalities of the NATO nations in the
fields of science and technology relating to aerospace for the following purposes:

- Recommending effective ways for the member nations to use their research and development capabilities for the
common benefit of the NATO community;

- Providing scientific and technical advice and assistance to the Military Committee in the field of aerospace research
and development (with particular regard to its military application);

- Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence posture;

- Improving the co-operation among member nations in aerospace research and development;

- Exchange of scientific and technical information;

- Providing assistance to member nations for the purpose of increasing their scientific and technical potential;

- Rendering scientific and technical assistance, as requested, to other NATO bodies and to member nations in
connection with research and development problems in the aerospace field.

The highest authority within AGARD is the National Delegates Board consisting of officially appointed senior
representatives from each member nation. The mission of AGARD is carried out through the Panels which are composed of
experts appointed by the National Delegates, the Consultant and Exchange Programme and the Aerospace Applications
Studies Programme. The results of AGARD work are reported to the member nations and the NATO Authorities through the
AGARD series of publications of which this is one.

Participation in AGARD activities is by invitation only and is normally limited to citizens of the NATO nations.

The content of this publication has been reproduced

directly from material supplied by AGARD or the authors.

Published March 1996

Copyright 0 AGARD 1996

All Rights Reserved

ISBN 92-836-1032-6

Printed by Canada Communication Group

45 Sacrk-Caur Blvd., Hull (Qukbec), Canada K I A OS7

ii
 The Design, Qualification and Maintenance of
Vibration-Free Landing Gear
(AGARD R-800)

Executive Summary
Aircraft landing gears are crucial for safety, comfort (both for passengers and pilots) and for weight
considerations. As the element responsible for safely moving the aircraft on the ground, the landing
gear has to fulfill several, sometimes conflicting, requirements.
I
Landing gears that shimmy (shimmy can be defined as a self-excited instability during take-off, landing
or taxiing, involving up to three vibration motions: angular wheel motions about a vertical axis - yaw -,
and a fore and aft axis - roll -, and lateral displacement of the wheel) are unacceptable. In fact, a severe
occurrence of shimmy can damage the landing gear and its attaching structure, resulting in significant
I repair costs and airplane down time. Some assurance is therefore needed that landing gear designs will
be free from shimmy under all operating conditions including the normal wear and tear experienced in
service.
- One of the difficulties of shimmy analysis is that real landing gear systems exhibit many non-linear
r
characteristics. Tests on life-size aircraft are obviously expensive and risky, and tests on test-rigs
(namely drop-test facilities) allow only for limited information about the landing gear's dynamics; the
"
interaction between aircraft and landing gear is especially difficult to assess. On the other hand,
'- simulation offers a means to examine the behaviour of the landing gear as part of a complex system at a
reasonable cost. Both rigid and elastic body motions can be modelled.
In view of the problems involved in correcting landing gear vibrations emerging late in the design
process or even after delivery of the aircraft to service, the Workshop participants concluded that the
state of the art of analyzing landing gear vibrations is not quite up to other similar subjects and that the
cooperation of experts across aircraft, landing gear and tyre industry institutions is badly needed.
Workshop participants did not expect rapid progress unless there was a well planned and coordinated
approach to the problem. They unanimously identified AGARD as the only institution which had
enough power and authority to promote such a coordinated effort in the interest of NATO member
nations.

iii
 Letude, lhomologation et la maintenance des
trains datterrissage B amortissement
(AGARD R-800) I1 1

Synthkse
Le train datterrissage dun aCronef est un ClCment determinant pour la sCcuritC et le confort des
passagers et des Cquipages, en plus des considCrations de coefficient de chargement. Cest le train
datterrissage qui autorise le dkplacement de lavion au sol dans des conditions de sCcuritC acceptables.
I1 doit, par condquent, satisfaire B plusieurs critkres, qui sont parfois contradictoires.
I
I

seraient, bien entendu, trop cofiteux et trop dangereux, tandis que les essais effectuCs sur des
installations fixes (et notamment les Cpreuves de chute) ne donnent que des informations limities sur
les caractCristiques dynamiques du train; linteraction entre 1aCronef et le train datterrissage est
particulikrement dklicate B Cvaluer. En revanche, la simulation permet dCtudier le comportement du
train datterrissage en tant quC1Cment constitutif dun systkme complexe, et ce, pour un coot abordable.
La simulation permet la modClisation des mouvements rigides et Clastiques du fuselage.

iv
 Contents

Page

Executive Summary iii

Synthbe iv

Preface vi

Structures a n d Materials Panels vii

Technical Evaluation Report T

by A. Krauss

Reference

SESSION I: LANDING GEAR DYNAMICS

Preliminary Design Optimization of Carrier a n d Land Based Fighter Landing Gears 1

by B.M. Crenshaw and S.C. Brown

A Review of Aircraft Landing G e a r Dynamics 2

by W.E. Krabacher

Self-Induced Oscillations of Landing Gear as a n Integral Landing Gear Aircraft System 3

Problem
by W. Luber, G. Kempf and A. Krauss

Analysis a n d Control of the Flexible Dynamics of Landing Gear in the Presence of Antiskid 4
Control Systems
by E. Denti and D. Fanteria

Fuselage Vibration Control Using Semi-Active Front Gear 5

by H. Wentscher, W. Kortiim and W.R. Kriiger

Dynamic Behaviour of Motorbikes 6

by F. Bohm and H.P. Willumeit

SESSION 11: LANDING GEAR SHIMMY

Landing G e a r Shimmy - De Havillands Experience

by J. Glaser and G. Hrycko

Unsteady Tire Dynamics and the Application Thereof to Shimmy and Landing Load
Computations
by K. Koenig

Influence of Nonlinearity on the Shimmy Behaviour of Landing Gear 9

by P. Woerner and 0. Noel

A Nonlinear Model for Landing Gear Shimmy with Applications to the McDonnell Douglas 10
FIA-18A
by J. Baumann

V
 Preface

Fully reliable procedures for designing vibration-free landing gear still do not exist. This is in large part due to the absence of
accurate dynamic models for describing the tyres in ground contact, the complexity of the (generally non-linear) dynamic
behaviour of the structural systems involved as well as the dynamic interactions with steering and braking systems which are I
contributing stability factors.

The Workshop focused on the various vibrational and stability problems (e.g. shimmy, antiskid induced vibrations) that must
be considered in the early design phase of landing gear systems, thereby especially addressing problems which are related to
vibrations of the combined structural system formed by the landing gear, its tyres and the flexible aircraft structure. The
intention was to indicate the impact of (combined) landing geadaircraft vibration problems on aircraft design and operations.
A further aim of the Workshop was to bring together specialists from aircraft, landing gear and tyre manufacturers to discuss
the state-of-the-art technology in this area and to define possible future steps of development.

Dr. R. Freymann
Workshop Chairman

vi
 Structures and Materials Panel

Chairman: Prof. 0. Sensburg Deputy Chairman: Prof. S. Paipetis

Chief Engineer Prof. of Applied Mechanics
Daimler Benz Aerospace School of Engineering
Militaefflugzeuge LM2 Dept. of Mechanical Engineering
Postfach 80 11 60 University of Patras
81663 Munich 261 10 Patras
Germany Greece

SUB-COMMITTEE MEMBERS

Chairman: Dr. R. Freymann

Ministhe de la Force Publique, LU
Bahnhofstrasse 27
85386 Eching
Germany

Members: D. Chaumette - FR H. Pemer - FR

L. Chesta - IT C. Perron - CA
M. Curbillon - FR 0. Sensburg - GE
H.H. Ottens - NE

I - -
PANEL EXECUTIVE

Dr. J.M. CARBALLAL, SP

Mail from Europe: Mail from US and Canada:

AGARD-OTAN AGARD-NATO/SMP
7, rue Ancelle PSC 116
92200 Neuilly-sur-Seine APO AE 09777
France

Tel: 33 (1) 4738 5790 & 5792

Telefax: 33 (1) 4738 5799
Telex: 610175F

vii
 Technical Evaluation Report
A. Kraw

Daimler-Benz Aerospace AG
Military Aircraft LME24
8 1663 Munich
GermiWJJ

INTRODUCTION especially not if this aspect would have led to

aircraft interface or even rcdcs~gnproblems
Landing gear is an invaluable aircraft system, albeit
quite unpopular with most aircraft designers: Therefore, the idea brought forward by AGARD
In extended position. it spoils the aerodynamic S M P to perform a workshop on "THE DESIGN,
shape of the aircraft. Retracted, it uses internal QUALIFICATION AND MAINTENANCE OF
space which "could much better have hecn devoted VIBRATION-FREE LANDING GEAR" w a s
to fuel or other useful things". Moreover, its dead highly appreciated and widely supported. Present
weight impairs flight performance. Looking at author then volunteered for 'lk.corde.r", hoping to
landmg gear from a struchue point of view. it pro- be able to draw up a fairly consistent pichue of the
duces large concentrated loads and provides for a state of the art as well as of a viable path to be
lot of difficulties by requiring voluminous landing followed from early design to trouble-free operation
gear bays and doors interrupting the smooth flow of of vibration-tke landing gear. The author's initial
loads and stress. There is also the possibility that intention plas to arrange this technical evaluation
the optimal position of the landing gear with regard report in parallel to the workshop title. However,
to e.g. nosewheel liftoff differs from that required having browsed tluougb the papers several times, it
for a satisfactory behaviour as a ground vehicle, and apparrdmo~~appropriatetoconcen~oncnrcial
, both positions might be unfavourable with regard to bchnical subjects along all tbe development process
structural attachment. Another stanza to this and to provide a view on more general subjects
lamentation could be devoted to the themes of a l o n g with a summary of the Round T a b l e
wheel size, brake accommodation, tyre size, tyre Discussion.
mechanical characteristics, and tyre pressure.

Present author has spent almost thirty years of his TYPE AND CAUSE OF VIBRATION
professional life in nircraft and landing gear design
and analysis. It appeared to him that to arrive at a In s u m m a r y , m o s t a t t e n t i o n w a s p a i d to
design satisfying straightforward and clearly written conventional shimmy, which implies a combined
physical requirements was hard enough. It was even lateral I torsional I tilting motion of the wheel (3
harder to defend such design from the particular degrrrs of freedom). In particular, [21, [81. nnd t 101
interests of other design disciplines. However, at concentrate on this type of landing gcar vibration.
least in military landing gear performance and Cases of forelaft oscillations (1 d.0.f.) were
design requirements the situation with regard to presented at 131 and 141. Though not a cax of self-
"vibration-free" landing gear was even more inducal oscillatiom, there was a study on the use of
difficult. There was not only a lack of guidance with semi-active landing ge& technology (control of
regard to acceptable methods of design and stroke damping force coefficient) to damp fuselage
analysis, the requirement f o r a vibration-free vertieal bending oscillations o f a passenger
landing gear itself was compromised by accepting transport aircraft [5]. A view on forced (vertical)
"shimmy" to the extent that pilots could still control potentially r e s o n a n t a i r c r a f t l l a n d i n g g e a r
the aircraft, that shimmy loads did not exceed oscillations on rwgb p u n d is given at [I 1. In one
structural limits, and that the phenomenon was ade- of the two classical shimmy cases presented at t71.
quately taken into account in fatigue analysis (see the analytical model included a stick model of the
for instance British requirements). What wonder wing which the landing gear is attached to. At [Z],
that budgets for shimmy analyses were limited at there is also evidence that a predictedly stable
best and that an analytical prediction of landing landing gear exhibited shimmy because of
gear shimmy was not considered a hard fact. Hence neglecting attachment/fuselage flexibility in the
the landing gear designer did not get much support analysis. Staning from isolated h n t wheel shimmy
in striving for a landing gear free of shimmy. of motorcycles, [6] nlso considers oscillationsof the
compl*6 vehiile.
It became evident bat the classical problem of
shimmy is becoming increasingly complex due to model which in its
coupling with other k r a f t systems. [SI indicates measurement of 36
that a fully active landing gear could cause stability
problems in vertical (stroking)d d o n ; in view of
[l]. [3], [41, f61,and 171 pnscat author expects a lot
of additional problmns if a fully active landing par
were to b e p r o v e n t o be v i b r a t i o n - f r e e . of modcl parameters is
notwithstanding ground coupling of flight control
system. The coupling of the flexible dynamics of
landing gear and antiskid feedback dynamics to
cause fore-aftvibrations ("gemalk")is studied at with 9 elementary tnnsfer
[4]. A r e l a t e d c a s e of s t r o n g braking f o r c e manufacturer dynamic tyre
oscillations (involving virtually rigid wheel motorcycles. [6] not only
suspension) is being &Mibed at [3]. At [7lthen is .o.f. but also reviews so-
indicated a case invalving feedback from the
s p n i n g acwam.

The author's conclusion is tbat latest at the time of

qualification the designer will have to substantiate
tharrhe landing par is v i W o n - f R a in all possibk
degrees of fmdom. purrheroa he will have to show nted at [4]. also including
this for dl active s y w m s with direct or indirect
coupling wiul me landing gear.

To prepare for this task be should have a t his

disposition a wide range of proven and accepted
models and he should stan to apply these models as
early as possible in the design stage. However it
appears that the development of vibration-free and linearized field. As long
landiing ear is still far frodl th.at. shimmy calculations proper

n is, that of the existing

TYRE MODELS ever been subject to a

Without any doubt, a well proven tyre model along

with a r e l i i l e set of tyre parame.ters for this modcl
is of vitd importance for expedient and successful
development of vibration-free landing g e a r .
H o w e v e r there a p p e a r s to be no c o m m o n l y
accepted. much less a validakd dynamic tyre modcl
for application 10 a full-size analysis of all kind of
landing gear vibration (vie. 6 d.0.f. wheel motion
with arbitrary deviations from steady state). As a
nafurel consequence of this situation. there is an
almost complete lack of routinely measured and
published (e.g. along with a tyre catalogue) tyre ility of a n a l y t i c a l
nrodel parameas. practically all papers

[Z] is an example where linearized tyre models (3

d.0.f.) attributed to von Schlippe and Moreland arc
applied to study shimmy stability of thm different
aim& landing gear data sets. Both Moreland and
von Schlippe tyre medels were used by [71 to

i
represent tyre dynamic pmeertics in solution of one Paper [Z]trea the relative influence of model
existing and one predicted sbimmy problem. 171 parameter vari tions on shimmy stability pmdic-
pmfers the von Schlippc medel due 10 difficlilties tions; it also p esents practical experience w.r.t.
encountered in defining the Moreland tyre time pndictionlnali disnepancics and reasons t h d .
constant. Paper [la] refers to the Moreland point- Paper [31 treat a case of apparent shimmy (which,
accordtng to the explanauon given i the paper. is reinforcement for increascd stiffness or in form of
rather a case of camouflaged gear walk), which mass balance for changing vibration modes and/or
vanishes at increased speed presumably due to frequencies.
nonlinearity of tyre circumferential force charac-
teristics. Also in [3] there is descripted a rig test There was not one serious attempt reported to
case (virmally rigid) w h m tyre and braUantiskid modify geometry of a landing gear affe*ed, for in-
system nonlinurities pmhibit linear treatment. [41 e implement a m m suitable
s t ~ c to idnation
deseribcs a similar phenomenon observed on and/or mil m.It appurs that volume end form of
aim& a& embarks on dcvaloptnsnt of a nonlinear landing gear bays are being defined to the laat
(time domain) model f o r calculation of tyre millimeter a t a much too e a r l y time i n t h e
longitudinal force. [5] discusses feasibility of development process. l e a v i n g landing gear
linenrimtion in context witb control optimization of designers at both airframe and l a n d i n g gear
a semi-active nose landing gear. Both linear and manufacturers a t a loss to re-establish a m l y
d idifferential equations are developd at [61 vibdon-free landing gear. This again empbasiiacs
for tbe wobble (shimmy) of a motorcycle front tbc necessity ofprrpsntioa and as m l y as possibly
wheel: nonlinear effects arise from inclusion of application of qualified and commonly accepted
wheel unbalance and periodic tyre radial force methods of design for vibration-free landinggw.
vpriation. In studying shimmy having occurred in
f l i g h trials, nonlinear efkts were also accounted
for by [7lin time domain analyses. Paper [8] p- ROUND TABLE DISCUSSION
sents a complete set of nonlinear equations of
unsteady tyre dynamics. Paper [9] discusses Thc Workshop was concluded with a Round Table
influence of nonlincarities on shimmy, presenting Discussion kd by Dr.R. Fnymum. The discussion
models f o r typical nonlincarities as stick-slip was characterized by deep concern about present
friction and freeplay. hxldon of aoaliaearitics in a state and f u m development of "TEEDESIGN,
simulation model of a cantilevered landing gear QUALIFICATION AND MAMTENA" OF
g.nrrey is prrscntcd at [IO]. VIBRATION-FREE LANDING GEAR'.
C o m p a r e d w i t h e f f o r t s p e n t on a e r o -
From a scan of the sub@ of d i t i e s , p v n t (servo)elaaticity it is obvious that the field of
author got the impression that there is widespread l d i n g g w r v i b r s t i o a s has bccn negkctadbadlyfor
concern about the reliability of linearized landing a long time. Tbcro was also snse discussionabout
gear models. However, increased modelling e f f m the reasona therefor. Nowcver, in view of the cost
a n d cost o f c o m p u t a t i o n appear to i m p e d e involved in correcting landing gear vibration
application of nonlinear moder8 except in cases problems emerging late in the design p r o m s or
when shimmy occurred on existing hardware, even not before aircraft delivery to service,
notwithstanding cases whcre linearization is imp- diseussion turned to future. It became clear tbat the
PrOpriaD (landing, braking). state of the art of mating leading gear vibrations
was not quite up to olber similar subjects and that
Interesting enough to present author none of the cooperation was ncGded of expem widely diatrib-
papers mentioned shimmy under nonsymmetric utEd across aircraft, landiig gear and tyre industry
b a s i c conditions, e.g. a wheel running a t a (the latter was not represented at the workshop),
geometrically or elastically induced pre-set slip imtitutions, and Sum0ritie.s. Workshop participants
mgk ('be-in") or tilt angk. did not expect the rapid progress needed. unless
there wps a well planned and c- appmtdr
to the problem. Unanimously. AGARD was
c o ~ M R Q s u a E s identified the only instituhn which bad enough
power and wtbority to promote such a c m r d b t c d
In spite of all effort spem on design of supposedly effatintheinurratofNATOmMnbcrn;ltiear.The
vibration-free landing gear, the workshop was format of a Working Group was considered most
pmented with some practical e&lmplcs of landing suitable to acbicve the pmgrcss mmssery within a
gear shimmy ([2]. [31. [71, [IO]). It was interesting reasonrbLtim9r.Fnymnrmwasaskcdtoconvey
to note that in most c m the 1andiig gem affected this proposal to tbe appropriate AOARD M y . It
cmld not really be cured. Rather in most instances was conaidemi useful to bave tentative 'Tmms of
the butden had to be put on maintaining close Rckrena"writt.n for this task. John O l w r of de
tolerances w.r.t. f w p l a y and tyre pressure, on Havilland. William E. Krabacher of Wright
costly early tyrc replacement, on frrsucnt checks of Laboratory, and Amulf Krauss of Daimler-Benz
tyre unbalance and out-of-round. and on pilots Aerospace were n o m i d for an informal ''Point
having to h e spccial openting proccdurcs. In of coruct"group.
somc instances there has also mass b.en added to
tbe l a d i n g par, either in form of landing gear
T-4

PAPERS PRESENTED AT "HE WORKSHOP

[9] 'Tnfluence arity on the Shimmy
g,Gear". P. Wocrner
"Preliminary Design Optimization of Dowty S.A., VClizy.
Carrier and Land Based Fightcr Landing
Gears", B. M. Crenshaw and Susan C.
Brown. Lockheed Martin Aeronautical for Landing Gear
Systems. 86 South Cohb Drive, Marietta l i c a t i o n s to t h e
Ge 38063 USA as F / A - l 8 A " , Jeff
ngincer, Structural
"A Review of Aircraft Landing G e a r
D y n a m i c s " , William E . K r a b a c h e r .
WLIFIVMA, Wright Laboratory, Wright
plncrsoa AFB. obi 45433 USA

"Self-Induced Oscillatim of Landing Gear

as an IntegralLMding Gear Aircraft System
Problem". W.Luber, G. Kempf, A. Krauss,
Daimler-Eenz Aerospace AG, Military
Aircraft LME24, D-81663 Munich.
Y-
"Analysis and Control of the Flexible
Dynamics of Landiig Gear in the Presence
of Antiskid Control Systems", E. Denti and
D. Fantsria. Dipmimento di lngegneria
Aerospaziale, U n i v e r r i t i di Pisa. Via
Diotisahri 2,56126 Pisa. Italy

m 'Tuselage Vibration Control Using Semi-

Active Front Gear", H. Wentscher. W.
Kortiim. W. R. Krliger, DLR Deutsche -
Forsehungsanstalt for Luft- und Raumfdut,
Insaitutfiir Robot* und Sys~mdynamik.D
82234 Wcssliig. Germany

"Dynamic Behaviour of Motorbikes", F.

BOhm and H.P. Willumeit, Department of
Traffic Engineering and Applied
Mechanics, Technical University Berlin,
Germany. Paper was presented by P.
Banaarie

"Landing Gear Shimmy DE -

HAVILLAND's Experience", John Glaser
and George Hrycko. Structural Dynamics
Group, de Havilland Inc.. 123 Garratt
Blvd., Downsview. Ontario, M3KIY5,
Canada

"Unsteady Tire Dynamics and the

Application Thereof to Shimmy and
Landing Load Computations", Klaus
K a n i g , Daimler-Benz Aerospace Airbus
GmbH, HiinefeldsuaSc 1-5, Bremen,
Y-
 PRELIMINARY DESIGN OPTIMIZATION
OF CARRIER AND LAND BASED FIGHTER
LANDING GEARS

B. M. Crenshaw
Susan C. Brown
Lockheed Martin Aeronautical Systems
86 South Cobb Drive, Marietta Ga. 30063 USA

SUMMARY LIST OF SYMBOLS

The differences in requirements of land based (LB) and g acceleration of gravity -in./sec. squared
I
carrier based (CB) aircraft landing gear are reviewed L aerodynamic lift (per strut) - Ib.
with respect to landing impact and ground surface N load factor - nondimensional
roughness. Frequently the issue of operational S strut stroke - in.
roughness requirements vs. taxi requirements arises. V sink speed - in./sec.
More often, it seems, attention is being focused on W weight per strut - Ib.
operational runway roughness requirements. The MIL- 8, tire deflection - in.
SPEC roughness amplitudes and wavelengths may not 7 strut efficiency - nondimensional
represent the operational capability of an aircraft when 7 tire efficiency - nondimensional
combined sources of loading are considered. There are
severe loads on both main and nose gears during 1. INTRODUCTION
landing rollout on a rough runway surface if braking is
used; however, levels of braking and roughness An effective general approach to minimizing vibration
combinations are not always clearly defined in effects on landing gear and preventing transmission of
procurement specifications. ground generated high frequency loads into other
aircraft structure is to utilize the lowest practical tire
The more robust landing gears sized for carrier pressures, and to keep the strut airspring constants low.
operations are examined to determine their potential To achieve these goals, thorough preliminary analysis
- operational performance for land based roughness is needed to avoid later fixes which can make the
levels when combined loads from rollout braking are gears marginal energy absorbers for ground operations.
considered. While the weight penalty associated with
carrier qualified landing gears is commonly recognized, All too frequently, landing gears and their
and weight efficiency requirements of fighter aircraft retractionlextension mechanisms become overly
may ultimately outweigh cost considerations for complicated or compromised because the implications
commonality of landing gear components, it is of structural requirements were not studied thoroughly
nevertheless worthwhile to consider methods for enough during the initial aircraft layout to lay claim to
reducing costs through multiple application designs and the space and volume necessary to accommodate all
parts usage. Although without great previous success, landing gear needs. Sometimes inevitable aircraft
multi-service application of designs has long been an growth or mission changes can require a tire size
attractive concept. If severe runway roughness change or stroke adjustment, only to have an
capability is considered as an operational requirement interfering bulkhead preclude a straightforward change.
for land based gears, the weight obstacles to common In the tightly packed wheel wells of fighter aircraft,
landing gears may diminish. any unconservative gear sizing may even prevent such
changes as slight castedcamber adjustments to improve
A hypothetical future fighter aircraft is utilized to tire wear life or trail adjustment to optimize shimmy
compare runway roughness capability of gears sized for stability. It is therefore important to exercise all
carrier landing and arrestment loads with gears sized available computing and simulation techniques early in
exclusively for land based use, to compare relative the aircraft preliminary design stage to ensure sufficient
landing gear weights, and to develop concepts of gear stroke and wheel well clearances to support the
landing gears with the potential of multi-service usage. aircraft design maturing process without undue cast
A simple forward retracting vertical post strut penalties to the landing gear or the necessity for
arrangement has been examined from a conceptual cornpromise solut ions.
standpoint for low cost production in two stroke
lengths. For a long period, landing gear load/stroke
characteristics were determined primarily by landing
impact energy absorption requirements. Taxi loads

Paper presented at the 81st Meeting of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibration-Free Landing Gear, held in Ban& Canada from 4-5 October 1995, and published in R-800.
 1-2

were defined by a 2.0g requirement. Presently,

specifications MIL-A-8862 and MIL-A-8863 include
roughness definitions in terms of (1-cosine) dips or
bumps plus requirements for steps and other obstacles.

Transport aircraft designers usually concentrate on

maximizing step bump capability and operations from W
semi-prepared surfaces. Because of the low landing
and takeoff speeds of large transports, only the shorter
(1-cosine) wavelengths (below perhaps 100 ft.) usually
produce a speed/wavelength combination near enough To calculate the gear stroke required vs. sink speed and
to resonance to produce high gear and wing loads. At load factor, the landing impact is assumed to occur at
these short wavelengths, amplitudes are low. There is l g wing lift, and the term:
little published information available indicating how
designers have addressed the longer wavelength
roughness that becomes important at the higher landing
and takeoff speeds of modem fighters. It is
particularly perplexing how gears are designed with W
enough damping to meet the MIL-A-8863 requirement
if the aircraft is assumed to operate through resonant
speed on a surface of "continuous" (l-cosine) bumps. goes to zero. If it is further assumed that a tire will be
selected to produce 40% as much deflection as the
Conversations with procurement representatives, with shock strut, and that strut and.tire efficiencies are 85%
design and structures personnel, with test pilots, and and 45% respectively, Equation 1 becomes:
with potential users have not indicated a consensus of
opinions with respect to "design roughness" as
contrasted with "operational capability." In almost all
instances, current specification wording is generally
vague in this respect, in some places using the term --
"operate to and from --'I and in other places referring to
taxi only, without braking.
V2
With this uncertainty as to operational requirements vs. S = (3)
aircraft capability in mind, i t was decided to select two 0.45
1 + 0.4(-)(2)(386)(0.85)N
sets of landing gears, one sized for carrier operation, 0.85
and one sized for conventional land based operations
for studying potential roughness capability as a V2
function of strut stroke and design load level. The S = (4)
ground rules for carrier based gears and land based 795.16 N
gears have been generally followed at a preliminary
design level, concentrating on vertical strut loads.
Thus other load sources such as turning, pivoting, Using Equation 4, a plot of stroke requirements for
jacking, etc. have been ignored as they can be several load factors and sink rates can be developed as
determined in connection with establishing gear shown in Figure 1.
strength, setting wall thicknesses, lug sizes, etc. No
study of launch bar or holdback mechanisms has been Figure 1 may be used as a guide in selecting strut
made. Vertical struts have been assumed for strokes; however, a carrier based gear with a sink rate
computational simplicity. Non-vertical or articulated requirement of 20-25 fps. would not be designed for
struts are expected to yield similar results for the lowest load factor shown because the stroke would
equivalent vertical axle strokes. be excessive, nor would a land based gear with a sink
rate requirement of 10 fps. be designed to the highest
2. SINK RATES AND LANDING GEAR load factor shown, because of the lack of sufficient
ENERGY ABSORPTION stroke to absorb runway roughness. In keeping with
current acceptable aircraft strut sizing, a preliminary
The first step in sizing a landing gear is to establish the design 25 inch stroke for carrier gears and a 12 inch
stroke length. For vertical struts, Currie, (Reference 1, stroke for land based gears was selected for further
p35) shows that: study.
 1-3

Once preliminary stroke values are obtained, tires must runway roughness" but does not mention braking. In
be chosen. Reference 2 suggests that tires should have addition, an alternative 2.0 x MLG static load and 3.0
a deflection capability from static to fully compressed x NLG static load criteria is allowed by MIL-A-8862.
equal to the step bump height to be traversed; however, These alternate load levels allow very little energy
this is only practical for steps and short wavelength absorption reserve to be used for braking.
bumps. For longer roughness wavelengths with their
large amplitudes, it is believed to be reasonable to MIL-A-87221 is also considered ambiguous because it
choose a tire with a total deflection capability of specifies taxi analysis without braking in the taxi load
around 40 percent of the strut stroke as used in section, but requires consideration of runway roughness
Equation 4. in the braking section. A review of the structural
design criteria for several modem fighter aircraft did
During preliminary designs, it is usually sufficient to not reveal any allowance for combined braking with
consult tire manufacturers' handbooks and select an taxi loads.
existing tire if possible. For single tire struts, the tire
must have the required maximum static load rating. A I t should not come as a surprise, with the pressure to
little more freedom exists for multiple tires in that minimize landing gear weight, that gears designed to
there may be more sizes available which together will these specifications provide little braking capability
have the necessary load rating. It is desirable to utilize over the defined runways, and "operational" roughness
as much tire deflection capability as practical, yet keep capability today remains largely an undefined quantity.
the volume occupied by the tires small enough to fit in It is recommended that procurement documents in the
a minimal size and crowded wheel well. future provide more specific definitions as to the level
of "operational" runway roughness capability required.
For this preliminary study, an F-111 main tire was the
only listed size found with both reasonable tire Any additional roughness capability above that
diameter and the required static load rating for the achieved by current land based aircraft must come with
main gear. Dual C-130 size nose tires were selected, some weight penalty. The next section of this paper
keeping in mind the heavy nose gear loading resulting attempts to quantify this penalty by comparisons with a
from the carrier loading condition of free-flight gear sized for MIL-A-8863 carrier landing criteria
engagement by the arresting cable. To maximize based on a hypothetical aircraft. This comparison is
roughness amplitude capability of the study gears, the done only at a preliminary design level and is intended
same tire sizes are used for both carrier and land based to help quantify the potential weight penalty. It
designs. remains a customer's decision to accept higher landing
gear weight to achieve increased operational capability.
3. RUNWAY ROUGHNESS
4. AIRCRAFT AND LANDING GEAR
Once strut stroke lengths and tire sizes have been CONFIGURATION
selected, runway roughness response must be
addressed. The procurement authority specifies A hypothetical aircraft has been configured for this
roughness requirements, usually in terms of existing landing gear comparative study. The aircraft is not
military specifications. In the future the requirements representative of any existing aircraft; however, an
will likely be those in MIL-A-87221. Figure 2 effort was made to achieve realistic proportions,
compares portions of these specifications for (1-cosine) weights, and size anticipated for future fighters. In
discrete bumps for surfaces defined as either prepared addition to landing gear characteristics, the required
or semi-prepared at wavelengths up to 400 ft. Further parameters are landing and takeoff weight, pitching
breakdowns in speeds below and above 50 knots are moment of inertia, wheel base, and main gear tread
are made in MIL-A-87221; however, only speeds width. Tread width was chosen to ensure turnover
above 50 knots are considered in this study. Takeoff requirements are mel. An artist's concept of the
or landing dictates responses for higher speeds; but at aircraft is shown in Figure 3.
50 knots and lower "taxi" conditions, speed limitations
are feasible for controlling roughness induced loads. Two single tire, forward retracting main gear struts are
Table 1 summarizes several requirements in addition to used. A dual tire nose strut is used for compatibility
amplitude and wavelength. with carrier nose gear catapults. These configurations
are illustrated in Figure 4. Both struts have simple
Of the three specifications considered (References 3 - 9 , forward retraction mechanisms, and are mounted
only MIL-A-8863 explicitly requires combined braking laterally between spherical bearings to reduce bending
simultaneously with loads from the rough surfaces; loads transmitted to the airframe. On each gear, one
however, it does not specify a level of braking to be bearing is locked. The lateral positioning of the side
used. MIL-A-8862 defines the profiles as "acceptable brace should isolate any spin-up/springback and
 1-4

antiskid vibrations from transmission into the aircraft Figure 8 shows drop test simulations for nose and main
structure through installed bearing tolerances. No gears and free flight engagement simulation loads for
details were considered for catapult attachment. the nose gear. Since aerodynamic characteristics of the
Pertinent aircraft data is shown in Table 2. hypothetical aircraft were not defined, a conservative
tail down angle of 16 degrees was used for the free
Figure 5 shows the struts in more detail. Gas separator flight engagement simulation.
pistons are used in each strut to minimize any pressure
loss potential from gas going into solution in the 6. LANDING GEAR GROUND ROUGHNESS
hydraulic oil. There is no significant pressure CAPABILITY COMPARISONS
differential across this separator to cause leakage.
Although metering pins are shown, computations were Evaluations of the roughness capability of these gears
carried out assuming a constant orifice size. Metering during landing rollout have been based on techniques
pins should offer some improvement in efficiency, with developed for evaluating existing aircraft to operate
resultant lower loads, but would not be expected to from bomb damage repaired runways as reported in
affect relative comparisons of gear capability. Reference 7. In this method, a matrix of aircraft
Rebound damping using a flap valve below the strut velocity and roughness wavelength combinations is
upper bearing is used in all gears. Rebound damping simulated and the maximum landing gear vertical loads
generally follows recommendations given in Reference are tabulated. For purposes of this comparison, the
6. runway roughness amplitude is defined by the
equations of MIL-A-8862 that express the roughness
Both main and nose struts are assumed to be of the amplitude as a function of the wavelength. Velocity
same diameter for computational convenience. For a increments of five knots were used from 50 to 150
given load/stroke curve, the strut diameter does not knots, along with wavelengths from 50 ft. to 400 ft. in
determine vertical load, it establishes strut internal 50 ft. increments. MIL-A-8862 specifies use of single
pressures. and two bumps. In this study, two bumps were
assumed. Also only dips were simulated as dips are
5. LANDING LOADS usually found in other studies to produce higher loads
than corresponding bumps of equal amplitude.
Landing impact loads for land based gears are
established for 10 fps sink rates. Sink rates for carrier From the resultant 3-dimensional table of velocity,
landing gears are based on the multivariate analysis wavelength, and maximum gear load, contours of
outlined in MIL-A-8863. For the study aircraft, this constant load level can be drawn, producing a graphical
analysis results in a sink rate of 24.5 fps as shown by description of load trends. Contours of loads equal to
Figure 6, based on an approach speed of approximately or higher than limit load define combinations of
125 knots. Besides the landing impact loading, roughness wavelength and airplane speed that must be
allowance must be made for tire encounters with avoided. For full capability over the entire
carrier deck obstructions. Thus, additional tire speed/wavelength range, no load contour can exceed
deflection capability equal to the deck obstruction the limit vertical load of the landing gear. If limit load
height needs to be provided. exceedances are found, the bump amplitude can be
reduced until the gear loads are within the limit load
The main gear drop weight is calculated for an aft c.g. envelope. To minimize the number of computer
at design landing weight. The nose gear drop weight is simulations required, the amplitudes have been reduced
calculated for the most forward c.g. plus a forward by an equal percentage over the entire wavelength
acting c.g. acceleration of 10 fps squared, to account range studied. Thus when the results are discussed in
for pitching following impact with a tail down aircraft terms of a percentage of the specification roughness
attitude. amplitude, the quoted percentage represents the point at
which the highest load occurred, and more capability
Carrier nose gears are very highly loaded by the "free- exists for other noncritical combinations of speed and
flight engagement" condition where the tail hook wavelength.
catches an arresting cable at the instant of main gear
contact while at low sink speed and a high angle of Initial simulations were made using conventional static
attack. For this design point, the nose gear must react to compressed 3:l compression ratio strut sizing. The
impact loads plus absorb a large amount of aircraft land based gears sized for the alternate 2g/3g criteria
rotational energy introduced by hook loads. could not achieve more than approximately 55% of the
sem i-prepared roughness requirement based on single
Figure 7 shows landing impact load/stroke calculations bumps and the carrier based gear could not reach 100%
for both land based and carrier based nose and main of the semi-prepared surface single bump requirements.
gears along with nose and main tire load deflections. The airspring curves were then recomputed using
increased inflation pressures and lower compression Figure 10. For quantifying the weights, it is believed
ratio values. Because of high preload pressures, these that the conventional land based design would be in the
modifications represent what is believed to be the range of 2.2% of aircraft weight. The 0.9% relative
maximum capability for conventional single chamber increment shown for the carrier based design is
gears with these stroke lengths. believed to approximate the additional landing gear
weight necessary for 25 inch stroke gears to achieve
Figure 9 shows strut capability for these gears where full takeoff/landing/RTO capability on MIL-A-8862
sufficiently high roughness levels have been used to paved runways and limited operational capability on
produce at least limit gear load over some portion of semi-prepared surfaces. Full operational capability
the speed/wavelength range studied. Unrestricted with braking over MIL-A-8862 semi-prepared surfaces
operations would require a lower roughness level with two bumps or over the MIL-A-8863 continuous
where no limit loads would occur. (l-cosine) bumps could not be achieved with these
struts as configured on the hypothetical aircraft.
Figure 9 (a) shows the land based 12-inch stroke gear
results without braking. The roughness level was 45% 8. REDUCING ROUGH FIELD LANDING
of the MIL-A-8862 semi-prepared surface. High load GEAR COSTS
levels were produced for both main and nose gears in
the wavelength range above 300 ft. Escalating costs of aircraft procurement requires that
all possible methods of cost reduction be examined.
Since only limited semi-prepared surface roughness For landing gears, this means design simplicity and
capability exists for the land base gear, braking commonality wherever possible. Cost and weight
simulations were made assuming only prepared surface optimization must be considered jointly.
roughness as shown in Figure 9 (b). A braking
coefficient of 0.4 was assumed. Main gear loads were Forging costs represent as much as 5 - 10% of total
well below limit, but nose gear loads did exceed limit cost of modem landing gears. Applying design
at the highest speeds around a wavelength of 100 ft. initiatives to maximize commonality of forgings
between land and carrier based gears, and possibly
As configured in this study, the carrier sized 25-inch between nose and main gear components can reduce
stroke gear could not achieve 2-bump capability to production costs. To do so, the general arrangements
100% of the MIL-A-8862 roughness level at all of the land based and carrier based gears should be as
speeds and wavelengths. Figure 9 (c) shows that limit similar as is practical. Using the same basic geometry
loads are reached in the 300 ft. and above wavelength and trunnion attach points can facilitate commonality
range at 75% of the semi-prepared surface roughness of components and maximize interchangeability.
level without braking. The carrier sized gear does have
substantial braking capability over roughness as Major structural components such as the shock strut
compared to the land based design. Figure 9 (d) shows cylinder and piston could perhaps be machined from
that limit loads are just reached at 50% of the MIL-A- common forgings. Forging dies cut to produce
8862 roughness level. forgings for a 25 inch stroke gear could have excess
material removed when producing 12 inch stroke gears.
Undoubtedly, additional changes could provide some Common strut diameters could be achieved by
improvement in the roughness tolerance of these gears; machining to different wall thicknesses as necessary for
however, it is not believed that the changes would higher carrier landing loads. Navy requirements now
affect the relative capabilities of the carrier and land indicate Aermet 100 material, and this steel is
based designs. One potential method of obtaining presently a significant cost factor. Aermet 100 material
additional roughness capability is the use of double removed during machining can be recovered to recoup
chamber struts. In recognition of objections to double some raw material cost.
chamber struts by maintenance and gear overhaul
personnel, this survey was limited to single chamber Minor structural components such as drag and side
gears only. braces may also be made from common forgings, again
being sized for the heavier gear design. Allowances
7. WEIGHT COMPARISONS for wider lugs should be made during the design
process. Common trunnion pins may be used with a
It is not possible to estimate accurate landing gear larger inner diameter for the land based gears and a
weights until all load components, including drag, side, smaller diameter for the carrier based gears. Retract
and miscellaneous loads are known and the gear design and steering actuators may also be made from common
has progressed at least through the detail layout task. forgings, allowing for larger piston areas for heavier
Consequently, parametric weight estimating techniques gears.
have been used to compare these gears as shown in
1-6

Many nonstructural components may be designed to be maximum stroke lengths and tire sizes that could be
interchangeable from carrier based to land based anticipated for future, perhaps multi service usage of
designs. By using equal piston diameters, common the airframes.
components such as separator pistons, gland nuts, and
strut bearings may be incorporated. Also, system type Future studies should focus on the inherent capabilities
components such as steering valves may be completely of gears designed for shipboard operations to establish
com mon . benefits for use on rough runways.

To benefit from design commonality, both carrier based Additional work is needed to rapidly quantify landing
landing gear requirements and the maximum gear weight as a function of required capability through
operational runway/taxiway roughness should be improvements in automated loads and stress analysis.
considered at the beginning of the design process. Modeling tools of this type must be available and
History has shown that landing gears initially designed utilized from the earliest design stages.
for land base use are impractical for carrier operations
and entirely new designs are required. If carrier It is recommended that cost reduction methods,
requirements are considered initially, the landing gear including simplified gear retraction geometry, be given
may be more easily adapted to multi role uses in the high priority in the initial gear design. Sizeable c a t
future and at the same time much higher roughness savings might be realized, both in development and
tolerance potential can be achieved. manufacturing with more progress toward standardized
gear arrangements, and by reduced gear complexity
9. CONCLUSIONS which enable the use of common parts and forgings.

Specifications for landing gear design ground loads are 11. REFERENCES
not always explicit in describing operational
requirements on rough runways. The lack of clarity of 1. Currie, Norman S., "Aircraft Landing Gear
operational requirements with respect to braking levels Design: Principles and Practices," AIAA
while on rough surfaces can lead to designs with Education Series, American Institute of
reduced operational capability, i. e. limited to unbraked Aeronautics and Astronautics, Inc. ,
taxi or free rollout where the possibility of roughness Washington, D. C., 1988, p.35.
resonance exists.
2. Williams, W. W., Williams, G. K., and
Navy carrier based aircraft specifications (MIL-A- Garrard, W. C. J., "Soft and Rough Field
8863) require landinghakeoff operations from lower Landing Gears," SAE Paper 650844, October, - -
amplitude rough surfaces than the corresponding MIL- 1965.
A-8862 surfaces; however, additional capability is
inherent over corresponding Air Force gears from other 3. Airplane Strength and Rigidity, Landplane
requirements associated with Navy shipboard Landing and Ground Handling Loads, MIL-A-
operations. When severe roughness combined with 8862A, MarchJ971.
other load sources is considered, weight differences in
Navy and Air Force landing gears may diminish. 4. Airplane Strength and Rigidity, Ground Loads
for Navy Acquired Airplanes, MIL-A-
Airfield roughness capability of two landing gears with 8863B(AS), 6 May 1987.
different stroke lengths using MIL-A-8862 roughness
specifications have been compared for a hypothetical 5. Aircraft Structures, General Specification For,
aircraft. The landing gears were sized by Air Force MIL-A-87221, 28 February, 1985.
and Navy Landing Impact Criteria. The landing gear
weight penalty to provide operational capability with 6. Conway, H. G., "Landing Gear Design,"
braking to Air Force gears has been estimated from Chapman and Hall Ltd., London, 1958. , p.
this comparison at approximately 0.9 percent of aircraft 184
weight.
7. Anon, "Aircraft Operation on Repaired
10. Recommendations Runways," Report of Working Group 22 of
the Structures and Materials Panel,
Landing gear design and user technical personnel need NATO/AGARD, Paris France, AGARD-R-
to improve, clarify, and develop realistic ground 731, August, 1990.
roughness performance requirements definitions to
avoid potential problems in landing gear/vehicle
integration. Wheel well sizing should be based on the
 1-7

Figure 1. Stroke Requirements as a Function

of Sink Rate and Load Factor.

0 5 10 15 20 25
-
SINK RATE FT.lSEG.

30

f
25
/
4
10
4m
5

0 I I I I I I I
0 50 100 150 200 250 300 350 400 0 50 100 150 200 250 300 350 400

BUMP WAVELENGTH FT. - BUMP WAVELENGTH - FT.

(a) Prepared Surfaces @) Semi-Prepared Surfaces

Figure 2. Military Specification Roughness f a Prepared and Semi-prepared Surfaces.

TABLE 1

ADDITIONAL. RUNWAY ROUGHNESS REQUIREMENTS

MIL-SPEC Number of Bumps Braking Plus Roughness Landing Impact on

n Dips Roughness

8862 1 or 2 None Defmed

87221 Different Amplitude

Requirements for Single and
Double Bumps
2 Definitions:
Taxi without Braking
Braking over Roughness
No

I
 High Catapult and Arresting Loads

Free Flight Engagement: -oads

I

Deck Obstruction Loads

 .
Maximum Weight 1W,ooO Lb.
~

Landing Weight 75,000 Lb. (65,000 CB)

Landing Weight pitch Inertia 0.848E+09Lb. in?

Longitudinal Distance, Nose Gear to Main Gear 330 in.

Main Gear Tread Width 180 in.

Forward C. G. Position 61.8 in. forward of Main Gear

Aft C. G. Position 26.8 In. forward of Main Gear.

C. G. Height Above Ground 100.78 in.

Tailhmk Location 190.76 in. aft of C. G., 32.34 in. below C. G.

I 15Degrees I
I Turnover Angle (Forward C. G.) I 53.819 Degrees I
I Minimum Static Nme Gear Load I 8,121 Lb. I
I Maximum Static Nose Gear Load I 18,727Lb. I
- Nose Gear Load with Braking (10 ft./sec? Deal.) 28,211 Lb.

MIL-A-8863, Soft StNt MIL-A-8862 Soft Strut

Single Chamber, Gosoil Separated Single Chamber, Gasoil Separated
ITEM
MAIN GEAR NOSE GEAR MAIN GEAR NOSE GEAR

I Stroke I z in. I z in. I 12 in. I 12 in. I

I Piston Dia.lArea I 6 in.tB.274 in? 6 in.128.274 in? 6 in.tB.274 in?

I Extended Pressure 1 844 psi 426 psi

Extended Volume 903 in.' 794 466 in?
39 x 13 22PRn 47 x 18 36PRl1
(wWt.Est.j -16501b. 1 275 Lb. I 6ooLb. I
I Orifice Coefficient I 2.65 I 1.43 I 4.6 I 2.9 I
I A, Tire Coefficient I 9918.0 I 17,710 (2 Tires) I 7406.5 I 4742. (2 Tires) I
I A2 Tire Coefficient I 1.259 I 1.022 11.237 11.269 I
I Tire Pressure I 245 psi I 165 psi I1mpsi I 60 psi I
1 Strut Preload I 23,863 Lb. I 12,045Lb. I 25,051 Lb. I 8,8SOLb. I
I Strut Bottoming Load I 109,857 Lb. I 109,736 Lb. I 92,128 Lb. I 56,729 Lb. I
Tire Static Load 54,000 Lb. 24,600 Lb.fCire 54,000 Lb. 24,600 Lb.lIire
Rating
1-10

Front View

(a) Main Gear

Front View @) Nase Gear

Figure 4. Landing Gear Arrangement.

 METERING PIN

UPPER BEARING

SEPARATOR PISTON

SPACER \
LOWER BEARING

AIR CHMdBER SPACER

Figure 5. Landing Gear Strut Concepts.

 Figure 6. Carrier Based Sink Speed as
a Function of Approach Speed.

110 115 120 125 130 135 140 145

APPROACH SPEED - KNOTS

7,
7

X
20
b.- 20

4 15
w
$
f? 10

2
f n 5

0
0 5 10 15 20 25 0 5 10 15 20 25
STRUTSTROKE - IN. STRUTSTROKE - IN.
(a) Nose Struts (b) Main Struts
12

10

0 5 10 15 20 0 5 10 15 20
TIRE FORCE - LB. x io4 TIRE FORCE -
LB.X lo4
(c) Nme Tire Curves (d) Main Tire. Curves

Figure 7. Strut and Tire Load-Deflection Characteristics.

 1-13

1.2e+5 1.6~5 I

Free-Flight
Engagement 1.4et5
1.0~5
1.2~5
8.0e+4 m
J
w 0 1.0e+5
n
2o 6.0~4
4
9 8.0et4
U
5 +
E 4.0~4
2+ 6.0et4
v) ..
____....___. v)
4.0e+4
2.0~4
2.0et4

O.OetO
0 5 10 15 20 25 0 5 10 15 20 25

STRUTSTROKE IN. STRUTSTROKE - IN.

(a) Nose Gear (b) Main Gear
Figure. 8. Drop Test and Free-Flight Engagement Simulations.

400
\I -i 400

.!2
-
300
350

250
-

-
I@(
u
hl
84 5' 1' I
&
z
350 -92K

300 -
250 -
4 -2
- v)
200 - Nose Limit Load iz 200 - Main Limit Load
v) 57, Lb. 0 92,000 Lb.
W
$ 150 -
3
$ 150-
? U
g 100-
3
g ioo-
50
I I I I I I I I I 50 . I. I. I. l. i. I. I. i. I.
50 60 70 80 90 100110120130140150 50 60 70 80 90 100110120130140150
AIRCRAFTVELOCITY - KNOTS AIRCRAFTVELOCITY - KNOTS
Nose Gear Main Gear

Figure 9 (a). Land Based Gear Capability over Two Bumps at 45 Percent MILA-8862
Semi-prepared Roughness.
1-14

kI 40
-I
350

300
k 350
400

300
4 4
$

-
Nose Limit Load Main Limit Load
250 92,000 Lb.
57,000 Lb.
U
i? 200
I
:
w
I
w:
$ 150 $ 150

8 100
9g 100

50 IIII(IIIIII 50
50 60 70 80 90 100110120130140150 50 60 70 80 90 100110120130140150

AIRCRAFT VELOCITY - KNOTS AIRCRAFTVELOCITY - KNOTS

Nose Gear Main Gear

Figure 9 @). Land Based Gear Capability over MILA-8862 PrepaRd Surface Roughness with
0.4 Braking Coefficient.

400

k 350

300
4
$ 250 Second Contour: 150K ..
a Nose LimR Load
v)
200 200
Main Limit Load
3 w 180,000 Lb.
z 150 $ 150

g 100

50 11111111111
50 60 70 80 90 100110120130140150 50 60 70 80 90 100110120130140150

AIRCRAFTVELOCITY - KNOTS AiRCRAFTVELOClPl - KNOTS

Nose Gear Main Gear

Figure 9 (c). Carrier Based Gear Capability over 75 Percent MIL-A-8862 Semi-prepared Roughness.
 1-15

400 400

-
350 350
0
i
P 300 E
4 300 -
250 Second Contour: 90K
250 - Second Contour: 1501
d

$
W
200
Nose LImR Load
;
d

w
200 -

f 150 130,000 Lb. f 150 - Main LImR Load

2
g 100
zg 100 -
180,000 Lb.

50 I I I I I I I I I
50 60 70 80 90 100110120130140150 50 60 70 80 90 100110120130140150
AIRCRAFTVELOCITY - KNOTS AIRCRAFTMLOCITY KNOTS -
Nose Gear Main Gear

Figure 9 (d). Carrier Based Gear Capability over 50 Percent MILA-8862 Semi-prepared Roughness witb
0.4 Braking Coefficient.

"a 1.0%

5w 1.5%

:
L
1.0%

0.5%

0.0%
LANDBASED CARRIu(-BASu)

Figure 10. Landing Gear Weight Comparisons.

 2- 1

A REVIEW OF AIRCRAFT LANDING GEAR DYNAMICS

WILLIAM E. KRABACHER
WL/FIVMA Building 45
WRIGHT LABORATORY
WRIGHT PATTERSON AFB
2130 Eighth Street, Suite 1
Ohio 45433-7542, USA

SUMMARY C, - Structural Lateral Damping Coefficient

(lb-sec/in)
A review of two different landing gear shimmy
mathematical models is presented. One model uses CT - Rotational Velocity Squared Damping
the Moreland tire model and the other model uses the Coefficient (Rotation about the Strut Vertical
Von Schlippe-Dietrich tire model. The results of a Axis) (in-lb-sec2/rad2) OR Linear Viscous
parametric study using these models is presented Damping Coefficient (in-lb-sechad)
indicating the sensitivity of various parameters to
numerical variation. An identification of stability CFl - Friction Torque (in-lb)
critical parameters in the models is given. Three
different aircraft landing gear shimmy data sets are CF2 - Coefficient of Friction Torque as a Function
reviewed and model stability predictions are of Side Load at the Ground (in-lbAb)
discussed. A comparison is made with actual
experimental results. One of these data sets indicates FP - Peak to Peak Torsional Freeplay of the Gear
the nonlinear variation of various input parameters as about the Strut Vertical Axis (rad)
a function of strut stroke. In the course of the
presentation some design rules and cautions are I - Mass Moment of Inertia of the Rotating Parts
- suggested. about the Axle (in-lb-sec*/rad)

LIST OF SYMBOLS I, - Equivalent Mass Moment of Inertia of the

Lumped Mass System about the Trail Arm
The definitions of the various parameters used in Axis (in-lb-sec*/rad)
the mathematical models are defined below. Except
1 as noted, the parameters are used in both the I, - Equivalent Mass Moment of Inertia of the
I Moreland and the Von Schlippe-Dietrich models. Lumped Mass System about the Strut Vertical
Axis (in-lb-sec*/rad)
C - Tire Yaw Coefficient (radllb)
(Moreland Model Only) K, - Rotational Spring Rate about the Trail Arm
Axis Measured at the Wheel/Axle Centerline
C, - Moreland Tire Time Constant (sec) (in-lb/rad)
(Moreland Model Only)
K, - Cross Coupled Spring Rate (lb)
h - Half Length of the Tire Ground Contact Patch
(in) (Von Schlippe-Dietrich Model Only) K, - Tire Lateral Spring Rate (lb/in)
L, - Tire Relaxation Length (in) K, - Lateral Spring Rate of the Structure at the
(Von Schlippe-Dietrich Model Only) Trail Arm and Strut Vertical Centerline
Intersection (Ib/in)
C, - Structural Rotation Damping Coefficient
(Rotation of Mass about the Trail Arm Axis) KT - Torsional Spring Rate of the Structure about
(in-lb-sechad) the Strut Vertical Centerline with the
Damper Locked (in-lb/rad)
C, - Tire Lateral Damping Coefficient (lb-sedin)

Paper presented ut the 8lst Meeting of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibration-Free Landing Gear, held in Ban8 Canada from 4-5 October 1995, and published in R-800.
2-2

K,, - Torsional Spring Rate of the Steering Actuator critical to gear stability, to review some aircraft
about the Strut Vertical Centerline landing gear data sets indicating in one case the non-
(in-lb/rad) linear variation of the various input parameters as a
function of strut stroke, and to review mathematical
L - Mechanical Trail Length, defined as the model stability predictions for these data sets.
Normal Distance from the Axle Centroid to
the Strut Vertical Centerline (in) 2. THE MATHEMATICAL MODELS

La - Geometric Trail = R*sin(y) Seventeen years ago the author developed a

mathematical analysis for determining the
m - Equivalent Lumped Mass of the Wheel, Tire, shimmy stability boundaries of an aircraft nose
-.
and Fork (Ib-sec2/in) landing gear. Over the intervening years, this
analysis has evolved into a user friendly, robust
m, - Equivalent Lumped Mass of the Fork computer program that has been quite useful in the
(Ib-sec2/in) prediction of aircraft landing gear shimmy problems.
In presenting these mathematical models, it should be
R - Rolling Radius of the Wheelnire (in) made clear that no claim is made to these models
being the best or most sophisticated ones that can be
V - Aircraft Forward Taxi Velocity (idsec) used. The only claim made is that these models have
been consistently successful in the prediction of
W - Landing Gear Vertical Reaction (Ib) shimmy problems on the aircraft studied.

p, - Tire Moment Coefficient (in-lb/rad) From the landing gear structural standpoint, the
two models to be presented are completely identical.
y - Attitude of the Gear, Positive Wheel Forward The only difference between the two models is that
(rad) one uses the Moreland tire model (2) and the other
uses the Von Schlippe-Dietrich tire model (3)(4).
Other symbols used in the analysis The mathematical equations of motion for the
Moreland tire model are given by
F, - Ground Reaction Normal to the Wheel Plane
Ob) 1. m+d%/dt2- (m - mJ+L+d20/dt2=
F, - Ks+y - K,+a - C,+dy/dt
T - Torque Moment about the Strut Vertical Axis
due to Torsional Spring Rate of the Structure 2. Is*dzO/dt2= T + (V/R)+I*dddt - H+\V
(in-lb) -
- FN*(L+ La) + (m m J + L * d ~ / d t z
-
W+(A - R+a)+sin(y)

1. INTRODUCTION 3. I,+d2ddt2 = - K,*a - K,+y - C,+dafdt

- (V/R)*I*dO/dt - F,+R*cos(~)
In an earlier paper (l), a rudimentary outline was - W*(A - R+a)+cos(y)
presented for the development of an international
standard for dealing with various types of landing 4. C,*d2A/dtZ= - dy/dt - dNdt + R * d d d t
gear dynamics problems such as shimmy, gear walk, + (L + La)+dO/dt - V*O - Cl+dy/dtz
judder, antiskid brake induced vibrations, dynamic + C,*R*d2ddtZ+C,+(L + La)+dzO/dtz
response to rough runways, tire-outsf-round, and - CI+V*dO/dt- V+C+FN
wheel imbalance. The purpose of the present paper is
to sharpen the focus of the previous paper by 'Due to space limitations, a discussion of how each
presenting a detailed analysis of the specific problem parameter in the mathematical models is obtained will
of landing gear shimmy. The main content of this be deferred to another paper.
paper will be to review two different landing gear
mathematical models, to present the results of a (Note: For a general layout of the coordinate system
parametric study performed on a fighter aircraft nose used refer to Figure 1 below. Both models use the
gear indicating the sensitivity of various parameters in same coordinate system.)
the model to variation', to discuss a few parameters
 2-3

5. T= - &*[ IQ - 0,I - FP]*[Q - 8, I/[ 1

0 - 0,I ]
- [ CFI + CF2*FN]*[dQ/dt]/[IdQ/dtl ]
6. G*de, /dt*ld@, /dtl =
&*[ IQ @I I - FPl*[Q Qi I/[ IQ Qi I1
- GI%
7. FN = KO*@, - y + (L + LO)*Q)
+ Co*(dyddt -
+ (L + LO)*dWdt)
8. y = (IN)*[dy/dt + W d t - R*da/dt 8. = [ I N ] * [ dyddt - R * W d t + V*Q]
- (4 + L,,)*dQ/dt + V.6 ]
It is noted for reference (See Figure 1) that a special
where the degrees of freedom of the model arc equation can bc derived which connects the Moreland
Tire Model parameter A with ths Von Schlippe-
y - Strut Lateral Deflection at the Intersection of Dietrich parameter yI. This equation is
the Shut Vertical Centerline and the
Centerline of the Trail Arm

0 - Torsional Rotation about the Strut Vertical

Centerline

8, - Torsional Rotation at the Steering Actuator

/Damper

a - Tire Roll Rotation about the Fore-Aft Axis

through the Trail Arm

y - Tire Torsional Rotation about the Vertical

Axis through the Axle Centerline

A - Tire Contact Patch Centerline Lateral

Deflection with respect to Wheel Center
plsnc Intersection at the Ground
FIGURE 1
The mathematical equations of motion for the Von
Schlippo-Dietrich tire d l am given by A*cor(B) = yo - y + (L + L.,pain(s)
1. m*d?/dt' - (m - mJ*L*d%/dt' = or, making the small angle assumptions, cos(0) = 1.0
-
FN K,*y - K,*a - C,*dy/dt and sin@) = 0, it follows that

2. I,*d%/dt' =T + (V/R)*I*Wdt - p,*y

- FN*(L + L,,) + (m - mJ*L*d?/d?
-
-W*@, y + (L + L,J*Q -R*a)*sin(y) The degrees of freedom of this model arc

3. I,*dza/dt' = - K,*a - K,*y - C,*da/dt y - Strut Lateral Deflection at the Intersection of

- (V/R)*I*dB/dt - FN*R*cos(~) the Strut Vertical Centerline and the
- W*@,- y + (L + L a * @ -R*a)*cos(y) Centerline of the Trail Arm

4. ([(3*L,, + h)*hT/[ 6*V3]}*d'y,,ldt3 = Q - Torsional Rotation about the Strut Vertical

-yo + y - (L +Lo)*@ + (LK+ h)*O Centerline
- [GR+ W l * d Y d d t
- [(2*Ln + h)/(Z*V?]*d?ddt' 6, - Torsional Rotation at the Steering
ActuatorDamper
2-4

Q - Tire Roll Rotation about the Fore-AR Axis gear in severe divergent shimmy. The two computer
through the Trail Arm analyses also output the 0. Q,, a,A or yo. and torque
T variables in similar graphical plots.
v, - Tire Torsional Rotation about the Vertical
Axis through the Axle Centerline

yo - Lateral Distance from the Fore-AR Axis of

Undeflected Forward Motion to the Center
Point of thc Oround Contact Area

The method of initial excitation in both models is

7 SHIMMY DAMPED OUT

to give 8 an initial v a l q of 5.0 degrees.

In cornparins these two mathematical models, it

will be noted that the only difference between them is
essentially which equation is used to describe the tire
motion. When this shimmy analysis was originally
n w (SEC)
developed, only the Moreland Tm Model was used
and most of the landing gears analysed with this
analysis had Moreland tire data available. About ten FIGURE 2
years afler the original analysis was developed, a
landing gear needed to be analysed for which only
Von Schlippe-Dietrich tire data was available. Due to
the prior suoceas of the originel analysis, it was LIMIT CYCLE
decided to retain the original stnrctural analysis SHIMMY
portion and only modify the tire model portion. The
result is the Von Schlippe-Dietrich version presented
in this paper.

An important result of obtaining this second

analysis is that now a direct comparison of the
accuracy of the two tire models can be obtained since
all structural considerations are exactly identical for
the two analyses. Recently, it becume known that an TIME (SEC)
empirical procedun existed which could approximate
either tire model's shimmy parameten from the
LMwnr of tiro sizs and inflation pressure (5). The
availability of i h i ~
procedure lenda itself to providing FIGURE 3
a h i s for a direct comparison of the two models'
prediction capabilities. Some preliminary
comparisons of analysis predictions have been
conducted on a fighter known to have shimmy WVERGENTSHllvlllly
problems. However, at this time it is felt that these
comparisons are too few to make any general P
statements on trcnds found in these comparisons. 2 10

To provide examplea of the type of output Po -10

> -=
I
obtained from the two computer analyses based upon
the above two sets of differential equations, reference 30
is made to Figures 2, 3, 4. and 5 below. These four
-
graphs are plots of the y Lo@ variable for the cases
of a stable landing gear. a gear in l i t cycle
oscillation, a gear in divergent shimmy, and a
FIGURE 4
 2-5

Increases Stability
- Decreases Stability
- Decreases Stability until I, = 2.25 and then
Increases Stability
- Greatly Increases Stability
- Decreases Stability
- Increases Stability
- Increases Stability
- Greatly Decreases Stability
- Greatly Decreases Stability'
- Greatly Increases Stability
- Decreases Stability until m = ,225 and then
Neutral Stability
- Increases Stability
- Slightly Decreases Stability
- Greatly Decreases Stability
FIGURE S - Greatly Increases Stability
- Slightly Increases Stability
- Slightly Increases Stability
- Decreases Stability
3. SENSITIVITY ANALYSIS OF PARAMETER
VARIATION 4. LANDING GEAR STABILW CRITICAL
PARAMETERS
About f h e n years ago, a sensitivity analysis was
conducted using some tighter nose landing gear data In one particular shimmy investigation involving
and the Moreland Tire Model version of the shimmy a trainer nose landing gear a problem was uncovered
analysis. While the reaults of this study are only that identified stability critical parametersu.
totally accurate for the aircraft analysed, it has been Specifically, the vendor supplied a complete set of
found that the general trends in this study have been
a valuable guide in analysing other nose landing gear. ' The some of the original values used in the
The data d e f d g this nose landing gear from a parametric study were accidentally lost "he
shimmy standpoint is presented in Table 1 under the values listed in Table 1 arc approximately correct.
column for Fighter A.' However, they should not ba used in the analysis in
an attempt to reconstruct the results cited here. The
D e f d g the values in this table as the nominal results of this study should be viewed simply as a
values of the gear. the parametric study was general indicator of how dynamic response may
conducted by inormumtally varying one paramctcr at change as a parameter value is altered
a tima and increasing its value until its value was 1.5
times its nominal value. The computer output from ' Since the nominal value of L for this landing gear
the variation of each parameter was reviewed for is 0.8 inches, the maximum value uscd in this
stability effects and graded according to the scale: scnsitivity study is 1.2 inches. Therefore, greatly
greatly increases stability, increases stability, slightly decreasing stability makes physical sense in this case.
increases stability, no change, slightly decreases Generally, positive trail values below 3 inches tend to
stability, decreases stability, greatly decreases be destabilizing.
stability. Using this scale the following results were
obtained. ' The term stability critical parameter is taken to
mean that this parameter will produce greatly varying
C - Increases Stability analysis prediction results for small changes in the
C, - Increases Stability parameter value.
C, - Slightly Increases Stability ' The actual experimentally measured data for this
C, - Increases Stability
C, - Increases Stability trainer is presented in Table 1.
C, - Slightly Decreases Stability
I - Decreases Stability until 1 = 1.6 and then
 2-6

landing gear shimmy analysis data which was entered 5. REVIEW OF LANDING GEAR DATA
into a data file for the shimmy analysis. When the SETS AND STABILITY PREDICTIONS
analysis was executed, the analysis indicated the gear
would shear off in the rust quarter-cycle. (Refer to Table 1 below provides the shimmy data sets for
Figure 5 above for a plot of the actual dynamic two different tighter (Fighter B and Fighter C) and
response predicted by the analysis.) Since this level one trainer aircraft that have been analysed for
of unstable dynamic response had never been shimmy behavior. These three aircraft were selected
encountered before, it was determined that something
must be wrong with the data. After the vendor
supplied four additional data sets. all of which also
predicted the gear would-shear off in the first quarter-
cycle, an investigation was conducted by comparing
the values of the data in these five data sets with the
data sets from other aircraft nose landing gears to
determine the physical realism of the data. Using
the above introduced parametric study as a guide,
focus quickly narrowed onto the value of the
parameter K, since its variation greatly alters the
analysis predictions. The result determined was that
in all five data sets supplied the value of K,, relative
to the values of K, for other landing gears was in
error by being two orders of magnitude too large.

It was decided to measure the stiffness values of

& ,Ks ,K, I and K, as well as the torsional freeplay
of the gear on an actual aircraft (6). The outcome of
this measurement process was that in all five sets of
vendor supplied data, the= four stiffnesses were
grossly at odds with what was experimentilly
measured on the actual aircraft. The experimentally
measured values of K,, K,, and K, were
approximately one order of magnitude smaller and
the K, parameter was two orders of magnitude
smaller than the values supplied in the data sets.

While this particular vendor did not have a f ~ t e

element analysis capability for determining these
parameters, it should be pointed out tha: subsequent
encounters with these parameters on other aircraft
nose landing gears has generally established that
finite element analyses tend to error on the high side
in determining these four parameters. Therefore,
since these parameters are stability critical,
considerable caution should be used in determining TABLE 1
values for them in the design stage of the gear. As
an additional caution, a band around each such
determined value should be established and
investigated in the analysis to insure no stability because of a desire to provide examples of both
boundaries are crossed within this band or in the shimmying and non-shimmying landing gear. Each
combination of bands of these parameters. of these data sets will be reviewed and their shimmy
cheracteristics discussed.

The Trainer data set is an example of a stable

 2-1

landing gear when the values listed in Table 1 arc This procedure was repeated at increments of 0.05
used in the shimmy analysis. An example of the degrees of torsional Geeplay from 0.5 to 1.2 degrees.
typical dynamic response of this landing gear is
provided in Figure 6 below for the 8 variable. (Note: As a comparison. a static analysis of the gear was
8 in this figure is in degrees.) It is noted that this also performed to evaluate the prediction variations.
gear is very stable as there is little dynamic response The static analysis was conducted by selecting the 11
to the initial 5 degree deflection of 0. parameter values at the static stroke position of the
gear and incrcmenting the taxi speed in 5 knot
increments from IS to 150 knots while holding these
Y I I 11 parameters constant. Again the shimmy speed was
taken as the lowest speed at which the analysis
indicated sustained limit cycle oscillation of the gear.
The comparison between the dynamic and static
analyses predictions are shown in chart 14. Also
included in this charr arc the actual experimentally
determined shimmy boundaries of the gear. The
lower experimental boundary shows when shimmy
onset occurred in the gear and the upper experimental
boundary indicates strong limit cycle oscillation
shimmy.

It is particularly interesting to note that the

FIGURE 6 dynamic analysis boundary follows the contour of the
shimmy onset boundary quite closely and consistently
errors slightly on the conservative side. Whereas, the
The Fighter B data sCt is one of the most static analysis boundary lies above the shimmy onset
complete data sets for a landing gear that the author curve below 0.85 degrees of freeplay and ermrs
poswsses. This set contains 11 charts that show the increasingly on the conservative side aa the torsional
nonlinear variation of 11 of the input parameters to freeplay increaser from 0.85 degrees to 1.20 degrees.
the shimmy analysis. Also included arc charts 12 and Both the dynamic and static analyses indicated the
. 13 which show the load stroke curve of the shock shimmy frequency to be 27 HZ and this was
. strut and the load on the gear as a function of c o n f i i e d in experimental measurement.
aircraft forward ground speed.
The overall implication of these fmdmgs seems to
In an earlier paper (7). an investigation was suggest that the dynamic modelling of the gear as
conducted to determine the overall effect of these described above will significantly improve the
nonlinear vuiations on the prrdiotcd stability ~ m c of y ths a d y r i r prsdiotions and that nonlinear
boundaries of the analysis in comparison to the paramcter variation ir on important consideration in
experimentally determined shimmy stability obtaining an accurate shimmy analysis of a landing
boundaries of the gear. The manner in which this gear.
investigation was conducted was to have the analysis
execute 5 knot increment runs starting with 15 knots The Fighter C data set in Table I is provided for
and incrementing up to 150 knots. At each speed the 4 inch stroke position which corresponds to the
increment, the analysis would determine the load on static stroke position of the gear. Figures 7 . 8 . 9 , IO,
the gear from chart 13. Depending on the particular 11, 12, 13, and 14 show typical dynamic response of
parameter, the analysis would either then calculate the this nose gear. Figures 7 , 8, 9, and 10 arc plots of
parameter directly or it would determine the stroke -
the y L*Q or lateral displacement variable at the
position from chart 12 and then calculate the desired point A for the taxi speeds of 80, 90, 95, and 100
pwameter. After all I1 parameters were so knots respectively. (Note: In these figures what is
calculated, the analysis would execute and the -
called the x L*Q variable is identical to what is
dynamic response of the gear would be determined. -
referred to as the y L*Q variable in this paper.)
From the analysis results. the shimmy speed was then Figures 11. 12, 13, and 14 are the plots of the 0 or
taken to be the lowest speed at which the analysis torsional rotation about the strut vertical axis variable
indicated sustained limit cycle oscillation of the gear. at these same taxi speeds. (Note: C3 in degrees.)
 I -+
Y
I
Y

1::
l-
I -
P *-
I 1-

FIGURE 7 FIGURE 11

:
.
.I
1 Y

.
I

FIGURE 8 FIGURE 12

FIGURE 9 FIGURE 13

FIGURE 10
 2-9

In comparing these graphs it is noted that as the Referring to Figure 15, it is seen that the lateral
taxi speed increases from 80 to 90 knots a stiffness values obtained for a landing gear rigidly
progressively longer period of time is required for the mounted in a fixture (in the chart, strut only
vibration to damp out. At 95 knots the gear has the (measured)) vary from one and a half to three times
onset of limit cycle shimmy. At 100 h o t s the limit the values of the lateral stiffness actually measured on
cycle shimmy is definitely established. From these a landing gear mounted on the aircraft. This
time histories it is seen that the shimmy oscillation comparison establishes the fact that backup structural
frequency is about 27.06 HZ at 95 knots. When the flexibilities of the airframe must be taken into
actual aircraft was tested, it was found that the nose consideration when the four stability critical stiffness
gear went into a sudden onset of shimmy at 92 knots values are determined.
and had a frequency of 27.5 HZ. Repeated testing of
this landing gear, yielded reasonably consistent In addition to this primary cause of the difference
repeatability at this shimmy speed and frequency. between analysis predictions, a secondary cause was
also established. The vendor's analysis assumed a
In the course of the shimmy investigation for value of 6% of critical damping as the magnitude of
Fighter C, one very important point came to the structural damping available. Whereas, based upon
light concerning the measurement of landing gear personal experience with the variablity of available
shimmy parameters. The vendor's original analysis structural damping, the author decided to assume that
consistently indicated that this landing gear was stable the maximum amount of available damping was 1%
at all taxi speeds. The analysis provided by the of critical. At the time, the main reason for this
author indicated the 95 knot shimmy event at 27.06 selection was a desire to select a damping value that
HZ. The question arose as to what was different would expose any potential sensitivities the gear may
between the two analyses. After review of the input have toward shimmying. Further support for this
data a primary cause of this difference was position was established after the fact when Denis J.
established. The stability critical parameters KT,K,, Feld published his results on an analytical
K,, and K, used in the vendor's analysis were investigation of landing gear shimmy damping (8).
obtained by rigidly mounting the landing gear in a The essential conclusion of the analytical
fixture and measuring the resulting parameters. investigation was that the amount of available
Whereas, the author's analysis took into consideration damping at any instant during a taxi run could vary
- - the softening effects of the landing gear mounted in a from 1% to 10% of critical viscous damping.
- flexible fuselage and reduced the stiffness values Therefore, at the beginning of any landing gear
supplied by the vendor. The amount of this shimmy investigation an assumption of 1% critical
- -
reduction was largely based on experience with other damping is a reasonable starting point.
landing gear.
Before departing from Figure 15, another subject
The reason for taking fuselage flexibility effects is in need of mention. In the discussion of landing
into consideration was based on some data obtained gear stability critical parameters above in section 4, it
in another landing gear shimmy investigation. was pointed out that caution should be exercised
when obtaining stability critical stiffness values
because the finite element analysis predictions for
U T L R A L S C R I N O RATE these values tend to error on the high side. The proof
S R U T ONLY IYEAWREDI
of this error is established in the above figure for the
strut only (calculated) case where it is seen that the
largest error in stiffness values occurs between the
calculated values and the strut mounted on an actual
airframe.

6. CONCLUSIONS

In conclusion, a detailed analysis of the specific

problem of landing gear shimmy was presented in
which the two mathematical models described have
been consistently successful in the prediction of
FIGURE 15 landing gear stability characteristics to date. One of
 2-10

these models uses the Moreland Tire model and the TIRE MOTION AND WHEEL SHIMMY, Robert F.
other uses the Von Schlippe-Dietrich Tire model. Smiley NACA TR 1299 (1 957)
A classification of the various types of analysis
output was given which consisted of a stable gear, a 5. EMPIRICAL ESTIMATION OF TIRE
gear with limit cycle oscillation shimmy, a gear with PARAMETERS, Raymond J. Black, paper
divergent shimmy, and a gear with severe divergent presented at SAE Committee A-5 Meeting, 28 April
shimmy. The results of a sensitivity study based on 1980, Renton, Washington
Fighter A's landing gear data was given. The
parameters K,, K,, K,, and K, were identified as 6. THE EXPERIMENTAL MEASUREMENT OF
being stability critical in the analysis of gear shimmy. THE T-46 NOSE LANDING GEAR SHIMMY
The Trainer data set described a stable landing gear. PARAMETERS, William E. Krabacher, ASDENFSL,
The Fighter B data set described a gear with limit Wright Patterson AFB, OH (1986) (In-House
cycle oscillation shimmi at a frequency of 27 HZ Publication)
which was verified experimentally. Finally, the
I
Fighter C data set analytically described a gear also 7. A DYNAMIC SHIMMY ANALYSIS, William
with limit cycle oscillation shimmy at 95 knots with a E. Krabacher, ASDENFSL, Wright Patterson AFB
frequency of 27.06 HZ. Experimentally, this gear OH (1981) (In-House Publication).
went into limit cycle oscillation at 92 knots with a
frequency of 27.5 HZ. 8. AN ANALYTICAL INVESTIGATION OF
DAMPING OF LANDING GEAR SHIMMY,
In the course of the presentation a number of Denis J. Feld, SAE Technical Paper Series 902015
important points were made. First, the Fighter B data (1 990).
set indicated the nonlinear variation of 11 input
parameters to the shimmy analysis and suggested' that
dynamic modelling of the gear will significantly
improve the accuracy of the overall analysis C T I R E YAW C O E F F I C I E N T
predictions. The Fighter C data set established the
importance of taking into consideration the backup
fuselage flexibilities when determining the stability
critical parameters KT,K,, K,, and K,. This data set
also suggested that an initial assumption of 1% of
critical viscous damping is a good manner in which
to expose the potential senstivities of a landing gear x o .O1 f ; : : : ,
O ~ ~ : : : :
to shimmy. Lastly, the tendency of finite element 2 3 4 5 6 7 8 91011121314
analyses to error on the high side when calculating N L G V E R T I C A L L O A D X10*.3 LB
the stability critical parameters was established.

BIBLIOGRAPHY
1. AIRCRAFT LANDING GEAR DYNAMICS
PRESENT AND FUTURE, William E. Krabacher, C 1 T I R E TIME C O N S T A N T
S A E Technical Paper Series 931400 (1993).
4,
2. THE STORY OF SHIMMY, William J.
Moreland, Journal of the Aeronautical Sciences,
pp. 793-808 (1954).

3. SHIMMYING OF A PNEUMATIC WHEEL, B.

Von Schlippe and R. Dietrich ATI-18920
(1941) Subsequently printed as NACA TM 1365 2 3 4 5 6 7 8 91011121314
NLG V E R T I C A L LOAD
X10**3 LB
4. CORRELATION, EVALUATION, AND
EXTENSION OF LMEARIZED THEORIES FOR
 2-1 1

CD - T I R E LATERAL C F l FRICTIONAL TORQUE

DAMPING
700 T

- . . , , , . , . . . , , .
0 1 2 3 4 5 6 7 8 9101112
0 2 4 6 8 101214 NLG VERTICAL LOAD
LOAD X 10**3 LB XlO"3 LB

K3 TORSIONAL K4 C R O S S COUPLED
STIFFNESS S P R I N G RATE

c
aa 0 2 4 6 8101214 .- .

- - 1 3 5 7 9 1 1 1 3 1 5
NLG STRUT STROKE
(INCHES FROM FULLY NLG VERTICAL LOAD
EXTENDED) XlO'"3 LB

KD TIRE LATERAL SPRING KS LATERALSPRING RATE

RATE

0 3
c
x 2
o h 3 q : ; : ,
0 2 4 6 8101214 0
0 2 4 6 8 10 12 14 16
NGVHTICAL LOAD NLG STROKE (INCHES FROM
X 10-3 LB FULLY EXTENDED)
2-12

KT T O R S I O N A L S P R I N G RR TIRE ROLLING RADIUS

RATE
T\
n

5
a
U
8-
0- 0
0 2 -4 6 8 10 12 14 16
8.5 I . . . . . . . .
NLG S T R U T S T R O K E 2 4 6 8 10 12 14 16 18
(INCHES FROM FULLY NLG VERTICAL LOAD
E X T E N DE D ) XlO"3 LES

MUT TIRE MOMENT 40

COEFFICIENT 35
9 30
25
b
x' 20
2
5 10
5
0
0 2 4 6 8 101214 0 2 4 6 8 10 12 14 16
NLG VERTICAL LOAD NLQ STROKE (INCHES FROM
X 10**3 LB FULLY E X 1E N D ED)

1-12 I

SHIMMY ONSET
TAXI SPEED - LOAD CURVE \ 7S T R O N G SHIMM'I
I

--DYNAMICANALYSI
c I \
6
i
2000 STATIC ANALYSIS
I
I
' 0
0 40 80 120 160 0 02 0.4 0.6 0.8 1 12
G R O U N D SPEED (KNOTS) TORSIONAL FREEPLAY (DEG)
 3- I

Self-Induced Oscillations of Landing Gear as an

Integral Landing Gear Aircraft System Problem

W.Luber, G.Kempf, A.Krauss

Daimler-Benz Aerospace AG
Military Aircraft LME24
8 1663 Munich
Germany

SUMMARY the cost of testing it is desirable to be able to predict the

dynamic behaviour of landing gear systems.
Present paper is treating three topics: In the past methods were used t o c a l c u l a t e v i a
mathematical criteria, Ref. 1- 1 and 1-2. the stability of
Two of the topics deal with seemingly self-induced the nose landing gear against shimmy. Since landing gear
oscillations observed during aircraft and subsystem are very non-linear systems a time domain simulation
development tests. The third topic gives a short overview code must be applied to show the behaviour of the
on procedures to treat shimmy as a "flexible landing gear landing gear itself and also the involved subsystems. An
on flexible aircraft" problem. integrated approach to the modelling of the subsystems
The topic treated in section 2 (Shimmy on Nose Landing of the landing gear and the interaction of the elastic
- Gear) presents a case observed in taxi tests of a prototype aircraft is required to accomplish the task. To get reliable
aircraft. Although specific shimmy calculations were not results, which can also be validated by tests various
performed for this case, there are indications that a components must be considered and introduced in the
- purely lateral/torsional tyre model would possible not calculations like tyre, wheel, bogie, leg, oleo, braking
have shown the instability observed. Appendix A gives system, anti skid control, steering and flexible aircraft.
. - some evidence therefor. This report presents the straight forward trial to integrate
Section 3 (Brake Torque Oscillations at Brake Initiation) the elastic aircraft into the shimmy investigation for nose
treats a case. where quasi self-induced oscillations are landing gear of a fighter typ&aircraft development phase.
caused by interaction of tyre circumferential force
characteristics with the brake control system. Usual
means to suppress shimmy (e.g. dampers, modification of 2. SHIMMY ON NOSE LANDING GEAR
landing gear geometry) are ineffective in this case.
Landing gear vibrations may cause a variety of problems A prototype training aircraft is depicted in Fig.2- 1 and
in the rest of the aircraft. Section 4 (Integral Landing the major design data are presented in the table 1 below.
Gear/Aircraft Problem) presents a summary of potential
problem ramifications. It also gives a short overview on a I Wing SDan 1 10.46 m 1
viable approach to treat landing gear shimmy in context Total Length 10.90 m
with aircraft structure. Height 3.91 m
Wing Area 15.51 m2
Gross Weight 3540 kg
1. INTRODUCTION

An aircraft landing gear system has to be designed to

meet various requirements covering all ground-based
operations. It is a complex non-linear system which
incorporates many sub-components often procured of
sources which can produce structural dynamics and
loading effects on the gear as well on the aircraft.
The landing gear must be capable to carry heavy weight
for take off operations, hard impact of the landing as well Table 1 Technical Data Advanced Training Aircraft
as the kinetic energy of forward motion on landing to
bring the aircraft safely to rest. Due to the long terms of The aim of the development was to take part on the
developing an aircraft reducing the risk of redesign and competition of the next generation of the Air Force and
Navy training aircraft. The design of the aircraft was

Paper presented at the 81st Meeting of the AGARD SMP Panel on "The Design, Qualification and Maintenance
of Vibration-Free Landing Gear", held in Ban8 Canada from 4-5 October 1995, and published in R-800.
3-2

therefore based on the requirements of the ' I Joint 2.3 TAXITEST

Primary Advanced Training - System ".
To keep the development of the A/C in time attention Several taxi tests were performed on the smooth runway
was paid to a new aerodynamic design with good at Dasa flight test centre Manching to show that the nose
handling and performance quality. Reducing the risk of landing gear is free of shimmy. During the test no special
redesign proven components of similar aircraft were used equipment for shimmy excitation was installed on the
for fast prototyping. Therefore the nose landing gear runway. The aircraft was equipped with 13 transducers
from T-46A training aircraft was used. The modified on wing and tail to measure the important dynamic eigen
nose landing gear is shown in Fig. 2-2 with the most modes. Two additional transducers were installed on the
specific data. This paragraph describes the investigation nose landing gear to measure the x- and y- acceleration (
to prove the modified nose landing gear on the advanced Fig. 2-2 ). The result of a low taxi test is shown in Fig. 2-
training aircraft is free of shimmy. 3. This time history reveals a heavy oscillation in x- and
y- direction at a frequency of 25.7 Hz.
2.1 THEORETICAL ANALYSIS MODEL

2.4 EXPLANATION

Shimmy onset was measured at an aircraft speed of 35

kts during deceleration of the aircraft. The speed was
stabilised and the aircraft was moving straight forward
with no pilot input on the stick.
Fig. 2-3 shows the time history of the measured nose
landing gear transducers as well as the wing tip and
ailerons. The analysed frequency is at 25.7 Hz and the
maximum recorded accelerations are f log in x-direction l
and +7g in y-direction. From these accelerations it is
suspected that a primary fore/aft oscillation induces y-
oscillations due to asymmetry of nose wheel suspension
Mode Frequency (see Fig. 2-2 and Appendix). All other pick ups show no =
[ Hz ] change in the time history before and during shimmy.
Elevator Rotation 3.65 The oscillation lasted about 2.4 seconds.
~~~ ~

Rudder Rotation 3.55 The speed of the aircraft was slightly decreased to 32 kts. -
Fin 1 st Bending 7.02 A more detailed sketch of the pure shimmy, only the x-
and y- behaviour of the nose landing gear is depicted in
- ~

Wing 1 st Bending Symmetric 8.85

Fuselage 1 st Vertical Bending 12.62 Fig. 2-4.
It should be mentioned that a rigid body mode of about 3
Wing 1 st Bending Antisymmetric 16.15 Hz is overlapped this shimmy oscillation. This aircraft

Aileron Rotation Symmetric 26.20

mode were always measured during taxi tests and is not
responsible for this landing gear oscillation. i
Fig. 2-5 shows the Power Spectral Density of the
Wing 1st Torsion 36.59
ailerons. The dominant frequencies are the fuselage
bending , aileron rotation.

Table 2 Major Eigenrnodes and Frequencies 2.5 SOLUTION OF THE PROBLEM

To avoid shimmy on the nose landing gear the following

2.2 NOSE LANDING GEAR points were considered:

The nose landing gear strut is a single axle forward a) Change of trace line
retracting gear with an oil filled, nitrogen charged shock Comparison of the original nose landing gear on T-46A
strut. The gear was originally developed for the Fairchild aircraft shows that the inclination of training aircraft
Republic T-46A Trainer A/C. It was designed for a sink landing gear leg is the same. To change the inclination of
rate of 13 fps limit and 16.2 fps ultimate cases. To the nose landing gear either the actuator rod end or the
maintain the same trail of the nose landing gear as on T - attachments of the strut must be modified.
46A aircraft the original actuator rod end has been
modified by Daimler Benz Aerospace. The nose landing b) Influence of elastic fuselage modes
gear wheel and tyre is from the Pilatus PC-9 and Due to the shimmy frequency of 25.7 Hz it is assumed
delivered via BFGoodrich, Ohio, because the original I
that the eigen-modes of the fuselage, 1st( 12.6 Hz) and
wheel and tyre was not more available.
 3-3

2nd (5 1 .d Hz) vertical bending are not involved in this fastest possible brake force rise at high speed. A fast
oscillation problem. brake force rise was considered to yield. at least in
theory. shortest possible landing ground run distance.
C) Tyre Pressure
Under the circumstances that the original tyre was not The test set-up "dynamometer"consisted of one main
available and the used tyre has a different characteristic, wheel with brake running within a heavy drum. where the
an increase of pressure could solve the problem due to rotational inertia of the drum was to represent respective
shorten the relaxation line. mass per braked wheel of the aircraft. This arrangement
not only gives a representative picture of aircraft
d) Installation of shimmy damper deceleration by the brake but also provides correct
The shimmy damper is a hydraulic unit which reduces kinetic energy to be dissipated by the brake. The brake
the tendency of the wheel to oscillate from one side to was actuated by representative hydraulic components.
side. This dampers are usually constructed in one of two Brake control was performed by one lane of the brake
general designs, piston type and vane type, both of which control computer, pilot's pedal input being replaced by a
might be modified to provide power steering as well as synthetic signal. Anti-skid functions of the brake control
shimmy damper actions. computer were implemented and active.

e) Mounting of massbalance
Tuning the eigenfrequencies of the landing gear 3.2 DESCRIPTION OF THE PHENOMENON
massbalance weights have to be mounted on an arm
forwards or rearwards on the strut. Massbalance is a At brake initiation ("pilot" quasi "jumping" onto brake
useful tool if there is enough space to mount the weights. pedal) there was not the expected crisp yet steady rise of
brake force with eventual subsequent anti-skid system
f) Landing gear torsion stiffness activity. Rather there was a sequence of rapid on/off
Possible effects of torsional backlash is a reduced switching of the brake which lasted for about 1.5 seconds
effective torsional stiffness of the strut. Generally there is at a frequency of about 6 Hz. Thereafter the system
an undue sensitivity by reduction in torsional stiffness. stabilised to a normal behaviour.

Taxi tests after increasing the tyre pressure reveals no Fig. 3-1 shows a principal sketch of those three test
shimmy. Fig. 2-6 shows this sufficient improvement measurement traces which are considered essential for
- -. because the nose landing g e a r is almost free of description and explanation of the phenomenon. From
oscillation. top to bottom, these are Wheel Speed. Brake Pressure
No further changes were investigated. (i.e. pressure on brake piston face), and Brake Torque.
- -
It should be noticed that worn and cold tyres made The very first increase of brake pressure is needed to
undercarriages less stable than it was with new tyres. overcome the piston lifting spring. To arrive at the net
This was reported by pilots during taxi tests and is likely pressure acting on the brake pad, that "spring pressure" is
due to the change i n tyre stiffness as the tread wears to be subtracted from the value measured: thus, at the
away. There is also a possibility that out of balance first three troughs in Brake Pressure trace, brake pads are
f o r c e s in the worn tyre induces high frequency practically unloaded. This becomes apparent in Brake
oscillations which nullifies the effect of friction in the Torque trace, where for all practical purposes torque
landing gear. reduces. to zero at the corresponding non-zero troughs of
Brake Pressure.

3. BRAKE TORQUE OSCILLATIONS AT The initial phase of Brake Pressure deserves a further
BRAKE INITIATION comment: While brake servo valve output pressure (not
shown) follows solenoid current with negligible delay.
The case following demonstrates that not only unsuitable this is not true for the brake pressure proper as long as
combinations of structural stiffness, damping, and brake pistons are moving to close the gaps between rotor
pneumatic tyre characteristics may lead to unexpected and stator disks of the brake package. The flow of
vibration problems on landing gears. Rather, an unlucky hydraulic fluid from servo valve to brake is restricted by
combination of brake system design with the peculiarities a quite narrow Restrictor orifice which limits loss of
of circumferential force development by a tyre can also hydraulic fluid in case of e.g. rupture of a flexible hose.
produce a serious vibration problem. The Restrictor leads to a pressure drop between servo
valve output and brake piston as long as the piston is
moving. This pressure drop is the larger the faster the
3.1 THE SITUATION pilot tries to actuate the brake. However, when fluid flow
stops, pressure drop across the Restrictor vanishes and
In a series of development rig tests for a fighter aircraft full servo valve brake pressure gets through to brake
brake system. some test, were scheduled to demonstrate
3-4

pistons within fractions of a second. Thus indirectly the between corresponding tread and ground points at exit
safety feature "Restrictor" is responsible for the from the ground contact patch. This "exit" distance de is.
extremely steep brake pressure rise following on the under stationary speed conditions, calculated to be
brake filling phase.

3.3 EXPLANATIONS

From measurement traces not shown here i t became where ground speed
apparent that oscillations observed were produced by vg
vc wheel circumferential speed
anti-skid system action. However, given the software and
the parameters for skid identification installed in the length of ground contact patch
IP
control algorithm, the anti-skid system worked correctly.
The real reason for that disagreeable and inappropriate Please note that circumferential force in this regime is
series of anti-skid actions at brake initiation is a dependent on a speed ratio rather than on actual speed.
mismatch of brake system control and tyre physics.
b) At increasing speed difference, adhesive friction
3.3.1 Physics of Tyre Braking Force capability will be exceeded in small portions of the
footprint mainly at the exit end of the ground contact
Explanations following aim for a basic understanding of patch. Within the footprint spots affected, rubber will
tyre mechanics with regard to developing circumferential start to slip relative to ground albeit at a very low slip
force. speed and for very short distance. This means that
circumferential force will still increase with increasing
When a tread element of a straight-rolling tyre enters the speed difference. However, rise rate will decrease with
ground contact patch it will stick to that ground element. friction limited areas covering increasing portions of the
On a tyre rolling exactly at the forward speed of the tyre footprint.
wheel axle, the position in space of tyre tread element
and ground element will be practically identical at entry c ) At speed difference approaching ground speed
to and exit from ground contact patch. Nevertheless practically all of the rubber in the footprint will be -
within the patch there will be slight distortion due to the skidding at nearly ground speed and for extended
tyre tread being forced from originally circular shape to distance. Since skidding friction depends on skidding - -
straight (ground) shape. Contact patch distortion speed and distance covered, circumferential force of a .
changing sign at half patch length leads to approximately non-rotating tyre (full skid) will be less at high ground
zero net circumferential force on the wheel. speed than at low ground speed. Furthermore, at fixed - -
ground speed circumferential force of a skidding tyre
Non zero net circumferential force develops when wheel will also be less than the maximum achievable at a lower
circumferential speed differs from axle ground speed. differential speed, where almost all of the footprint area
Due to speed difference the tread point and the ground may also be skidding yet at a higher coefficient of
point which coincided at entry to the contact patch depart friction.
from each other on their travel through the contact patch.
As long as the relative distance of the two points is not
too large. tread rubber will adhere to the original entry Circumferential coefficient of friction (pc) of a tyre
ground point: elastic deformation of the tread element usually is presented as a function of "Slip Ratio" (SR),
increases on its travel through the contact patch, thereby where
increasing the local circumferential load on the tyre. If
Va - v,
wheel circumferential speed is sufficiently less than SR = -
ground speed, at some point along the contact patch
shear force on the tread element will exceed adhesive
friction capacity of the element. As a consequence, the
This presentation compresses all tyre angular rates from
tread element will start to slide on the ground. Sliding
friction of rubber in quite complex manner depends.
-
rolling at ground speed (SR 0) to full skid (SR = 1 )
into an abscissa from 0 to 1. However. as shown above
amongst other parameters. on sliding distance covered
and sliding speed. In general it may gradually fall well pc with an increase of slip ratio becomes increasingly
below adhesive friction with increasing sliding distance dependent on ground speed. Hence there will be different
and speed. This admittedly coarse view of rubber tyre pc vs. SR curves for different values of ground speed.
mechanics leads to the following conjectures about Fig. 3-2 presents a qualitative picture of this speed effect
circumferential force on a braked wheel: on circumferential coefficient of friction.
Figures concerning tyre circumferential force are in
a) At low speed differences, circumferential force on the essence modifications of figures presented at Ref. 3-2.
tyre should be approximately proportional to the distance
 3-5

Ref. 3-2 treats this subject in much more detail than it is During perfectly steady braking wheel angular
done in present paper. d e c e l e r a t i o n is proportional to a i r c r a f t l i n e a r
deceleration. Hence, on first view one could assume that
3.3.2 Conclusions with Regard to Sensible Brake any wheel deceleration beyond this value could be
Control interpreted as an incipient skid. In theory this assumption
holds only if the brake is operating at the maximum
3.3.2.1 Brake Control during Braking wheel torque transmittable to ground, because any loss of
transmittable wheel torque will drive the working point
Fig. 3-3 is an somewhat exaggerated variation of Fig.3-2. to the right (unstable) side of the curve (Fig.3-3 and 3-4).
However, the ordinate has been renamed "Wheel
Torque" after, in ideas, having multiplied circumferential If the brake operates on the stable side somewhere below
friction coefficient with wheel load and ground-to-axle maximum transmittable wheel torque a reduction of
distance. transmitted wheel torque will be followed by a "useful"
increment of wheel deceleration which brings transmitted
Best possible deceleration of the aircraft by wheel wheel torque back to brake torque. Therefore, anti-skid
braking could certainly be achieved if retarding moment action should only be triggered on exceedance of this
produced by the brake (!'brake torque") was just equal to "useful" wheel deceleration increment. As stated above,
the maximum driving moment achievable by the tyre admittable trigger level is zero if brake torque equals
("wheel torque"). However, this maximum point will transmittable wheel t o r q u e ; t r i g g e r l e v e l r i s e s
move around very quickly due to wheel load fluctuation (progressively) with the ratio of transmittable wheel
on uneven ground, due to fluctuation of ground to tyre torque to brake torque.
friction on varying ground roughness. due to variation of
ground-to-axle distance. and due to other effects (e.g. Assumed that a brake control system in its anti-skid
side load on a braked wheel). branch contains a fixed wheel deceleration trigger
criterion, the control system should also provide for the
Although brake torque is measurable directly, wheel appropriate torque reserve between brake torque and
torque is not. Hence it is virtually impossible to exploit maximum transmittable wheel torque. Since maximum
100 percent of achievable wheel torque throughout the transmittable wheel torque is not measurable directly,
- braked ground run. Therefore, brake control systems in anti-skid systems of the type considered here contain
general are aiming for a working point below the algorithms which reduce brake pressure output from the
- . maximum wheel torque on the left (stable) branch of the servo valve according to frequency and intensity of
. wheel torque vs. slip ratio curve. However, if by chance
maximum wheel torque falls below brake torque or if
previous anti-skid actions. Brake pressure will be
cautiously re-increased (eventually to the level
- - brake torque is increased beyond maximum achievable corresponding to pilot's command) if no more anti-skid
wheel torque, this statically unbalanced torque difference action was triggered in a sufficiently long time interval.
will decelerate the wheel and thereby increase slip ratio.
Provided torque difference is large enough and is acting 3.3.2.2 Brake Initiation
for sufficiently long time, slip ratio will be increased to
the unstable part of the wheel torque curve. Up to the point of brake initiation, the wheel rolls freely
at negligible wheel torque just balancing rolling drag.
In order to avoid this critical situation it is necessary in Therefore at brake initiation, rise of brake torque cannot
the first place to timely recognise an incipient skid and to immediately be counteracted by wheel torque. Rather,
reduce brake torque fast enough such that slip ratio is momentary difference between wheel torque and brake
kept on the stable side. torque leads to angular deceleration of the wheel. This
effects an increase of slip ratio and wheel torque. If brake
I
However, a backup procedure is needed in case that slip t o r q u e i s l i m i t e d to a v a l u e b e l o w m a x i m u m
ratio has eventually reached the unstable side of the transmittable wheel torque, slip ratio will eventually tune
wheel torque vs. slip ratio curve. If this happened the in to stationary balance of brake torque by wheel torque.
wheel would continue to decelerate to an eventual
standstill as long as brake torque is larger than actual Wheel deceleration occurs during brake initiation as well
wheel torque, even if maximum achievable wheel torque as during a skid. Hence, if during brake initiation wheel
had meanwhile recovered beyond actual brake torque. To deceleration exceeds the threshold set for skid detection,
recover from this "deep skid" the safest way is to lift anti-skid control will unnecessarily lift the brake. The
brake completely until wheel speed has returned to the wheel will recuperate to or nearly to freely rolling
stable part of the wheel torque vs. slip ratio curve. conditions. When brake pressure is re-applied by anti-
skid control another undue skid prevention cycle may
For better understanding of the case presented i t is occur.
necessary to explain the principal method applied here to
recognise an incipient skid: Wheel angular deceleration For sake of a vibration-free brake onset it appears useful
is the most important parameter used for skid detection. to gain insight into the factors influencing wheel angular
 3-6

deceleration during brake initiation. To this end quite Fig. 3-5 shows simulation results of a braked wheel
basic considerations may be helpful: during brake initiation. The model used is quite simple. It
contains just one degree of freedom representing rotation
a ) Step increase of brake torque: If brake torque is of a wheel moving at constant ground speed. Brake
assumed to follow a step function then extremum wheel torque is assumed to form a ramp type function of time
deceleration will occur at t = 0 and will equal brake while K is assumed constant. i.e. brake torque is assumed
torque divided by wheel moment of inertia. to not exceed the linear part of wheel torque vs slip ratio
function. Eventual dynamic deviations of wheel torque vs
b) Creeping increase of brake torque: If brake torque is slip ratio from quasi-stationary behaviour were not taken
applied very slowly, this will lead to low wheel into account.
deceleration values as well. because the wheel is being
given ample time to adapt to the slowly increasing Numerical results from simulation confirmed principal
demand on slip ratio. considerations presented above. Hence a systematic
evaluation of simulations was performed and summarised
c) Influence of "brakes on" aircraft speed: In a diagram in Fig. 3-6. This figure presents kind of a design chart.
showing wheel torque versus slip ratio (see e.g. Fig. 3-3) With maximum design ''brake on" speed given it shows
the initial linear part is virtually independent of ground which brake torque rise rate is at best admissible if skid
speed. However. analysis of the equation defining slip detection threshold shall not be exceeded during brake
ratio. viz. initiation. This type of diagram can easily be set up for a
specific aircraft using its tyre and wheel characteristics
(mainly moment of inertia and wheel torque curve) in
c o m b i n a t i o n with b r a k e a n d a n t i - s k i d c o n t r o l
characteristics (primarily skid detection threshold).

indicates that wheel torque rise rate over time will reduce In developing Fig. 3-6 it was tacitly assumed that brake
with an increase of ground speed. torque shall not exceed the straight part of the wheel
torque vs slip ratio curve. However, in reality this is
Assumed wheel torque is linearly dependent on slip ratio, rather improbable, because maximum brake torque
i.e. achievable is mostly close to or even beyond maximum ~

wheel torque achievable on dry runways. If brake torque

v g - vc
WT=K-SR=K-- ceiling is below maximum wheel torque achievable, - -

R wheel angular deceleration will increase on the .

or degressive slope and fall sharply off to the value
corresponding to aircraft deceleration when brake torque - -
ceiling is reached. If brake torque ceiling is beyond
maximum wheel torque a c h i e v a b l e , then wheel
deceleration will not fall off at brake torque reaching its
then wheel torque rise rate becomes ceiling; rather it will progressively increase on the
negative slope of the wheel torque vs slip ratio curve.
Unless skid detection level had been exceeded already on
the degressively positive slope of the curve, i t will be
exceeded here, provided it is small enough. In addition,
since brake torque rise rate is not at all perfectly
From this equation i t is easily concluded that wheel controllable (e.g. due to brake temperature influence on
deceleration will increase proportionally to "brakes on" brake torque vs brake pressure characteristics), variations
g r o u n d s p e e d ( b r a k e torque r i s e rate a s s u m e d of rise rate will a l s o lead to variations of wheel
independent of speed). deceleration.

It is also concluded that wheel deceleration may vary Assumptions taken for Fig. 3-6 are therefore on the
with tyre type. e.g. if different tyre construction and/or optimistic side. It is advisable to apply more realistic
tread material changes the slope K of the wheel torque vs assumptions (especially on wheel torque vs slip ratio) in
slip ratio, K may also change with operating conditions. developing brake and anti-skid control system algorithms
e. g. dry or wet ground (Ref. 3-2) and parameters.

Furtheron. wheel deceleration increases on transition Nevertheless, the simplified approach which was taken
from the quasilinear part of the wheel torque slope into here yields sufficient insight into the oscillation problem
the degressive part while approaching maximum observed during dynamometer tests of rapid brake
transmittable wheel torque. application at high aircraft speed.
 3-7

3.4 RECOMMENDATIONS AND system efficiency throughout all braked ground

CONCLUSIONS roll. Ground roll distance lost at soft yet non-
oscillatory brake initiation is only a fraction of
In the case treated here, both brake pressure and brake that which can be gained by improvement of
torque rise at enormous rates. Brake torque rise rate brake efficiency by correspondingly lower skid
alone would have been sufficient to trigger anti-skid detection threshold.
action. In addition. presumably due to slight time delays
in the system, brake torque is allowed to by far overshoot Quickest possible brake reaction at braking onset
maximum wheel torque achievable. Although in the next a p p e a r s d e s i r a b l e w i t h r e g a r d to f l i g h t
cycle the brake torque ceiling is lowered it is still too performance. However. nose diving provoked by
high due to the excessive overshoot in the previous cycle. abrupt braking will eventually produce critical
In cycles following it is hard to decide whether sharp loads and/or reduce aircraft capability to cope
rises of brake torque or brake torque overstressing wheel with rough ground (e.g. aircraft operation on
torque capacity lead to further anti-skid actions. bomb damaged and provisionally repaired
runways)[Ref.3. I ] . If brake torque rise rate is
In any case. a substantial reduction of brake pressure and reliably controlled by brake system control then
brake torque rise rates during initial brake application this can be advantageously used in structural
would be beneficial in two ways: First, undue anti-skid design (Fig. 3-7).
action could be reliably avoided thereby. Second. Adjustable control of brake torque rise rate is considered
eventual brake torque overshoot could be drastically a sensible means not only to avoid oscillatory anti-skid
reduced with the effect that brake pressure ceiling could action at brake initiation but also to improve aircraft
adapt to wheel torque capacity within a single step and ground handling and performance.
with significantly less pressure amplitude. In summary,
reduced brake torque rise rate is suited to overcome the 4. INTEGRAL LANDING GEAWAIRCRAFT
brake torque / brake force oscillation treated here. PROBLEM

From all insight gained into the problem there is no Sustained (e.g. self-induced) landing gear vibrations
solution to be expected from any kind of bolt-on often are considered to form an exotic and rather local
- structural damping device. class of dynamic stability problems. However, the
ramifications of a vibration problem on landing gear may
well extend across the entire aircraft.
- 3.5 POINTS OF IMPROVEMENT IN THE First of all. that vibrations can present severe problems
BRAKE SYSTEM with regard to dynamic loads on landing gear and its
- - attachments.
The brake torque oscillations observed in dynamometer Even if oscillatory landing gear loads do not exceed
tests at extreme "brakes on'' speed and at extreme step loads determined by quasi-static load cases. they may
input to the system can be avoided through improvement well become a problem with regard to structural fatigue
, by design as well as through improvement by operation. (e.g. Ref. 4.1).
Furtheron, landing gear vibrations may impair pilot's
Improvement by operation means that pilots are ability to control the aircraft.
instructed to initiate brakes such that brake torque rise Excessive aircraft on landing gear vibrations deteriorate
rate is kept below critical values. Improvement by passenger comfort.
operation is viable as long as an average pilot is not Early and comprehensive consideration of landing gear
overcharged by the operation required. At best. vibration is therefore an indispensable part in the
oscillations should not occur if the pilot just avoids development of a successful aircraft.
"jumping" onto brake pedals.
"Shimmy" is not the only but one of the most frequent
Nevertheless there are aspects of the problem rendering and most dangerous sources of landing gear vibration.
improvement by design a desirable alternative. Shimmy is an oscillatory wheel mode induced by a
mechanism similar to that of a wing in flutter.
Experience shows that aircraft get heavier in course The shimmy mode is important because it can lead to a
of time. As a consequence maximum "brakes on" dynamic stability problem.
speed shows tendencies to increase, too. If safety Primary source of the phenomenon is the flexibility
margins against oscillations shall be kept constant between tyre footprint and the wheel plane. Other factors
then brake torque rise rate and/or skid detection such as wheel speed, tyre geometry and torsional
threshold must be changed. If brake torque rise constraints have a contributing effect. From flutter
rate is a hardware constant then the necessary calculations we know that a component investigation
lifting of skid detection threshold will lead to show general existence of instability. Total aircraft
reduction of skid detection quality. In turn, analysis reveal always the same flutter mode a s
reduced skid detection quality reduces anti-skid component calculation but very often at lower speed.
 3-8

Through interference of aerodynamic and interaction of Third. it turned out that the elastic fuselage modes are
coupled components new flutter modes are possible. not important , if the leg mode frequencies are well
Similar to this flutter behaviour it is believed that the separated from the aircraft mode frequencies.
eigenmode of the flexible aircraft can influence the
stability of landing gear.
The aim of the landing g e a r dynamic behaviour REFERENCES
programme is to obtain dynamic data to support the
design specification of an aircraft. To do this sufficiently Smiley R.F.
the c o d e must be capable to reconcile different Correlation, Evaluation and Extension of
subsystems (structure. flight control system, hydraulic Linearized Theories for Tyre Motion and
etc.) and improve data with test results. Wheel Shimmy
At the beginning of a project the data available for sub NACA TR 1299; 1957
systems are very limited. The analysis can be used to
obtain preliminary results by assumed or old data and Pacejka H.B.
can be updated with more detailed information during the The Wheel Shimmy Phenomenon
development. The analysis at Dasa is now organised as a A Theoretical and Experimental Investigation
series of individual modules for dynamic investigation with Particular Reference to the Non Linear
for components and subsystems. This system allow easy Problem
but time intensive upgrading of modules and trend Technical University Delft, Netherlands
I
I calculation by using generalised modes.
Schweiger J., Weiss F., Dobbs S.
Each full analytical model of the leg has about 50 Aeroelastic Design and Flight Test Evaluation
eigenmodes, depending on the configuration selected. of the RANGER 2000 Training Aircraft
The high order of the model is a direct consequence of International Forum on Aeroelasticity and
including all the features that are necessary to represent Structural Dynamic 1995 Manchester U.K.
the interactions with other sub-systems during landing
and ground roll simulation. Krauss, A. J.
Considerations on Optimality of Landing Gear
Lower order models are required to permit efficient time Arrangement and Design.
response evaluation. Reduction in modal order is Paper #8 from AGARD Conference
achieved by removing the dynamics with the higher Proceedings 484: Landing Gear Design Loads
frequencies modes. This can be achieved by dynamic
condensation of the mass and stiffness matrix. For this 3.2 Brewer, H. K.
application, with relatively small finite element models, Parameters Affecting Aircraft Tire Control
the method is preferable to that of R. Guyan, because the Forces.
dynamic condensation or also the residual flexibility AIAA Paper No. 74-966
methods preserves the mode shapes very well and
permits a simple procedure for the selection of modes. 4.1 McBrearty, J. F.
Up to the lowest five modes of the leg model. selected in A Review of Landing Gear and Ground Loads
the manner described above are retained by transforming Problems.
the equations into modal coordinate system. It is also AGARD Report 1 18
possible to determine the lowest eigenmodes on a free
Ground Vibration Test (GVT) with retracted [ 4.21 Gu yan,R.
undercarriage. With those derived modes a stiffness Reduction of Stiffness and Mass Matrices
matrix can be calculated, which corresponds to the mass AIAA Journal February 1965
matrix but is not useable for static calculations.
A damping matrix corresponding to the generalised co-
ordinates can be easily established. because each mode is
selectable. This generalised matrices are used as input to
the stability criteria of R. Smiley in connection with the
linearised tyre characteristic and the applicable ground
force.
Generally speaking it turned out that the most important
parameter in the shimmy investigation is the relaxation
length of the tyre, because changes of this length results
in large changes of the landing gear stability.
Second, the torsion (friction) damping of the piston
against the cylinder, which is derived from measurement
of the landing gear producer has also an imminent
influence to shimmy.
 3-9

APPENDIX guiderail. reality this woul nean a reduction of

landing gear leg forward rake.
Consider a wheel equipped with a pneumatic tyre. The
wheel axle shall he connected to a straight guiderail by a Improvement can also be achieved by a quasi "software"
linear springidamper element. This guiderail shall be modification. i.e. by pumping up the tyre to increase RR.
moved across the ground at constant speed; funheron the
guiderail shall be inclined against the ground.
AZ. The "Slip Ratio" Assumption
Motion of the wheel axle relative to the guiderail is
influenced by radial tyre force (also represented by a As already discussed in present paper, circumferential
linear springidamper element) and by circumferential force on an almost freely rolling tyre is approximately
tyre force. Modeling of circumferential force depends on proportional to slip ratio SR.
assumptions made.
If this assumption is introduced to the model then linear
Al. "Cog-Wheel"Assumption displacement of the wheel axle and rotation of the wheel
are two separate degrees of freedom which are coupled
The wheel is assumed to rotate at that angular velocity by the circumferential force.
which is defined by the ratio of ground speed to
deflected tyre radius. Circumferential force is then This system is not as readily analyzed as the "Cog-
defined by rotational acceleration of the wheel enforced Wheel" system. However, from the main part of this
by tyre radius variations. Since this radius depends only paper it is concluded, that one and the same axle linear
on the coordinate of the wheel axle relative to the velocity relative to the guiderail will effect different slip
guiderail, the equation of motion reflecting the "Cog- ratio rates, slip ratio rates becoming smaller with
Wheel" assumption has one degree of freedom. viz the increasing ground speed. That means that the ratio of
linear displacement of the wheel axle on the guiderail. circumferential force oscillation amplitude to axle
displacement amplitude reduces with increasing ground
With regard to potential self-induced oscillations of this speed.
system it is interesting to study the various damping
. terns in that equation of motion. Apan from any phase shift effects it is therefore expected
that an eventual destabilizing effect of tyre
. With circumferential force on axle foreiaft motion will cease
with speed.
DC Damping coefficient of the guiderail spring
DT Damping coefficient of tyre radial spring
W Wheel moment of inertia A3 Conclusion
VG Ground speed
XB Axle displacement relative to guiderail Comparation of the "Cog-Wheel" model and the "Slip
RR Distance axle to ground Ratio" model perception lead to the conclusion that
p Inclination of guiderail (positive front end up) sustained foreiaft bending oscillations may well occur
under unfavourable landing gear design parameters.
the complete damping term reads However. this type of self-induced oscillation should be
limited to the low to medium aircraft speed range.

In a realistic analysis of the phenomenon, care should be

taken that all important effects are included in the model.
For instance, effective "guiderail" inclination can be
- IW.sin p .cosp .(VG + X B .cosp) influenced by fuselage bending and / or by stroking
(RR+XB.sinp)' motion of the oleo strut.

The damping term resulting from the "Cog-Wheel"

assumption is negative (destabilizing). Since it increases
approximately proportional to ground speed, the
equation indicates that there might exist a critical ground
speed at which overall damping becomes negative.

This critical speed can he increased by a "hardware"

modification to the model by reducing inclination of the
3-10

Fig. 2-1 Advanced Training Aircraft

81 . b 3

66.60.

C G R O U N D LINE
WL 59.70
STA
144.40
-
Fig. 2 2 Nose Landing Gear
 3-11

<
E

. A
-
Fig. 2 3 Time History Nose Undercarriage

1 second
., .. ..
I
1 ....; ...............;.............. j............... ~ ;_.__..._..
_._...........
.,. I , .

-
Fig. 2 4 Shimmy
3-12

X 1

-
Fig. 2 5 Power Spectral Density

.....
...............

....... ...........
f

....

-
Fig. 2 6 Time History after Modification
 3-13 I

BRAKE PRESS&

BRAKE TORQUE

I second
U

.. Fig. 3-1 Measurement Traces from Brake Fig. 3-3 Brake Working Points in Relation to
Dynamometer Test Wheel Torque versus Slip Ratio

z I

I I I I I
0 0.1 0.h 0.6 0.8 1.0
U

Fig. 3-2 Circumferential Friction Coefficient Fig. 3-4 Wheel Torque Gradient K
Dependence on Slip Ratio and Speed versus Slip Ratio
 W.

200.

100.

0
-
0.5 1.0 1.5 2.0

TIME

Fig. 3-5 Histograms of Wheel Deceleration

Rr
W TORWE RISK RATE
Fig. 3-6 Design Chart for Brake Torque Rise Rate
 3-15

AKE RISE TlME

BRAKING DRAG (QUALITATIVE)

0.00 ' b.40

1 1
0.m
I
1.20
I I
I .CO b.00
I
2.40
i
2.@0 5.20
I
S.60

Fig. A-1
 ANALYSIS AND CONTROL OF THE FLEXIBLE DYNAMICS OF LANDING GEAR
IN THE PRESENCE OF ANTISKID CONTROL SYSTEMS

E. Denti, D. Fanteria
Dipartimento di Ingegneria Aerospaziale, Universits di Pisa
Via Diotisalvi 2. 56126 PISA. Italy
Tel. +39.50.550200 - Fax t39.50.553654

SUMMARY

This work is part of research carried out to investigate the vibration problems due to coupling between rile flrsihle
dynamics of aircraji landing gear and the dynamics of antiskid control systems. The last objective is the dei~elopnrerrt
of new antiskiddevices for the suppression of landing gear vibrations during braking.
In the paper the above mentioned problems are described together with the models and methods available for their
investigation. Preliminary results of the research are also presented. Namelv, an analytical and nwirerical model ofthe
tyres in runway contact dynamics, that is one of the most critical models involved, is developed. A Iineuri,-ed nrodel
of the system composed of the wheel and the landing gear is developed and the system poles are imestigoted.

LIST OF SYMBOLS landing gear-aircraft), is needed. Accurate analytical

models of system components are also fundamental in
FN Vertical ground force order to design new control laws. In literature models
F, Longitudinal ground force and methods are mainly developed around automotive
FN Vertical ground load distribution problems while few specific works on aircraft braking
Re Effective rolling radius are available. The difference between the automotive and
sx Slip coefficient the aeronautical point of view is mainly in the elasticity
T Brakingtorque of the link between the vehicle and the wheel; wheel-hub
V, Aircraftspeed vibration due to the elastic motion of the aircrafts
- Vc Wheelcentrespeed landing gear has in fact a great influence on the dynamic
V, Wheel circumferential speed behaviour of the wheels.
Vs Wheel slip speed This paper is composed of three parts. In the first part
- Vs,. Local sliding velocity (sec. 2, 3, 4) the above mentioned problems are
XG Long. deflection of landing gear leg described, together with a review of the literature on
- P Tyre friction coefficient available models and methods for their investigation.
0 Theoretical slip coefficient Suitable models for numerical simulation are also
R Wheel spin velocity critically discussed. The second part (sec. 5,6)is focused
on the tyres in runway contact dynamics. A dynamic
1 . INTRODUCTION brush numerical model, based on a Lagrangian point of
view, is developed and investigated. The Eulerian
The growth in aircraft size and landing speed has led to version of the basic equations is analytically solved and
the development of braking control systems to protect a method to obtain the transfer function of braking force
tyres from skid damage and to reduce the stopping for small parameter variations, both at free rolling and in
distance. Early antiskids were on-off systems generic conditions, is given. Finally (sec. 7) a linearized
developed only as tyre savers to prevent lock-up skids; model of the landing gear-wheel system is developed and
later, with the introduction of the proportional servo- a preliminary investigation into the system poles is
valve, it became possible to develop improved systems presented.
with modulated braking control. Modem antiskids are
complex electronic systems that can actually improve 2 . COUPLING PROBLEMS
landing performance on dry and wet runways. This
performance may however be compromised by the On aircraft equipped with antiskid devices a dynamic
coupling of the flexible dynamics of landing gear and instability phenomenon, known as gear walk. may
antiskid feedback dynamics that may cause large wheel- occur. This is reported in detail in [MOSEL]and [ENRIG].
hub acceleration and landing gear vibrations known as The gear walk consists in strong vibration ofthe landing
gear walk. gear, typically below 20 Hz, that involves large wheel-
The subject of the present paper is the development of hub accelerations and seems due to a negative matching
mathematical and numerical models for the study of this between the frequency spectrum of the longitudinal
phenomenon, with the final aim of developing new ground force and the natural frequencies of the gear.
antiskid control laws for the reduction of vibrations and Further documentation on gear walk, collected during a
loads on landing gear during braking. braking test on a turboprop executive, is given in Fig.
T o achieve this objective a numerical code, for the 2-1 and Fig. 2-2; when the antiskid current sharply
dynamic simulation of the whole system (antiskid- decreases (and the braking torque rises) strong undamped

Paper presented at the 81st Meefrng of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibrmon-FreeLnndrng Gear. held in Ban8 Canaak from 4-5 October 1995, and published in R-800.
0-2

fundamental for analysis and synthesis of new antiskid

laws for the suppression of gear walk.

3 . SYSTEM MODELLING
The dynamic system to be studied is composed of:
aircraft, landing gear, wheels, the braking system and the
antiskid control system, connected as shown in Fig. 3-1.
Starting from the simulation code described in [DENTI],
a computer program has been developed for the
numerical simulation of the system. This program is
based on the following preliminary choices. The
dynamics of the aircraft is represented by a 3 degrees of
freedom (DOF) rigid model. The elastic dynamics of the
landing gear is represented by a 1 DOF second order
linear model. The vertical and horizontal ground forces
are computed as functions respectively o f tyre
0 IO TIME (s) 20 deformation and slip coefficient; these functions are
Fig. 2-1 An example of the gear walk phenomena supplied to the program by tables as input data. The
dynamic model of the wheel spin takes into account the
horizontal ground force, the rolling friction and the
braking moment. The braking system model takes into
account only the servo-valve dynamics, represented by a
second order linear model, while the braking moment is
assumed to be proportional to the output pressure of the
servo-valve. The antiskid control box implements a
basic control strategy and it is possible to build other
strategies of the users choice. The time history of the
pilots braking command is an input datum.
All these models are discussed in the following.

Fig. 2-2 Zoom of the critical zone

wheel-hub vibration starts. In [MOSEL] the gear walk
also occurred in tests where the braking level was below
the antiskid working threshold. The hypothesis put
forward in [MOSEL] is that the wheel-hub fore-aft elastic
motion causes high wheel angular speed variations that
can be interpreted as an impending wheel skid by the
antiskid system. Another problem can be the lag
between the real angular speed of the tread and the Fig. 3-1 Antiskid-landing gear-aircraft system
angular speed measured at the rim; this lag is due to the
Aircraft dynamics has been taken into account in order
torsional elastic mode of the tyre whose natural
to evaluate the time histories of aircraft speed and
frequency is of the order of 30 Hz. A further possible
vertical ground loads, so as to consider the modifications
cause of gear walk may be brake torque oscillations, due
to the random behaviour of brake friction [TANNE,
of the working conditions of the antiskid control
system. For this purpose the model chosen seems
MOSEL]. In any case, a possible solution might be an
adequate. Note that the aircraft pitch model gives the
antiskid with a feedback on aircraft and wheel-hub
vertical loads, on nose and main landing gear. as a
longitudinal acceleration, in addition to wheel spin
function of the longitudinal braking force, working as a
parameters. This allows evaluation of the real slip
low-pass filter.
conditions and in any case makes the synthesis of a
control on landing gear acceleration possible. The landing gear elastic dynamics model mentioned
Investigation into these phenomena is quite difficult. above seems adequate for research purposes [ENRIG].
The collection of experimental results is expensive and Anyhow, if previous modal analysis of the landing gear
critical from the point of view of correct data is possible, a n-DOF second order linear model based on
measurement and interpretation. Furthermore numerical modal superimposition can be used.
simulation needs a very accurate modelling of all system Braking system. Literature on brakes and braking
components because physical reasons for gear walk are systems is limited. Useful data about braking systems
unknown and it might be impossible to simulate gear are to be found in [GERDE, GILES, TANNE] and about
walk if it is due to neglected details of the behaviour of brakes in [ENRIG. MOSEL, PERRO, TARTE]. However.
these components. On the other hand accurate analytical these papers contain little information for the system
and numerical models of system components are also modeller, especially on the physical phenomena that
 4-3

govern the dynamic generation of the braking moments. wheel hub velocity ( V,) and spin velocity (a), that are
Friction is a non linear phenomenon depending upon both measurable quantities; R , depends upon V C .
temperature, with histeresis and random components. So vertical load on the hub ( F N ) and tyre features. In
a lot of work would seem necessary to improve the generic braking conditions, circumferential velocity ( V , )
simple model chosen. In any case, the way of modelling is defined as the product R R e , while slip velocity is
this system is not essential for the present work because given by the relationship: Vs = VC - V,. (Fig. 4-1 ).
it does not affect the dynamics of the wheel-landing gear
subsystem. However it becomes fundamental when the
antiskid control loop is eventually closed.
Antiskid control box. The antiskid system is a
dynamic box that allows simulation of both basic and
newly designed antiskid control laws.
Recent research on antiskid control systems mainly
concerns ground vehicles (cars, bikes and trucks). For
example new antiskid control laws are studied in
[LINDO, MIYAS, WATAN], where experimental and
simulation results are given, too, and in [TANCH,
YEH-I, YEH-21 from a more theoretical point of view.
In [HOHNO, M A W ] systems based on Fuzzy Logic and Fig. 4-1 Classical model
Neural Networks are presented. Practical data for car
antiskid device simulation are also available in Local sliding velocity and longitudinal global force are
linked to wheel dynamic variables such as V c , R , FN ,
[BOWMA, FANCHI.
In the aeronautical field the development of antiskid, braking torque (T) and time; so, in steady state
from the early antilock on-off systems (1950) to the conditions, we have p = p ( F N , T , V C . R ) w i t h
modem digital antiskid devices, is described in [CURRE, F , = ~ F N .Since F, , T and R are linked by the wheel
spin equation of motion, the previous relationship may
HIRZE] and NASA research on antiskid is reviewed in
[TANNE]. But the strategic importance of the matter b e written again 8s p = p ( F ~ . V c , f 2 ) o r
leads to a virtually total lack of recent research and p=p(FN,VC.Vs). In literature [CLARK, PACEZ.
detailed information and data about commercial antiskid SAKAI] global friction coefficient p is given as a
devices. So, an accurate simulation of these devices may function of FN , V c and slip coefficient s , ~ =Vs/Vc (or
of FN , Vc and theoretical slip coefficient U= V,y/V,.).
~

be difficult and may need a new synthesis based on what

- is known about their control logic. These curves, known as tyre stationary characteristics.
depend upon working conditions such as: inflation
. * Wheel. The classical wheel model adopted so far is pressure, environmental temperature, soil status. They
quite good for the evaluation of the ground behaviour of can be used to evaluate longitudinal forces in quasi
aircraft, but not for the simulation of wheel transient stationary vehicle dynamics simulation, hut their
dynamics; so, improvements are necessary for the study experimental evaluation is heavy work and need adequate
of gear walk phenomena. In the following a review of tools.
the literature on this subject is given and improved
models are developed and examined. *Recent developments. In order to synthesise tyre
stationary characteristics an empirical formulation has
4 . REVIEW OF MODELS AND METHODS been proposed in [BAKKI, BAKKZ], whose coefficients
FOR WHEEL SIMULATION have a physical meaning. Since this model, known
nowadays as the Magic Formula tyre model, gave
Ample literature exists in which several tyre models are good results and was easy to use, it has been improved
developed from different points of view. Recent upon [OOSTE, PACE11 and implemented in computer
developments in classical models and advanced models simulation facilities [SHURI]. Recently, in [SLAGM] a
are briefly reviewed in the following. Advanced models Magic Formula tyre model was suggested to represent
are divided in two classes: the first concerns models of wheel dynamics in computer simulation for landing gear
contact forces, the second load transfer from contact parameter estimation and ground load prediction.
patch to wheel hub. Global models resulting from the Contact and Force Generation Models (Brush Models)
integration of the above mentioned aspects are described are divided into two main categories: Steady State (or
too. quasi steady state), for slowly varying conditions, and
* Classical models. In classical modelling it is assumed Unsteady state, capable of simulating high frequency
that tyre structure conveys the whole load, from contact phenomena.
patch to wheel hub, without any lag [CLARK, DUGOF, - The steady s f a f e brush model is an idealised
PACEZ, SAKAI]. Longitudinal forces, arising in the representation of the tyre contact region that results from
wheel contact patch, closely depend upon the the following basic hypotheses [CLARK, PACEZ.
distribution of the local sliding velocity of the tread in SAKAI]:
relation to the ground. Steady state experiments show I) the vertical ground load distribution is an input
that, at free rolling, the absolute velocity of the tread in data evaluated experimentally or numerically by
contact is zero and the Eflective Rolling Radius ( R e ) means of other models;
may be defined [CLARK, PACE21 as the ratio between 2) rolling resistance is disregarded nor is the model
4-4

itself able to predict it; B elastic link between rigid belt & rim
3) belt and tread are considered to be flat in the C elastic link between rigid belt & rim + residual
contact zone and tread is undeformahle stiffness
orthogonally to the ground; A) An elastic ring tyre model was developed by
4) the belt velocity, in the contact zone, is assumed [SGONG, ZEGEL] to solve the problem of correctly
to be uniform and equal to slip velocity Vs; representing the high frequency vibrational phenomena
5 ) at free rolling ( Vs=O) the tread is longitudinally of the tyre belt. Particularly, the need for a model
undeformed and no shear stress exists; capable of simulating the observed Standing Waves
6) in generic braking conditions ( V @ ) , for the phenomena was felt [CLARK, HUANI, HUANZ]. The
adhering part of the contact patch, the horizontal model consists of an elastic ring, treated as a circular
shear stress is assumed to be linearly dependent elastic beam, a mass, representing the wheel rim, and
on the local longitudinal tread deformation, while radial and torsional springs, connecting the ring to the
for the sliding part, shear stress depends on the rim, whose aim is to reproduce tyre side-wall and
local sliding velocity ( V J L )on the basis of the pressurised air behaviour (foundation stiffness). The
dynamic friction law IpAcU]. ring is assumed to have the radius of the belt and ground
To discretize the problem the tread in contact may be loads are concentrated at the intersection between the
regarded as a brush (Fig. 4-2) whose bristles are ring itself and a straight line perpendicular to the ground
attached to the belt and move through the contact zone and passing by the undefomed ring centre. The model
with the relative velocity V,. At least one bristle is gives good results [SGONG, ZECEL] and makes it
always adhering at the beginning of the contact patch possible to take into account the fundamental elastic in-
and the adhering zone extension is determined by the plane dynamics of the wheel. On the other hand. if one
static fnction threshold. is not interested in high frequency modes, only
The brush model, whose fust formulation can be found translational and rotational rigid displacements of the
in [DUGOF](1970), gives good results and is widely belt with respect to the rim are really important and the
used, both in its simplest form [DUGOF. PACEZ] and model becomes too sophisticated.
with experimental or empincal improvements [SAKAI, B) In this case the ring modelling the tyre belt is
SAKAZ, SAKA3, SAKA41. linked to the rim by radial and torsional springs as in the
previous model but it is assumed to be rigid [CLARK,
SAKAI, ZANTI, ZANTZ]. The ring radius is assumed 10
be equal to the effective rolling radius [ZANTI, ZANTZ]
to fit the kinematic relationships at free rolling [PACE21
and the previous hypotheses about load introduction are
retained. The rigid ring model provides good results in
wheel dynamics simulation as demonstrated in [ZANTI]
and [ZANIZ].
C) Recently, an intermediate model has been proposed
in [ZEGEL].The first test results show a good agreement
with experimental data for medium range frequency
phenomena. The ring is rigid and it has the radius of the
undeformed belt, but, between the load introduction
Fig. 4-2 Brush model point and the ring. more elasticity has been introduced.
- Unsready Models of contact force generation are rare Such elasticity (referred to as residual stiffness)
in literature. Examples can be found in [ZANTl] and consists of vertical, longitudinal and torsional springs
[ZANTZ] where the model adopted for computer whose stiffnesses are calibrated on the basis of
experimental wheel elastic properties [ZEGEL].
simulation is an unsteady brush model, implemented
from an Eulerian point of view, which gives very . I ~ ~ ~ ~ ~ ~ ~
, ~
~ ,A, ~
c ::l ~
~ ~ , , ~ : ~

presentea. extreme, there are detailed physical models with both

Finally it should be noted that all contact models and tyre structure and tread-ground interaction represented
their results are closely connected to the modelling of [SAKAl, SAKAZ, SAKA3, SAKA41, while at the other,
rubber-soil friction phenomena. Reviews on this subject there are classical tyre and Magic formula models. In the
are to be found in [SAKAZ] and [CLARK]; furthermore in middle asimple physical model can be found, essentially
[PACE21 simplified models for analytical studies or for obtained by combining the wheel spin equation of
.. the computational load in numerical simulation
reducing
are discussed.
motion with the above cited Brush Model.
Unsteady models combining a B or C tyre structure
* Tyre Structure and Transmission Properry Models may model with the unsteady brush model give a very good
be divided into the following main categories: fit between simulation and unsteady experimental results
A elastic link between elastic belt & nm [ZANTI, ZANTZ, ZEGEL]. Recently, a global dynamic
tyre model Is .en orooosed in IEGRU]; it consists of
I . Brush s
a complex representation of the elastic and dumping Belt ; s p x s -
features of tyre S I I U C N E (Fig. 4-3) with detailed contact
modelling. It makes it possible to evaluate the vertical
ground load distribution and to take into account rolling
resistance phenomena disregarded in all physical models
\ &-XSL -
reviewed. The price to pay is in terms of computational
load and complexity in the interpretation of results. Soil

Physical model ,,c7

Fig. 5.1-1 Bristle sketch and physical model

analogous to that usually used for brush modelling
Fig. 4-3 Radial spring-damper model [ZANTI]. The equations governing bristle dynamics are
Conclusions summarised in the following table. where f l ( V . 7 ~ )is
For the numerical analysis of gear walk, a " B or " C
tyre structure model coupled with the unsteady brush Non sliding
model (Fig. 4-4) seems to give a fairly good
representation of the physical phenomena.
From the point of view of the analytical modelling of
the system, in all the works reviewed, a constant wheel
forward speed V , is assumed. Since such a hypothesis
cannot be accepted if the wheel hub is connected to an
elastic landing gear leg, improvements are then
necessary to obtain dynamic characteristics of the system
(zeroes and poles). The poles of the brush subsystem are
given in [ZEGEL] but only at free rolling; so, poles in v,yr. = 0 ; 5 = v,y
generic braking conditions must be determined in order F, = K5
to synthesise antiskid control systems.
Discontinuities in bristle dynamic behaviour have been
handled as shown in the flow chart in Fig. 5.1-2. The
adhering bristle starts sliding, with respect to the
ground, if the sum of the elastic and dumping forces
becomes greater than the static friction threshold. The
sliding bristle stops moving when a change in velocity
sign occurs and the sum of the elastic. dumping and
inertial forces (if any) is below the static friction
threshold.
Stsrt
I

Fig. 4-4 The model selected

5. A GRANGIAN BRUSH MODEL FOR

CONTACT DYNAMICS SIMULATION

5.1 Bristle models

The first step in order to build a Lagrangian brush model

is to make certain dynamic models of the bristles
available. A sketch of the bristle is shown In Fig. 5.1-1 Non sliding bristle model Sliding bristle model
together with the physical model being considered.
Three possible parameter choices, referred to as the I I
"MCK Model" (with M,C,K#O), the "CK Model" (with 44
ffeium
M=O) and the "K Model" (with M=O,C=O), have been
investigated by numerical simulation. The K Model is Fig. 5.1-2 Bristle K-model management
4-6

In Fig. 5.1-3 qualitative responses to a step in slip 5 . 2 Brush models and software structure
velocity ( V,) are shown for the three models presented
above. The K Model (Fig. 5.1-3 a) behaves in a rather The software represents a brush built with a set of
simple way even if it is quite "unphysical": as bristle bristles behaving according to one of the three
sliding happens, the sliding velocity V, suddenly previously described models. The basic hypotheses are
reaches its steady state value (V,), while the relative the same as those mentioned in sec. 4 for brush models.
displacement of the bristle shows a discontinuity due to with the assumption of a parabolic distribution of
the fact that the spring force (proportional to the relative vertical forces (see Fig. 4-2). The simulation is
displacement of the bristle) must balance ground friction performed by means of a Lagrangian approach: each
which decreases as VsL increases. The CK Model (Fig. bristle is followed as an independent dynamic system
5.1-3 b) avoids the discontinuity of the kinematic during its motion along the contact patch. The roots of
quantities, by introducing a V, dynamic transient the bristles are attached to the belt and equally spaced:
before the steady state value V, is gained. The MCK they move all together with the slip velocity ( V.7 ) rhat
Model (Fig. 5.1-3 c) shows very complex dynamics is one of the two inputs of the system, the other heing
whose effects in a dynamic brush model simulation are wheel centre velocity ( Vc). When a bristle exits from
not valuable a priori. the contact patch, a new bristle enters, at the same time.

0
1 7 :
Velocity

+ Position
xsL=xs (imposed)

Time
~
from the other side.

5.3 Results from the Brush K Model

In this section the numerical results obtained with the

brush model based on the bristle K Model are reponed.
Data for simulation were taken from various sources
[CLARK, SAKAZ, ZEGEL]constituting a set describing a
0 reasonable tyre. The values used for the tyre parameters
t Force Time
I
are listed below while the function giving rubber friction
versus sliding velocity is shown in Fig. 5.3-1:

"
I A- - Vertical force ( FN ) = 3200N/m
- Footprint half length ( a ) = 0.08 m
a) Mk.d.d Time - Bristle stiffness ( K ) = 1.0 IO6 N/m3

t Velocity 0.4
0.3
0.2
0.1
I me
4 Position 0
-0.1
-0.2
-0.3 ..........
Force Time -0.4
-0.5
-10 -5 0 5 10

Fig. 5.3-1 Rubber-soil friction law

lis,,q.. ,q b) CK-Model Time Braking rest. Fig. 5.3-2 shows transient responses on

-
Velocity total longitudinal force Fx (per tyre width unit) to step
inputs in V,, with constant V ~ = 3 0m/s. The unusual
shapes of the transient responses are understandable if
~

the time histories of longitudinal force distributions

n
along the footprint shown in Fig. 5.3-3 are analyzed.
Time
f Position Two opposite phenomena happen whose balance depends

0 I (Vs-Vsd
K(xs-xs~+C Time
on the intensity of the braking input: increases in Fv
values are due to the growth of bristle deformation in the
adhering part of the contact patch, while reductions in
F, values depend upon the increase in the extension of
the sliding zone. For small steps of sx ( bSr= 0.0303)
F, increases gradually to its steady state value because
the sliding zone is negligible. For greater steps, the Fv
time history shows an overshoot which becomes
Time
C) MCK-Model stronger as the steady state extension of the adhering part
of the contact patch decreases.
Fig. 5.1-3 Bristle responses From the steady state values of the time histories in
 4-7

occurs when the sliding zone is about fjO% of the

footprint length with sx ~ 0 . 1 5 The
. integral values of
the graphs in Fig. 5.3-5 were analytically calculated,
too, by the means of the relationships [PACEZ]:
(1)

with 2Ka2
2,r=2a(l- vu); q = -
3p o FN
solving Eq.(I), the following function Fv= Fr(a)was
obtained:
-2( ~py]}.(3)
~=F,(3rov~(l-vu)~+p[3(11~)
In Fig. 5.3-4 the results of the analytical calculation for
V, = 5, IO. 1 5 , 2 0 , 2 5 , 3 0 m/s are shown. The 30 mls
nm.1.1
curves coincide with the ones obtained by numerical
Fig. 5.3-2 Transient response in longitudinal simulation. It is interesting that, when plotting fi as a
force to a braking step input function of slip velocity V,, at high slip values p
Fig. 5.3-2, p(u),~ ( 8 , )and p(V,) curves were built depends only upon V,.
for Vc=30 mls and are shown in Fig. 5.3-4. The shape
of these curves is fully understandable on the basis of . Transient response analysis. Fig. 5.3-6 shows the FI
responses for small steps of s, starting from different
steady state longitudinal force distributions along the
footprint shown in Fig. 5.3-5. The integral value of steady state braking conditions; tests have been carried
such distributions depends upon the extent of the sliding out at constant Vc=30 mls. These tests aim at
portion of the contact patch (where long. stress has a deepening system behaviour for small perturbations
parabolic shape) with respect to the non sliding one ahout steady state conditions, both at free rolling and in
(where long. stress has a linear shape). A maximum any braking stationary condition whatever.

Fig. 5.3-3Transient variation of longitudinal load distribution over the contact patch for a braking step input
4-8

distribution along the contact patch. In order to obtain a

deeper understanding of this influence. tests for higher
sx step inputs were carried out and both the transient
responses and the time histories of the load distribution
obtained are shown in Fig. 5.3-7. The initial down step
is clearly due to the sudden variation in local friction
forces on sliding bristles p ( VsL ) , caused by the sudden
change of VsL from its initial value to the final one
(jump from condition 1 to 2 in fig. 5.3-7). The increase
and decrease in the F value (passing from condition 2
to 3. 4, etc. in fig. 5.3-7) is essentially due to the
balance between the rise in the number of sliding
0 0.2 0.4 0.6 0.8 1 bristles (due to the advance of the transition point) and
the rise in the deformation of the adhering ones.
The investigation carried out also shows that the F,
vc (dS)
response reaches its steady state values at a time which
0.25 .. 5 .......... depends on the non sliding percentage of the contact
0.2 .... 10 ......... ................... patch. Such a time is equal to the replacement time
15 necessary to replace all the bristles that are not sliding at
0.15 ....... the initial stationary state, given by:
0.1 f r = M V C - vs) . (4)
0.05 .... The variation of t , versus sx is shown in Fig. 5.3-8.
Finally, to complete the analysis of the system
0 dynamics, F responses for little steps of s , ~ ,with
0 0.5 1 1.5 2 2.5 3
constant V,, were evaluated, starting from the same
0.35 I ~~~

.. .. .. .. .. .. steady state braking conditions considered above. The

results are given in Fig. 5.3-9. The response starting
from s,=0.3333 is not given because, at such high slip
values, F, does not depend on Vc if Vs=constant (see
Fig. 5.3-4).

Conclusions
The brush model presented, based on the bristle K
Model, is not able to adequately represent sliding
velocity VsL which, at the transition point, suddenly
0 1 2 3 4 5 6 7 8 changes from zero to the value of slip velocity V,y. This
unphysical behaviour does not fit in with experiments
Fig. 5.3-4 Friction coeff. versus slip c0eff.s that show (at least for small sx) a linear shape variation
and slip vel. in steady state cond. of VsL along the sliding zone [CLARK]. However all
the above does not seem to affect too much the system
The steady state values of F variations are of course response on the total longitudinal force which fits
different according to the local shape of the p(s,) curve experimental measurements quite well [ZANTI].
at the point corresponding to the initial steady state This may be due to the fact that for small s . ~values. the
condition. The shape of the transient responses on F, sliding zone is very small, while, for larger slip
are obviously linked to the time histories of load coefficients, the sliding zone is larger but the badly
modelled one is confined to a restricted region after the
transition point. In conclusion, the model seems to be
able to represent contact dynamics in a numerical
simulation of wheel behaviour.
Besides, the bristle K Model is simple enough to permit
the development of analytical models of contact
dynamics (see section 6).

5.4 Brush CK and MCK Models

Brush models based on the bristle MCK and CK Model,

have not been investigated in detail yet. Judging from
m
0
%.
,ox 2- 1pi m Io* 4 ,m no* sm mm
preliminary results, the CK Model seems to be
interesting when a constant FN distribution is used,
while, with parabolic FN distribution, it produces large
Fig. 5.3-5 Long. load distribution over the sliding zones at the beginning of the contact patch that
contact patch in steady state cond. do not exist in experimental observations. The MCK
 0.012 0.014 0.01s
Time ($1

Fig. 5.3-6 Transient variation of the longitudinal force for small braking steps

Fig. 5.3-7 Transient responses to "non small" braking steps

4-10

Mod' ieems to give satis :tory VsL distributions over

L shown in the :h in Fig. 6. In this case Fv
the sliding zone even if it is very difficult to calibrate. variations are simply proportional to Vc and Vs
variations; therefore to simulate the transient behaviour
of Fx a more refined model is needed.
.........

0 0.05 0.1 0.15 0.2 s,

Fig. 5.3-8 "Replacement time" vs. slip coefficient
t
h W., rm* en.* Uru"r.4 eenmn 4"-

.................................... Fig. 6.1-1 Contact dynamics p(a) model

,w .......................................
,66
lrn[
................................
~
6.2 Brush model

In this section the dynamics of the braking force. for

small disturbances from steady state conditions, is
analytically investigated.
If we approach the tread as a continuum. adhering tread
motion is governed by the following equation [CLARK]:

"
-= V , -RR, -RRe- at;
Jt ax
where c(x,t) is the longitudinal deformation of the
tread,(O; x,z)isaframeofreferencemoving with the
footprint (see Fig. 4-2) and the initial and boundary
conditions are:
c(.,o)=r,,(X,O); t;(O,t)'O. (6)
Linearization of Eq. (5) with conditions ( 6 ) (with the
assumption that R, is constant) leads to the equations:
de de
-
at
+
= v C - OJR,(I U,,) - R,, R,-
ax (7)

e(x,O) = 0 ; e(0,r) = 0
&. L a
7 1 0.k O.&
Drn.18,
o.& 0 . k o.&
where the symbols with subscript "0" indicate the
reference steady state condition quantities and the lower
case symbols indicate small variations around the
reference condition, i.e.:
e(x,t)iA{(x.t); Y C - A V C ; w * d R .
Close to the free rolling condition u,,=Oand the tread
slidingportion can be disregarded; so Eq.(7) is valid over
the whole footprint and can be solved, using the Laplace
transform technique, yielding the following result:
e(s.x) = Y S b){I-exp[-(xs)/(R, Q,,)]}/s (9)
where e(x,s) is the Laplace-transform of e(x.r) and
v~ (s) v,-(s) - R, w ( s ) the Laplace-transform of the
Tim,a,o.014 0.0?8 0.018 input. Now, with the same assumptions as in the K-
Model, by integrating the product K e ( s . r ) along the
Fig. 5.3-9 Transient variation of the long. force footprint, the following transfer function for the
for small steps in wheel speed longitudinal force (f,) can be obtained:

6 . AN ANALYTICAL MODEL OF
CONTACT DYNAMICS
with t9*2a/ReR,,. Fig. 6.2-1 shows the time
6 . 1 p(a) model response to a vs step obtained from Eq. (IO); it should
be noted that this response is identical to that shown in
If F, is expressed as a function of Vc and Vs by Fig. 5.3-2. evaluated numerically for an s,, variation
means of p curves, a linear model can be obtained as from zero to 0.0123.
 4-1 1

1.2
I f ltlt I
1

IF
.,(.............
~

/ . I/
0.8 ................
.\ / \2 :

Fig. 6.2-1 Analytical response on the longitu-

dinal force to a vs step at free rolling
L I
Linearized transfer function and system poles
Linearizing Eq. (10) by means of a first order Pad? Fig. 6.2-2 Contact dynamics brush model
technique, the following expression can be obtained:
Note that in a free rolling condition 2 + c a .
Furthermore, K,=O and V s remain the only input of'
the system (see Eq. (13)); this does not mean that the
system does not respond to V c , as V , = Vc - R R , .
which can be written in the form:
7 . LINEARIZED MODELS OF THE
LANDING GEAR-WHEEL SYSTEM

or, since 2Ka2 = *I

as., S,,=O
(see Eq. (3)), in the form: In this section a preliminary analysis of the linear
dynamics of the landing gear-wheel system shown in
Fig. 7-1 is presented. Certain assumptions are made in
order to simplify the analysis. Since the aim of this
linearized model is to provide a tool for anti-lock system
design, a constant aircraft speed V , together with a
constant vertical load on the gear leg is considered. Such
hypotheses seem quite reasonable, since both vertical
Eq.s (13) and (14) give the transfer function of the load and aircraft speed are slowly varying with respect to
contact dynamics and its correlation with the rolling tyre gear walk frequencies. Besides, the wheel model is
stationary characteristics at free rolling. It is not difficult simplified by disregarding both foundation and rcsidual
to demonstrate that these Eq.s are equivalent to the stiffness.
"relaxation system" proposed in [ZEGEL].
The transfer function in generic braking conditions +
%
(a,#O) can be determined by extending the above
mentioned procedure to the case of a non-negligible Ai2
sliding zone, still with the same assumptions as in the A T -r .
. Ad
K-Model. The dynamic displacement of the transition
point E,,.(t) from the steady state reference position
can be evaluated by linearizing the equation: F.,, AFX

KC(x,,,t) = PU,PN(X,r) (15) WHEEL DYNAMICS

that represents the static friction threshold condition.

The contribution of the adhering part to f , can then be ~G
LANDING GEAR
evaluated by applying Eq. (7) between 0 and T t r ( 0 ) ,the At, DYNAMICS *
contribution of the sliding part integrating the function
p( Vs)P~(x) between Y,,(t) and 2a.
This procedure leads to the linear model of the contact
dynamics shown in the sketch in Fig. 6.2- 1, where:
The transfer functions of the blocks in Fig. 7- I are:

LANDING GEAR 2 = I
M~.S?+C~.~+K~;

WHEEL SPIN
DYNAMICS AQ=L(-AT-
SJ,, F,,,,,A~,.+AFJ~)

0 00 =o ROLLING
P= a,,Ks +(0"+l)K,
00 #O
FRICTION
0"
For the box of the contact dynamics, both the p(o)
4-12

model and the brush model (see sec. 6. I and 6.2) may be Dynamic Simulation and Load Prediction f o r
used. a Flexible Aircraft in Taxiing. Tuke-off and
In generic braking conditions, poles and zeroes of the Landing, AAAF - Proc. of the Int. Forum on
whole landing gear-wheel system, based on the brush Aeroelasticity and Struct. Dynamics,
model of the contact dynamics, have not been Strasbourg, v.2, pp. 839-856, 1993.
investigated in detail yet. When the p(o) model is [DOYLE] G . R . D o y l e , A Review <;f Computer
used, the presence of the wheel augments the period and Simulations f o r Aircraft-Surfcice Dynamics,
the dumping of the two complex poles of the landing AAIA Journal, v. 23, pp. 257-265, 1986
gear; a third real and stable pole, due to the wheel spin
dynamics, is also present and moves towards the origin [DUGOF] H . D u g o f f , P.S.Fancher, L.Sege1, A n
as (ap/ao)l0 increases, becoming unstable when Analysis of Tire Tractiori Properties arid Their
(a,u/ao)lo<O. When the brush model is used, at free Influence on Vehicle Dynamic Performance,
rolling one more pole, real and stable, due to the contact SAE Paper n o 700377, 1970, i n SAE
dynamics, is present. Transactions 1970.
It is interesting to note that, with the simplifying [ENRIG] J . J . E n r i g h t , Laboratory Siiuulation of
hypotheses assumed, it is possible to analytically Landing Gear Pitch-Plane Dynamics. SAE
express the poles of the wheel system as functions of Paper no 851937, 1985, in Selected Paper
system parameters. SAE PT-37, Ed. by J. A. Tanner, 1989.
[FANCH] P.S.Fancher, C.C.Macadam, Computer
8 . CONCLUSION Analysis of Antilock System Perfiormance in
the Braking of Commercial Vehicles, Braking
The review of the literature has shown that investigation
of Road Vehicles, I m e c h E C o n f .
into gear walk is difficult and few papers on specific
Publications, Published by Mechanical
problems of aircraft braking are available. Concerning
Engineering Pub. Ltd., 1976.
tyre modelling, a rigid ring model with residual
stiffness, for tyre structure, coupled with an unsteady [GERDE] J . C . G e r d e s , D.B.Maciuca, E.Devlin,
brush model, seems to be a good choice for numerical J.K.Hedrick, Brake Systeni Modelling ,for
investigation into gear walk. IVHS Longitudinal Control, Advances in
In the present paper a numerical unsteady brush model, Robust and Nonlinear Control Systems,
based on a Lagrangian approach, and an analytical model ASME, Dynamic Syst. & Control Division
of the tyre in contact dynamics have been developed. The DSC V. 53 pp. 119-126, 1993.
numerical model has been extensively tested providing a [GILES] N.A.Giles, Integrated Braking Systems, -
good qualitative understanding of contact phenomena. Aerospace Design Engineering, June 1993.
The analytical model is an extension of the ones in the
literature, as it is able to take velocity variations of the
[HIRZE] E.A.Hirze1, Antiskid and Modern Aircraft,
SAE Paper no 720868,1972.
wheel hub into account; this is essential when such a
hub is connected to an elastic landing gear leg. [HOHNO] H . 0 h n o , T . S u z u k i , K . A o k i . A.
Furthermore the poles of the brush subsystem in generic Takahasi, G.Sugimoto, Neural Netkrwrk
braking conditions are given. Cont ro 1 f o r A u torna tic B raking Con t ro I
System, Neural Networks, v. 7, n o 8, pp.
REFERENCES 1303-1312, 1994.
[HUANl] S . H u a n g , C.Su, In-Plane D~ncimics of
[BAKKl] E . B a k k e r , L.Nyborg, H.B.Pacejka,
Tyres o n the Road Based on a i l
Tyre Modelling for Use in Vehicle Dynamics
Experimentally Verified Rolling Ring Model,
Studies, SAE Paper no 870421, 1987, in SAE
Vehicle System Dynamics, v. 21, pp. 247-
Transactions 1987.
267, 1992.
[BAKK2] E . B a k k e r , H.B.Pacejka, L.Linder, A
New Tyre Model with an Application in [HUAN2] S . C . H u a n g , The Vibration of Rolling
Vehicle Dynamics Studies, SAE Paper no Tyres in Ground Contact, Int. J. of Vehicle
890087, 1989, in SAE Transactions 1989. Design, v. 13, no 1, pp. 78-95, 1992.
[BOWMA] J . E . B o w m a n , E.H.Law, A Feasibility [LINDO] W . C . L i n , D.J.Dobner, R.D.Fruechte,
Study of an Automotive Slip Control Braking Design and Analysis of an Antilock Brake
S y s t e m , SAE Paper no 930762, 1993, in Control System with Electric Brake Actuator, I
SAE Transactions 1993. Int. J. of Vehicle Design, v. 14, no I , pp. 13-
43, 1993.
[CLARK] S.K.Clark, Mechanics of Pneumatic Tires,
National Bureau of Standards, Depart. of [MAUER] G . F . M a u e r , G . F . G i s s i n g e r , Y.
"

Commerce Washington D.C., 1971. Chamaillard, Fuzzy Logic Continuous unci

[CURRE] N.S.Currey, Aircraft Landing Gear Design: Quantizing Control of an ADS Braking
Principles and Practices, AIAA Educational S y s t e m , SAE Paper n o 940830 i n SAE
Series, 1988. Special Publication SP- I O 18.
[DENTI] E . D e n t i , C.Brogi, L.De Francesco, [MIYAS] N . M i y a s a k i t M . F u k u m o t o , Y .Sogo
 4-13

H.Tsukinoki, Antilock Brake System ( M - 21, 1991.

ABS) Based on the Friction Coefficient
Between the Wheel and the Road S u f a c e ,
. .
[SGONG] S . G o n g , A. R Sa v koo r , H B. Pacej k a,
The Influence of Boundan Conditions oii the
SAE Paper no 912174, 1991. Vibration Transmission Properties of Tires,
[MOSEL] D.D.Moseley, T.J.Carter, Performance SAE Paper n o 931280, 1993, i n SAE
Testing on an Electrically Actuated Aircraft Transactions 1993.
Braking System, SAE Paper no 881399, [SHURI] D . J . S h u r i n g , W.Pelz, M.G. Pottin-
1988, in Selected Paper SAE PT-37, Ed. by g e r , The Bnps Model - A n Airtoiir~itecl
J. A. Tanner, 1989. lniplenientation of the Magic F o r i n i r l ~
C o n c e p t , SAE Paper no 931909 in SAE
[NEGRU] D.Negrut, J.S.Freeman, Dynamic Tyre
Special Publication SP- 1016.
Modelling f o r Application with Vehicle
Simulation Incorporating Terrain, SAE Paper [SLAGM] M.T.P.van Slagmaat, Tire Models i n
no 940223 in SAE Special Pub. SP-1016. Aircraft Landing Gear Simulation, Tyre
Models for Vehicle Dynamics Analysis,
[OOSTE] J.J.M.van Oosten E.Bakker, Determina- supplement to Vehicle System Dynamics, v.
tion of Magic Tyre Model Parameters, Tyre 21, 1991.
Models for Vehicle Dynamics Analysis,
supplement to Vehicle System Dynamics v. [TANCH] H.Tan, Y.Chin, Vehicle Antilock Brciking
21, 1991. and Traction Control: A Tlieorerictrl Strtd.~.
Int. J. of Systems Science, v. 23, no 3. pp.
[PACE11 H . B . P a c e j k a , E.Bakker, The Magic 35 1-365, 1992.
Formula Tyre Model, Tyre Models for
Vehicle Dynamics Analysis, supplement to [TANNE] J . A . T a n n e r , Review of N A S A Antiskicl
Vehicle System Dynamics. 2 1, 199 1. Braking Research, SAE Paper no 821393,
1982, in Selected Paper SAE PT-37, Ed. by
[PACE21 H . B . P a c e j k a , R.S.Sharp, Shear Force J. A. Tanner, 1989.
Development by Pneumatic Tyres in Steady [TARTE] J . F . T a r t e r , Electric Bruke S?:stenr
State Condition: A Review of Modelling Modelling and Simulation, SAE Paper n o
Aspects, Vehicle System Dynamics, v. 20, 91 1200, I99 I , in SAE Transactions I99 I .
pp. 121-176, 1991.
[WATAN] M . W a t a n a b e , N . N o g u c h i , A N e w
- -
. . . .
[PERRO] J M P e r r o n n e , M Re nne r , G G iss i n - Algorithm for Abs to Compensate f o r R o d -
ger, On Line Measurement of Braking Torque Disturbance, SAE Paper no 900205, 1990, in
Using a Strain Sensor, S A E Paper no 940333 SAE Transactions 1990.
in SAE Special Publication SP-1018.
I . . [SAKAl] H.Sakai, Theoretical and Experimental
[YEH-I] E . C . Y e h , C.Kuo, .Sun, CoirjugLite
Boundary Method for Control Larv Design of
Study on the Dynamic Property of Tyres, Part Antiskid Brake Systems, Int. J. of Vehicle
I : Review of Theories of Rubber Friction, Design, v. 1 1 , no 1 , pp. 40-62, 1990.
Int. J. of Vehicle Design, v. 2, no I , pp. 78- [YEH-2] E.C.Yeh, G.C.Day, A Paranrerric Stird,~
110, 1981. of Antiskid Brake Systenis Using Poincari
[SAKA2] H.Sakai, Theoretical and Experimental Map Concept, Int. J. of Vehicle Design, v.
Study on the Dynamic Property of Tyres, Part 13, no 3, pp. 210-232, 1992.
2: Experimental Investigation of Rubber [ZANTl] A.van Zanten, R.Erhardt, A.Lutz,
Friction and Deformation of a Tyre, Int. J. of Measurement and Simulation of Trunsient
Vehicle Design, v. 2, no 2, pp. 182- Tire Forces, SAE Paper no 890640, 1989.
226,198 1.
[ZANT2] A.van Zanten, R.Erhardt, A.Lutz,
[SAKA3] H.Sakai, Theoretical and Experimental Measurement and Simulation qf Trunsients in
Study on the Dynamic Property Of Tyres,
i Parr 3: Calculation of the Six Components of
Longitudinal and Lateral Tire Forces, SAE
Paper no 900210, 1990, in SAE Transactions
Force and Moment of a Tyre, Int. J. of 1990.
Vehicle Design, v. 2, no 3, pp. 335-372,
1981 [ZEGEL] P . W . A . Z e g e l a a r , S . G o n g , H . B .
Pacejka, Tyre Models f o r the Stricly of In-
[SAKA41 H.S a k a i , Theoretical and Experimental
Plane Dynanzics, Tyre Models for Vehicle
Study on the Dynamic Property of Tyres, Part Dynamics Analysis, supplement to Vehicle
4: Investigations of the Influences of Running
System Dynamics, v. 23, 1993.
Conditions By Calculation and Experiment,
Int. J. of Vehicle Design, v. 3, no 3, pp. 333-
375, 1982.
[SAVKO] A . R . S a v k o o r , Boundary Conditions on
Models f o r Predicting Tyre to Road Traction,
Tyre Models for Vehicle Dynamics Analysis,
supplement to Vehicle System Dynamics, v.
- .
 5-1

Fuselage Vibration Control Using Semi-Active Front Gear

H. Wentscher, W. Kortiim, W.R. Kriiger

DLR - Deutsche Forschungsanstalt f i r Luft- und Raumfahrt

Institut f i r Robotik und Systemdynamik, Posgach 11 Id
D - 82234 Wessling, Germany

Summary: room for improvement. With the advent of microelectronics

both in digital computers and in controller development the
This paper presents results of research performed in the field of idea of a controlled landing gear gained new momentum. Both
semi-active suspensions for advanced landing gears and opti- systems controlling strut stiffness and damping coefficient
mization of their associated design parameters to achieve mini- (fully active) and controlling only the damping parameters
mum weight, maximum comfort under strict requirements with (semi-active) are subjects of current research.
respect to safety and even increase lifetime by reducing the As stated above, aircraft landing gears are crucial for safety,
loads during landing impact and taxiing. comfort (both for passengers and pilots) and for weight consid-
erations. As the device responsible for safely moving the air-
A detailed mechanical model is derived for an existing aircraft craft on ground, the landing gear has to fulfill several,
(AIRBUS A300) which will be used for simulating the stan- sometimes conflicting, needs. Jenkins, [11and Young, [2] have
dard design as well as the active optimized landing gear. The given a detailed presentation of those requirements. In short,
simulation is performed with SIMPACK, DLRs prime multi- the landing gear must on one hand absorb vertical and horizon-
body computer code. tal energy during landing impact, and on the other hand keep
the aircraft in a stable position during ground manoeuvres.
Results will be presented on using the multi-objective parame- Commercial aircraft should also provide a smooth ground ride
ter optimization software ANDECS. The design case study will during taxiing both for passenger comfort and safety reasons.
concentrate on taxiing of a flexible aircraft. Reduction of the
so-called beaming effect (e.g. dynamic coupling of runway 1.2 Landing Gear Simulation
- excitation with elastic fuselage eigenmodes) is the major
design goal. The design of a new aircraft takes a long time span which can
last up to a decade from the first proposals to customer deliv-
ery. The development process includes different manufacturers
1. Introduction and suppliers who provide many components to the aircraft.
Simulation has long been a necessary step in the development
1 . 1 Landing Gear Requirements process. In addition to mechanical simulation (e.g. test rigs)
numerical computer simulation gains more and more impor-
Among the numerous complex components that make up an tance and acceptance.
aircraft the landing gear is often regarded only as a bothersome, The simulation used in the work presented in this paper was
but necessary attachment. It must not be forgotten that, while based on the Multi-Body-System (MBS) simulation tool SIM-
the predominant task of an aircraft is without doubt to fly with PACK. SIMPACK is the central MBS-tool of DLR and it is
the best performance achievable, it will also spend a good part being applied there for aircraft (landing gears), robots, space-
of its life on the ground. Typical cases are taxiing as well as craft structures, railway and road vehicles.
take-off and landing. According to todays airlines specifica- SIMPACK offers fast numerical analysis capabilities due to an
tions an aircraft should reach up to 90 000 take-offs and land- extended O(N)-algorithm for tree-configured rigid body MBS.
ings as well as 500 000 km of ground roll during its life time. Dealing with closed loop systems the equations of motion are
Statistics also show that accidents prior to or directly after take- accompanied by a minimal set of algebraic constraint equa-
off and touchdown relate their fair share (more than 50%) to tions, resulting in a differential-algebraicdescription of the sys-
the overall numbers. Hence, the importance of aircraft ground tem.
handling and therefore of landing gear design should not be For the examination of elastic vibration control it is of course
underestimated. necessary to use elastic instead of rigid bodies in the model
Tests on life-size aircraft are obviously expensive and risky, setup. In SIMPACK the kinematics of elastic bodies as well as
and tests on test-rigs (namely drop-test facilities) allow only stress stiffening effects of elastic deformations are taken into
limited deduction of information about the landing gears account.
dynamics; especially the interaction between aircraft and land- Extensive libraries of coupling elements like joints and force
ing gear is difficult to assess. On the contrary, simulation offers elements as well as excitations aid the engineer in setting up a
a means to examine the behavior of the airplane as a complex model. This includes the kinematics of different suspension
system in its environment at a reasonable cost. With a modem systems or other complex joints, force models for hydraulic
computer mechanical systems with a large number of degrees components, various tire models etc.; subsystem modelling
of freedom, both for rigid and elastic body motion, can be mod- techniques enable the user to establish complex non-standard
elled. kinematic and force laws. User written subroutines extend the
The conventional landing gear type for large aircraft, consist- modelling options.
ing of a set of tires, sometimes a bogie, and almost always an
oleo pneumatic shock absorber (often abbreviated as oleo)
today is a highly sophisticated device that leaves only limited

Paper presented at the 81st Meeting of the AGARD SMP Panel on The Design, QualiJcation and Maintenance
of Vibration-Free Landing Gear, held in Ban& Canada from 4-5 October 199S, and published in R-800.
5-2

7 1 ProENGINEER
Level 1
A CAD I

I
NAsTRAN \ / MATLAB
ANDECS User
Code
Block

Level 2 SystemBuild

I I
Figure 1: SIMPACK and its Interfaces Figure 2: SIMPACK - MATRIXx Interface

SIMPACK Interfaces: 1.3 Landing Gear Control

One of the most important features of SIMPACK is the fact that Active control of vehicle suspensions is not only an issue in the
SIMPACK is an open system which possesses various inter- aircraft landing gear development. Cars, trucks and railway
faces to external standard software products (see figure 1) from vehicles are the major applications to profit from active suspen-
the domain of finite element (FEM), computer aided design sions, and great efforts are made among the respective manu-
(CAD) and control system analysis (CACE) programs. facturers to develop reliable and affordable systems.
The principles are largely independent of the specific vehicle . -
FEM: employment. Although passive suspensions, mostly consisting
To read in arbitrary flexible geometry a file interface to FEM- of a spring and a damping device, today have a high standard, -
programs was developed. This gives access to mass, stiffness they suffer from a disadvantage that lies in their principle.
and damping matrices and to the above mentioned terms of Optimized to isolate the vehicle body from the ground best at a
second order. Vice versa, loads computed by SIMPACK can be certain frequency range, their performance often diminishes at
transferred to the FEM-code. other frequencies. Additionally, the need to retain tire ground
contact at all times poses a design conflict with comfort
CAD: requirements.
For the purpose of incorporating physical and graphical CAD- For an aircraft the implications of the phenomena mentioned
data, SIMPACK can be linked by a function call interface to above is noteworthy: designed to absorb the energy of a hard
CAD-packages. This means an addition to the interactive SIM- landing impact, aircraft suspensions perform quite poorly in
PACK model setup tool. The link enables data consistency reduction of ground-induced loads during taxi and take-off.
between MBS- and CAD-data at each step of the model estab- Not surprisingly, supersonic aircraft and a new generation of
lishing process. In general this feature is not guaranteed by just stretched civil transport aircraft suffer the most under ground-
using file interfaces. induced structural vibration because of their increased struc-
tural flexibility inherent in their design with slender bodies and,
Control System Design and Analysis: at supersonic aircraft, their relatively thin wings.
Due to its ability to numerically linearize the system equations, Active and semi-active suspensions promise a solution to this
SIMPACK can be used as a simulation block within linear con- problem. An active suspension can be defined as a suspension
trol design tools. SIMPACK may also be linked as a fully non- layout which controls the forces acting in the shock absorber
linear block into nonlinear optimization and control tools like by control of energy dissipation, or, if required by the control
MATRIXx, MATLAB,ANDECS, etc. (figure 2 shows these law, by generating additional force. The usual means of control
links for the connection of SIMPACK to MATRIXx). is a closed-loop control with a control law acting with respect
This gives the user the possibility to use SIMPACK and its to measured states (often velocities or accelerations) at certain
parameter variation capability in a closed optimization loop to points of reference. A suspension termed semi-active has the
efficiently design dynamical systems. restriction that it cannot input energy into the system. Usually
realized as damping control, it is therefore only able to control
the amount of energy dissipation. Only forces of the same
direction as that of the instantaneous relative damper velocity
can be generated. Nevertheless, since fully active control sys-
 5-3

tems usually require a heavy and costly force generation device A study of the properties of fully and semi-active oleos led to
(for aircraft landing gears mostly pressurized oil reservoirs the decision to take the semi-active damper as the system of
have been proposed) they are unlikely to be quickly introduced choice (see table 1) for the work presented here, see [IO].
in production aircraft. Semi-active suspension systems offer
considerable advantages of light weight and less complicated
mechanical requirements without suffering a great loss in per- passive semi-active active
formance. low weight low weight high weight
Summarizing, the main reasons for the introduction of semi- good performance good performance good performance
active landing gear control are: only for design over broad range over broad range
1, Minimize the load on the airframe structure, minimize case
force peak values and vibrations that can result in relative low com- medium com- high complexity
fatigue and reduce the life of the airframe. plexity plexity
2. Minimize accelerations acting on pilot and passengers
stable stable potentially un-
since the induced vertical and horizontal accelerations
stable
and vibrations can lead to passenger discomfort and
crew disorientation.
Table 1: System Features Comparison
1.3.1 Activities
There have been several preceding efforts to the improvement The semi-active damper seems to provide the best compromise
of aircraft of ground ride. A typical measure to adopt the land- between performance and additional weight handicap. The
ing gear force-deflection curve toward a lower slope is the use investigations of ground-load reduction are concentrated on the
of a two-chamber-oleo. In 1977, Somm/Straub/Kilner [4] pro- nose gear, since results show that this has the most potential for
posed an adaptive landing gear system for several military ride quality improvement.
transport aircraft to improve taxi performance on rough run-
ways. They worked with a secondary air chamber that could be 1.3.2 Control LAWS
pressurized shortly after touchdown in order to generate the Different control laws are investigated in the literature, mostly
desired softer pneumatic spring rate as a function of aircraft for cars and trucks. During aircraft operations, however, associ-
weight. Today, with reliable, inexpensive and powerful elec- ated parameters change during a wide range. Take-off weight
tronic signal processing available, research has moved toward will always differ drastically from landing weight, aircraft
. computerized closed-loop control. In 1984 a feasibility study ground velocities vary between very low speed at taxi and 60 to
for a series-hydraulic active control landing gear intended for 80 m/s at take-off and landing. Oleo and tire properties may
supersonic military aircraft was published by McGehee and vary depending on age and maintenance. Finally, runway qual-
- - Moms, [ 5 ] . Additionally to theoretical analysis a number of ity differs from airport to airport.
tests had been performed on a test-rig to permit experimental Suitable control laws have to be robust against these changes.
verification of the concept. The setup is typical for a fully Several solutions to the problem exist and have been discussed
active control system and has before and since been investi-
I gated several times. The gear force applied to the airframe is
regulated by the hydraulic pressure in the piston of the oleo
thoroughly, ranging from sky-hook damping to nonlinear and
adaptive control laws (for a state-of-the-art review see [ 1I]).
For many problems the algorithms have to go further than to
which is used as an actuator. solely observe center of gravity motion. Other points influenc-
The results obtained led to the modification of a F106 nose ing the results are maximum applicable actuator force and sus-
landing gear which was also tested in test-rig setups, [6]. pension travel limitations.
Another investigation that combined an analytical and an For the problem discussed in this paper, the sky-hook damping
experimental approach was conducted by Freymann, [7]. approach has been chosen.
Object of research was the simulation of a nose gear oleo.
Again the force generation was realized by pressurized oil, 1.3.3 Hardware
even though the valve layout was different. A laboratory setup
Since the idea of active and semi-active landing gear is rela-
was used to experimentally verify the basic control concept.
tively new, and no system for aircraft is as yet in the production
Even though the results obtained in both cases might be overly
stage, the hardware is still in the phase of concept evaluation.
optimistic when it comes to application to a realistic aircraft,
In the field of fully active oleo control so far heavy laboratory
the investigations showed nevertheless that a fully active land-
setups have been used to validate simulation results.
ing gear is feasible and can lead to significantly reduced air-
For semi-active control systems, some manufacturers, espe-
frame loads. Another somewhat more recent approach is the
cially in the domain of heavy-duty trucks, have proposed and
analysis of improvements gained with closed loop semi-active
built technically feasible layouts. Other suggestions include dry
oleo control. Studies by Kamopp [8] for automotive applica-
friction damping (which, for large aircraft, is likely to be unre-
tions show that the performance of a semi-active damper is
alistic due to the large energy dissipation). A relatively new
only marginally smaller than that of a fully active system, pro-
proposal is the use of electro-rheological fluids, i.e. fluids that
vided that an adequate control law is used. Catt/Cowling/Shep-
change their viscosity in response to an electrical field.
herd [9] come to a similar result in a simulation study of
Objectives for the final choice of a system will surely be led by
aircraft suspensions. The semi-active damping device is usu-
a consideration of performance against penalty. Surely low
ally proposed to be an oleo with a modulated damping orifice
weight is desirable, and high safety against failure is essential.
cross-section, completely omitting the metering pin.
A good performance also requires a high bandwidth, low reac-
5-4

tion time actuator. The possible achievable actuator peak force Hertz band above and below the critical range. A list of the
might also be a decisive criterion. rigid and flexible eigenmodes can be found in [lo].
An active controlled landing gear introduced to improve the
ride quality of a large aircraft would therefore have to concen-
2. Motivation trate on reduction of vertical accelerations due to elastic body
oscillations in the 4 to 8 Hertz frequency range. The excitation
2.1 The Problem of these oscillations is caused by runway roughness either due
to wear and tear or to other unavoidable reasons. Qpical exci-
2. I . I Ride Comfort Assessment tations used in the simulations are presented in chapter 3.1.4.
In chapter 1 it could be seen that the improvement of crew and
passenger ride comfort will be of growing concern for manu- 2.2 A Proposed Solution
facturers of large aircraft.
As base measurement for ride comfort assessment the vertical Recent research indicates (see chapter 1) that active gear tech-
accelerations are widely used. Since the individual perception nology may pose a possible solution to the problems of aircraft
is additionally influenced by other factors of the biomechanical ground loads and that the semi-active controlled front landing
human system, scales have been developed to weight the fact gear promises to be the most likely solution to be incorporated
that certain frequencies are perceived to be more uncomfort- in a production aircraft by the aircraft industry.
able than others for a given amplitude. In IS0 2631, [12] the The proposal of this paper to solve the ride comfort problem is
frequencies between 4 and 8 Hertz are denoted as the most cru- therefore: design the controller for a front landing gear, modify
cial for comfort (figure 3). Ride comfort improvement has a conventional landing gear to incorporate controlled damping
(thus creating a semi-active front landing gear), analyze it and
compare it with the conventional gear.

Acceleration a, (rms), lm/s21 2.2.I Multi-Objective Parameter Optimization using

MBS Models
8. -- The main part of the controller design, the optimization of the
controller parameters, was performed using ANDECS, [ 131,
developed also at the DLR. ANDECS is an open Computer
Aided Engineering (CAE) software for the analysis and design
of controlled systems. It offers interfaces with other CAE-tools,
Ih
in particular with modelling environments such as SIMPACK,
ACSL and DYMOLA. Besides these interfaces ANDECS pro-
2Sh vides as its core design environment MOPS (Multi-Objective
.5 -\ 4h Programming System).
8h
Multi-objective parameter-optimization has become more and
+. more a central design strategy for complex dynamical systems
where best solutions are searched in the parameter space
.-- such that certain - sometimes conflicting - performance specifi-
cations have to be met, [ 141.
The design strategy for multi-objective parameter optimization
Frequency [Hertz] is briefly sketched in figure 4. As usual the dynamic system is
Figure 3: Fatigue-Decreased Proficiency Boundary (IS0 2631-
1974(E))

therefore to concentrate on vertical accelerations in this fre-

quency range, measured in the cockpit and passenger compart-
Engineer Desision Domain
I
ment.

2.1.2 Airframe Eigenfrequencies

Normal accelerations of the airframe originate in longitudinal
rigid body modes as well as in elastic airframe eigenmodes.
Rigid body eigenmodes:
The important rigid body eigenmodes are found at frequencies
where humans are less susceptible. The rigid airframe eigen-
mode of large aircraft in vertical direction is likely to be at a Figure 4: Synthesis Loop of Optimization
frequency below one Hertz, whereas the eigenfrequency of the
unsprung mass (i.e. wheels, brakes, bogies), also denoted as
wheel hop frequency, is found in the range above 50 Hertz. defined by the plant and the controller models; both are com-
Elastic airjrame eigenmodes: pletely free and can be linear or nonlinear as appropriate.
Five eigenmodes display frequencies where the human body is A number of so-called synthesis-parameters, e, are set free
most sensitive, with an additional four eigenmodes in a two within certain limits; they are tuned to yield an optimal
design. It is indeed arbitrary whether these are some (or all)
control law parameters or some parameters characterizing the The following chapter will cover the model setup, the optimal
mechanical system. e.g. damping or stiffness parameters of controller design as result of the optimization process will be
passive suspension. The various performance criteria ci. with presented in chapter 4.
respect to which the parameters should be optimized (mini-
mized), are summarized in a criteria vector:
T
c = [ C 1 C 2...
il

The complete strategy exists of two imbedded loops. The outer

loop is initialized by the system definition and the criteria. The
inner loop stans on the basis of simulation and analysis initial-
ized by a provisionary parameter vector pa with a correspond-
ing criteria value CO. Figure 5: Software Interaction in the ANDECSlSlMPACK
Design Environment
The desired design-direction is defined by a design parame-
ter vector (design director), d:
7

d = [dl,d2 ...,d i l l 3. Modelling

3.1 Aircraft Model Built-up

where di is a chosen design command level corresponding to
criterion ci; di are upper limits for the desired stepwise descent 3.1.1 Modelling Approach
of the corresponding criteria: The AIRBUS A300, which was selected for the model buildup,
represents a configuration widely used for commercial airlin-
c 1. ( -p ) I d 1. i = 1,..., 1 (2.3) ers: two wheel front gear, two four wheel bogie main gears,
wide body fuselage, two engines and a maximum take off
As a design comparator for detecting better designs the max- weight around 140 metric tons.
function The literally central body of the AIRBUS model is the wing
sparframe, the fuselage frame situated at the connection of the
C. aft wing spars to the fuselage. It is modelled as a rigid body
1
a = max- i = I, ..., 1 (2.4) without mass or geometric extent. It serves as connection point
di between the elastic airframe bodies and the lneliial Reference
Frame. The landing gears, built up by several smaller entities.
is chosen. The strategy is to find a minimum for this a in vary- are linked up to the airframe bodies on the one side. On the
ing the synthesis parameters p: other side no joints, but tire force laws form the attachment to
* the runway surface.
a = min (a@)) (2.5)
P 3.1.2 The Elasric Airframe Bodies
For the incorporation of elastic bodies, the MBS package
The implemented strategy in MOPS does not only guarantee SIMPACK contains the possibilities to include FEM and beam-
the monotonous descent of all criteria but also tries to solve like stmctures. For the use of the BEAM preprocessor the body
efficiently the optimization goal for which a(e) is minimal. is divided up into several sections with piecewise constant stiff-
After a sufficientnumber of steps within the automatic synthe- ness, damping and distributed mass. This feature and the shape
sis iteration loop, an improved design is reached which is of the four airframe members modelled as elastic bodies (two
Pareto-optimal, i.e. no criterion can be fuliher improved with- wings, front and aft fuselage) made its usage feasible for the
out deterioratingothers. A300 model.
At this stage certain decisions are possible, e.g. to change the
design director in order to change the design direction. Thereby Fuselage:
the designer gains experience which criteria can be easily The fuselage model consists of a front fuselage pan with cnn-
improved and where the critical design conflicts (within the nection to the front landing gear and the aft fuselage part with
chosen stmcture) are located. Other decisions may be the the rigid elevator and horizontal stabilizer bodies. Both pans
change of the criteria or of the controller stmcture or the syn- are linked to the wing spar frame with zero degrees of freedom
thesis parameters. Within a number of such experiments the joints. They are divided into five sections each. The character-
potentials of the chosen stmcture is accumulated. istics of the fuselage were derived from design material pro-
The software modules and their interrelations as used for this vided by the airframe manufacturer.
paper are shown in figure 5.
The overall control of the computational process is done by wings:
MOPS; time-simulation of the nonlinear model is executed by Like the fuselage, the wing is partitioned into five sections.
DSSIM (Dynamic Simulation Module); the generation of the connected to the wing spar frame on the inner side and left free
MBS and controller equations is performed by SIMPACK. on the outer. The engine is linked rigidly to the third node, the
The end product of the optimizing process, the. satisfactory second node includes the attachment point for the main landing
compromise, is then used to visualize the final model behavior. gear. Only bending in z is taken into account for the wing bod-
ies.
5-6

Fed with this data, the BEAM prepmcessor calculates the * load variations small enough to allow linearization of defleo
desired eigenmodes of the four airframe members. These tion curve.
eigenmodes are then used by SIMPACK to evaluate the aircraft This tire model was used for most of the taxi simulations.
eigenmodes implemented in the simulation. The selection of
the number of eigenmodes plays an important role. High struc- Oko-Pneumatic Shock Absorber Force Imv:
tural vibrations result in small integration stepsize and thus Conventional (Passive) System:
increase CPU work. On the other hand the SHANNON theo- As primary force generating element each aircraft landing gear
rem determines the maximum eigenfrequency implemented in is equipped with one central oleo, acting on the principle of a
a cenain model to be at least twice the maximum frequency one stage or single slope oleo like those used in the landing
under consideration ( 1151 and 1161). The eigenmodes for the gears of the A300 aircraft. The passive oleo displays two sepa-
taxi analysis were chosen in a way to guarantee a safe analysis rate force generating elements, the gas chamber that generates
up to 25Hz. the spring force and the set of orifices responsible for the
damping force. The total oleo force is composed of a gas spring
part (calculated using the general gas law). a damping part and
a friction part.
The damping force was calculated using the squared value of
the relative velocity of the sliding oleo member and the main
fitting attached to the fuselage. It is dependent on the direction
of the relative motion because of different dampingcoefficients
for compression and for extension. The values for theses coeffi-
cients, depending on the oleo orifice ROSS sections, were deter-
mined in tests by the landing gear manufacturer.
Friction forces are difficult to assess and depend on a multitude
of parameters. Especially for the landing impact case, longitu-
dinal forces increase. the friction and can in fact block the gear
and prevent spring deflection. leading to high structural loads
and even structural failure. The gas chamber pressure influ-
ences the fitting pressure on the outer oleo wall, resulting also
in higher friction forces. For the taxi condition the gas pressure
variation and applied x-forces remain small, so a constant fric-
tion term supplied by the landing gear manufacturer was incor-
poIated.
Semi-Active System:
The main difference between a serm-active and a conventional
Figure 6:First and Second Fuselage Bending Eigenmode oleo is found in the oil orifices responsible for the damping
force. To provide control over the damping force, the oil Bow
and thereby the damping coefficient is regulated. This can be
In figure 6, the first two fuselage bending eigenmodes are. achieved by variable diameter orifices or by implementation of
shown. a servo valve. Since the actual design may exhibit quite differ-
ent oil flow regulators, but all of them control the damping
3.1.3 Force Laws coefficient, this coefficient was chosen as control variable. The
To simulate the dynamicalbehavior of the A300 airplane, SIM- control range, i.e. the maximum and nunimum damping coeffi-
PACK had to be given a set of force laws. With the aid of the cient values achievable. was deliberately chosen to lie between
relative displacements and velocities at the specified node 20 and 500% of the original value used in the conventional
pints of the multi body system, the according forces were A300 oleo because this range was considered to be technically
derived. The force laws are rcalizcd as user routines within feasible.
SIMPACK and are explained in detail in [IO].
Aemdynantics Force LAW:
lire force law: Stationary motion of an aircraft through the air results in an
The tire force model has to yield the longitudinal and lateral aerodynamic force acting on the aircraft. Taken into account
forces and the torque moment acting behveen the road surface the small influence of the aetodynamicfaces and moments on
and the wheel and depending on the slip between tire and the overall dynamicsof the system, the d y n a m i c force user
ground Nevertbeless. even with a relatively simple model, routines were limited to the longitudinalmotion. To account for
computational effort has proven to be substantial for an aircraft non-stationary moments, the pitch damping coefficient was cal-
with ten or more wheels. Lkpending on the simulation case it culated.
proved to be helpful to linearize the tire behavior if possible.
For an aircraft taxiing on ground without corneringor braking, 3.1.4 Runway sufme
a linear spring model was incorporated. The following supposi- To analyze the dynamic behavior of a general nonlinear MBS
tions apply: model, time histories have to be evaluated for a broad range of
* negligible slip (no braking). frequencies. The excitations used on the model should include
* no side slip, a wide frequencies range. Additionally, single frequencies of
* no loss of ground contact for any tire, special concern might be agitated separately.
 5-7

The following existing runway Nrbances have been taken A comparison between four measured runway ZDs (thin
into account in the present investigation: lioes) and the PSD of the implemented model runway (solid
General roughness: Especially on older ailports and on line) is shown in figure 9.
airports in the countries of the former Warsaw Pact, the
run- and taxiways display undulations of all frequencies
and with higher amplitudes than those measured on air-
ports in western countries, 1171. These conditions have Power spectru~Density [ft/radian/ftl
reportedly led to dangerous situations. Pilots of aircraft
equipped with Cathode Ray Displays (so called glass
cockpits) experienced difficulties to read the instru-
ments due to vertical accelerations during the take-off
Nn.
Concrete Plate Deforming: A widely used method to
construct fortified runways is the casting of large plates
using Liquid concrete. These plates are separated from
each other by gaps filled with rubber. Aging of concrete
runways causes the plates to settle unevenly, leading to
long wavelength bumps and steps at the gaps.
Center Line Lights: All run- and taxiways are equipped
with lights indicating the middle line during night time
opcsation. These lamps extend a few centimeters from
the ground (figure 7) and exert a shock when hit by a
1 - ModelRunway
-Measured Runway
tire. They are spaced regularly and can induce oscilla-
tions. 10-2 I U iff IO
Reduced Frequency [radianlftl

a: Housing Height (2...5 cm) Figure 9 Power Spectral Density of Measured and Model
Runway
b Length (I5...30cm)
c: Spacing 6.. 15 m)

4
Multiple Pulse Runway Model:
Step and pulse model inputs are especially suitable to induce a
broad band of frequencies into a model. In practical ground
. _ operation. pulse inputs can result from concrete plate gaps or
from center line light housings. In figure 7, a light housing built
Figure 7 Center Line Light Housings in a Runway into the ninway is sketched. In reality. the tire hitting the hous-
ing will deform, so that the impact of the heave motion of the
bottom of the tire spring will be muffled. The multiple pulse
To simulate these runway surface conditions, the following runway model uses this kind of pulse shape. Housing dimen-
three distinct models have been programmed. sions and spacing are programmed as variables and are defined
in the experiment model setup.
Qwi-stochastic Runwny Model
Thompson [17] gave a list of a number of measured Power Harmonic Wave Runway Model:
Spectral Densities (PSD) of runways in NATO countries. These The harmonic wave as excitation was included in the system
were used to build up a deflection curve using a set of harmonic analysis to perform a sort of worst case evaluation. With this
functions. The runway elevation curve resulting frnm this p m runway model, a distinct frequency acts on the landing gear to
cess is displayed in figure 8. induce potentially problematic oscillations like the main fuse-
lage bending eigenfrequencies. A simple sinusoidal wave was
used for the model. In figure IO, an example runway surface is

I
40. 80. im. 160. 200
Runway x-mrdinate [ml I
40. 80. im. 160. zw
Figure 8: Quasi-stochastic Model Runway Surface Runway x-coordinate [ml

Figure 1 0 Harmonic Model Runway Surface

5-8

given used for the excitation of the 3.6 Hertz fuselage bending using the Sensor Signal Processing and the control algorithm
eigenmode of the A3M) at a speed of 20 ds. implemented in the Contmller.
The velocity feedback used as controller input makes the cho-
sen control concept a variation of the sky-hook approach.
4. Control Concept Analysis
4.2 Optimization Process
4.1 Control System Lay-Out
4.2.I Criteria Selection
During the selection process of a control law to be imple- The goal of the implementation of a controlled front gear is to
mented, the rigorous safety requirements in aircraft design increase passenger and pilot comfort by reduction of fuselage
have to be taken into consideration. Thus, low system com- vibration. The criterion vector c must take into account the nec-
plexity is one of the design drivers for the control system. A essary states. The aircraft state vector Y
single input single output (SISO) control system requires the
minimum of one sensor signal and one actuator element and
was therefore chosen.
Aircraft currently in commercial operation are in part already
equipped with acceleration sensors. To keep additional system r= (4.1)
requirements low, a single acceleration sensor at the cockpit
position was assumed and used for the feedback signal. Figure
I 1 shows the resulting system layout.
contains the states describing the aircraft motion. Here,

Y, = [Y:], i = I, ...,nG (4.2)

denotes the position states for each of the nG joints,

Sensor

It/ / I
stands for the velocity states of the joints.

q = [;I , i = I, ...,nK (4.4)

' Semi-active Oleo

describes the deformation position coordinates for each elastic
body nK and

Figure 11: Control System Sensor and Actuator Location (4.5)

The block diagram in figure 12 shows the general principle of expresses the deformation coordinates in the velocity domain.
the feedback system. The Oleo Closure Rate remains the main For calculation and analysis the states of several distinct loca-
tions on the elastic bodies are needed. Unlike on rigid bodies.
where states of any location can easily be derived if the states
of the reference system of the body are known, the states of
arbitrary locations on the elastic body depend on the deforma-
Olrn tion position and velocity cwrdinates.
Closun Aircraft To observe all interesting eigenmode shapes it is necessary to
introduce measurementsof suitable locations on all elastic bod-
Aircraft swcrurc
ies (fuselage and wings). For example, the first fuselage bend-
ing mode can be observed by measurements in the far front or
far rear end of the fuselage, whereas for the second eigenmode
additional information is needed from stations between the
Sensor Signal front and the center of gravity of the aircraft. As a compromise
Conmlkr Recessing between the resolution of high frequency eigenmodes and sim-
ulation time. 16 locations have been selected on the A300 air-
frame.
Figure 12: Control System Block Diagram The semi-active landing gear influences primarily the vertical
motion of the aircraft. Vertical acceleration was therefore
input for the Landing Gear Oleo. It determines the forces picked as criterion state to be measured. The Root Mean
applied on the Aircrafr Sfrucrure depending on the Runway Square (RMS) value is calculated using
Pmfile inputs and the Damping Orifice Area Actuator. The Air-
craft Response coming from the Aircraft Strucfure is fed back
 5-9

The acceleration itself is determined as

with
CLRMS =
Jj(iz-Xm)
dt (4.6)

Yith
(4.13)

r, = vertical acceleration at sensor i (4.14)

xm= reference value for acceleration evaluation.tobe set by
operator where uft) stands for the model input, a function dependent on
ci,,+MS-criterion for s e m i time I and the controller parametersp.
To derive the free system parameters, the optimization strategy
10 = start of time integration interval
shown in figure 13 was used. The system was subjected to the
I, = end of time integration interval

Besides the Root Mean Square value (RMS value), which rates
amplitudes and their duration time, peak values are includes as
well:
cl,pe*k= m a x ( a b s ( x , - x m ) ) (4.7)

where cLwk denotes the peak-eriterion value. SIM'ACK Simulation and Analysis I
With this double criterion strategy, a trade-in of low R M S val-
ues for very high but short peaks is avoided. RMS values and
pealr values of all sensors were weighted equally, with slight cdtcriaq
3 * 16 RMS values of vertical accelerations
emphasis on the fmnt fuselage part 3 * 16 maximum amplituaes of vertical accelerations

4.2.2 Optimimtion Method

The numerical procedure used to solve this nonlinear multi
objective multi parameter optimization problem is based on the
technique of Sequential Quadratic Rogramming (SQP)
described in 1181 and [19] in an implementation in the
ANDES Control Systems Design Environment ( [% [Zl]). I],
?be aim is to determine a set of parametersp in a way that min-
imizes a cast function flxj under consideration of different
equality and inequality constraints g(x). The general form of
the cost function can be stated as
m a*=min [a(p))

Figure 13: Multi Model Optimization Process

f ( x ) = min ; X E %" (4.8)
three different excitation cases which were described in chapter
with the constraints
3 earlier. In the following step, a complete and self-contained
g,(x) = 0 ; j = 1,2,..., m, (4.9) simulation run of the nonlinear model was performed for each
excitation case. The criteria were evalmted using the steady
state time response. After being weighted with the design
gI. ( -x )-> o ; j=rn,+1,2 ,...,m (4.10) parameters d, the optimization run was evaluated and the next
optimization run initiated using changed parameters.
and the parameter domain boundaries
5. Results

5.1 Final Design Evaluation

Cost function and constraints have to be continuous. hut no The optimization process described above was used to derive a
specific structure is requid. For the acceleration minimiza- set of controller parametersbest suited to minimize the design
tion, the cost function can now be stated as parameters 4. On the following pages the system performance
using those controller parameters is shown for the three exdta-
tion cases. The performance is subsequently compared with
that of the original conventionaltype landing gear. A complete
presentation and discussion of the results can be found in [lo].
5-10

5.2 Quasistochastic Excitation Performance gear can especially be seen for the first fuselage bending eigen-
mode at 3.6 Hertz and for the second symmetric and antimetric
In figure 14 the vertical cockpit acceleration response of both wing bending eigenmode at 5.6 and 5.7 Hertz, respectively.

5.3 Center Line Light Excitation

Vertical Cockpit Acceleration [mls21 Figure 16 shows the system behavior for the repeated pulse
excitation as exerted by center line light cases or concrete plate
gaps.

Power Spectral Density of Cockpit Acceleration [mz/ s31

lo 7
-1 . -Conventional Nose Gear
- Semi-active Nose Gear

0. 1. 2. 3. 4. 5.
lime [SI
-Conventional Nose Gear
Figure 1 4 lime Domain Response for Quasistochastic
Excitation
10-31
- Semi-active Nose Gear

the conventional gear (thin line) and the semi-active gear @old 0 IO 20 30
line) in the time domain are shown for a total simulation time Frequency [Hertz]
of five seconds. The system state at the beginning of the simu- Figure 1 6 Power Speceill Density of Center Line Light
lation was determined by a pre-analysis time integration to Excitation Response
guarantee a steady state response throughout the whole analy-
sis period. It can be seen that no fading oscillations are present.
The controlled gear performs well, the amplitudes of the accel- While the system response of the semi-active gear remains
eration are diminished by a factor of five. below the one of the passive gear over the whole frequency
The power spectral density (PSD) of the cockpit accelerations range, two distinct peaks can be seen below lOHz for the air-
of this time history. shown in figure15 for the frequency range craft structural eigenmodes. At those points the system perfor-
mance of the semi-active oleo retains only a small advantage
over the conventional system. This behavior differs from the
one seen for quasistochastic excitation and may be amplified
Power Specaal Density of Cockpit Acceleration [m2/s31 by the total absence of inputs between the bumps and on the
main gears.

-Conventional Nose Gear 5.4 Harmonic Excitation

- Semi-active Nose Gear
The frequency domain response for the harmonic excitation in
figure 17 shows that a high level of accelerations at the fre-
quency of the exciting undulations of the runway for the con-
ventional gear has to be expected (thin line). Therefore the PSD
in this frequency range exhibits high levels. The semi-active
gear (bold line) transforms the response into small amplitudes
10-3
over a wide frequency range, leading to higher PSD levels at
higher frequencies than the conventiooalgear.
I I
It must be taken into account, though, that the total absence of
higher frequencies in the excitation was deliberately chosen to
0 IO 20 30
analyze this excitation form and will be accompanied by higher
Frequency [Hertz] or lower levels of distributed runway inputs in each real run-
way excitation.
Figure 1 5 Power Spectral Density of Quasistochastic Nevertheless it can be stressed that the semi-active system per-
Excitation Response
formance displays an impressive lead in the low frequency
range over the conventional gear. This result is even more
from 0.1 to 30 Hertz, displays significant improvements over important bearing in mind the susceptibilityof the human body
the whole frequency range. The influence of the semi-active with its emphasis on the low frequency range.
 5-11

Power Spectral Density of Cockpit Acceleration Im2/ s31 Power Spectral Density of Cockpit Acceleration [m2/ s3]

-Conventional Nose Gear

-Semi-active Nose Gear
I
lo 7

i
0 10 20 30
10-3

0 10
-Single Gear Excitation
-All

20
Gear Fkcitation

30
Frequency [Hertz] Frequency [Hertz]
Figure 1 7 Power Spectral Density of Harmonic Excitation Figure 1s:Comparison of Power Spectral Densities of Nose-
Response (3.6Hz) Gear-Only and All-Gear Excitation Responses

tion shows higher amplitudes than the response to the single

To assess the impact of the excitation of the main gears, com- gear excitation.
parative simulationswere performed. The results are presented Nevertheless. the relative impmvements of the all-gear-excita-
in the next subchapter. tion. semi-active gear against the all-gear-excitation. passive
gear are at least as good as for the nosegear-only excitation
5.5 All-Gear Excitation case (see [IO]).
This gave rise to the decision to use the single gear excitation
The general approach to the results obtained above was to for optimization and evaluation,reducing the computation time
excite the nose gear only. There are several reasons for this to levels compatible with existing hardware without risking to
assumption. First, it is a contribution to the fact that the center produce overly optimistic results.
line lights hit by the rolling aircraft would mostly be encoun-
tered by the M S landing
~ gear only. The main landing gear is
6. Summary
supposed to run on smooth surfaces. This argumentation is,
with some restrictions, also valid for the harmonic excitation, In the development of newer aircrafi a tendency can be seen
due to the wide aircraft track of 9.6 m. Second, the excitationof towards more flexible airframe structures and particularly
all three landing gears adds considerably to the model size. towards smaller fuselage bending stiffness. During ground
translating into higher numerical simulationtimc. The time dif- operations. this led to an increasing number of feedback prob-
ference to the single gear simulations are substantial. since the lems between elastic struch~reeigenmodes and landing gear
main gears represent a highly nonlinear subsystem. Therefore it dynamics, induced by runway unevenness.
was decided that the system performance would be evaluated The aim of this work was therefore to analyze the performance
using the single gear excitation case for the optimization N ~ S potential of a semi-active nose landing gear oleo to reduce
as well as for the subsequent evaluation simulations. fuselage resonance motion. This oleo is derived from an exist-
The followinginvestigationwas made to verify the justification ing oleo for the aircraft used for the analysis, an AIRBUS
of this assumption. It compares the A300 aimaft model with A300. It consists of the original gas spring and a modified oil
the conventional nose gear oleo with the model equipped with orifice capable of changing the damping coefficient at a high
the semi-active nose gear oleo. For the main gears conventional frequmy. A contmller was derived and optimized which cal-
oleos are implemented in both cases. The nose gear is subjected culates the necessary damping following the sky-hook damp-
to the center line light excitation as described in chapter 3. the ing approach. Here the absolute fuselage vertical acceleration
main gears are moved by the quasistochastic excitation is measured and processed by tbe controller.
described in the same chapter. The analysis was wormed using state-of-the-art simulation
The maximum peak values for both simulations,caused by tbe technology. A MultiMy Simulation (MBS) model was
center line light bumps, appear in both responses as nearly derived, including the highly nonlinear kinematics and force
equivalent. The difference between the excitation cases lies at laws of the landing gears and oleos and the airframe s t ~ c t u r e
the interval between two bumps. Here the single excited case elasticities. The controller was optimized using multi criteria
experiences almost no acceleration whereas the excitation of multi objective optimization software on three models with
the main gears for the all gear excitation case causes long excitations representing different runway surfaces and excita-
wavelength accelerations in the cockpit region. Thus. for low tion cases.
Frequencies higher PSD levels can be expected. The performance of the semi-active system was assessed by
Comparing the performance of the conventional gear with the time domain analysis with numerical integration as well as
semi-active gear (see figure 18). practically no difference is with frequency domain analysis using power spectral densities.
found for frequencies higher than 10 Hertz. In the range The behavior of the proposed system showed significant
between 0.1 and IO Hertz. the response to the all-gear excita- improvements for all cases under consideration and for the
5-12

whole frequency range of interest. The results highlight the [I31 HD

:. os: ANDECS ANt $isat: DEsign < Con-
multi objective parameter optimization approach for finding a trolled Systems (ANDECS 1.0). Technical Report,
well performing parameter set. Since an indisputable depen- DLR IB 515-W17.1994
dency of the system fuselage acceleration in the frequency [I41 Kortilm W., Lugner P.: Sysremdynamik und Regelung
domain on any parameter cannot be established it seems to be von Fahrzeugen. Springer Verlag Berlin, Heidelberg,
difficult to come up with a satisfying controller layout by man- New-York 1994
ually tuning ifs parameters.
[IS] H. Unbehauen: RegeIungstechnik I. Meweg Verlag,
Both optimization and performance evaluation simulations are
based on the highly nonlinear and complex MBS model. It Braunschweig, Wiesbaden 1989
lines out a representation of the real aircraft in many aspects [I61 H. Unbehauen: Regelungstechnik 11. Vieweg Verlag,
affecting the dynamical behavior. Nevertheless it has to be Braunschweig, Wiesbaden 1989
remembered that also this complex model uses assumptions [I71 W. E. Thompson: Measunments and Power Spectra
that may or may nor influence the dynamic behavior. of Runway Roughness at Airports in Countries of the
North Atlantic Treaty Organization. NACA Technical
Note 4303, WashingtonJuly 1958
7. References
[I81 Fletcher R.: Practical Methods of Optimization. Vol.
[I] S.F.N. Jenkins: Lunding gear design and development. 2. John Wiley and Sons. USA, 1981
Proceedingsof the Institution of Mechanicalhgi- [I91 Kraft D.: A Software Packagefor Sequential Qua-
neen, Vol203,1989 dratic P m g m i n g . DFVLR-FB 88-28, Oberpfaffen-
121 Donald W.S. Young: Aircraff landing gears: the post, hofen. Juli 1988
present, andfuture. Aircrafi Lunding Gear Systems. [U] G.Grlibel, H.-D. Joos. M. otter: The ANDECS
Society of Automotive Engineers, page 179-196,1986 Design Environmentfor C o m l Engineering. In IFAC
131W. Kortilm and W. Rulka and A. Eichberger R e n t World Congress, Sidney, 18-23 July 1993
Enhancementsof SIMPACK and Vehicle Applica- [21] H.-D. Joos: lnfonnafiomtechnische Behandlung des
tions. European Mechanics Colloquium, EURO- mchm'eligen opfimierungsgesfiltztenregelungstech-
MECH 320, Prague 1994 nischen Enrwurfs, h.D. thesis, Universilllt Stuttgart,
[4] P. T. Somm,H. H. Straub and J. R. Kilner: Adaptive Prof. R. RUhle, 1992
landing gear for improved taxi performance. Boeing [E]
H. Wentscher and W. Kortilm: Multibody model-
Aerospace CompanyJ977, APFDL-TR-77-119 based multi-objective parameter optimization of air-
[SI John R. McGehee and D. L. Moms: Active control craft landing gears. Pmceedingsof IUTAM-Sympo-
landing gear for ground load alleviation. Active Con- sium of Optimization of Mechanical Systems,
trol Systems - Review, Evaluation and Projections Stuttgart 1995
(AGARD CP 3&1, NASA Langley R. C. ,1981 [231 W. Kortilm and R. S. Sham (&is): Multibody Com-
~~

. -
[6] W.E. Howell. J.R. Mffiehee, R.H. Daugherty, W.A. puter codes in vehicle system ~ p a m i c sVehicle
.
Vogler: F-106 airplane active control landing gear System Dynamics, Supplementto Vol. 22. Swets &
drop test performance. Landing Gear Design Loads Zeitlinger, 1993
(CP W ) ,AGARD.1990
171R. Freymann: An active control landing gear for the
alleviation of aircrafttaxi ground loads. Zeitschrift fiir
Flugwissenschaft und Weltraumforschung, 1987
[8] D. Kamopp: Active Damping in Road Vehicle Suspen-
sion Systems. Vehicle System Dynamics, 12(6),Swets
& Zeithgex, Lisse, 1983
[9] m n e Catt, David Cowling and Alan Shepherd:
Active landing gear control for improved ride quality
during ground roll. Smart Stluctures for Aircraft and
Spacecraft (AGARD CP 531), Stirling Dynamics Ltd.
Bristol, 1993
[IO]H. Wentscher: Design and Analysis of Semi- Active
Landing Gears for Transport Aircraft. h.D. Thesis,
TU MUnchen, 1995
[ll]R.V. Dukkipati, S.S. Vallurupalli,M.O.M. Osman:
-
Adaptive Control of Active Suspension A State of
the Put Renew. The Archives of Transport. Vol. IV.
1992
[I21 International Organization for Standardization: Guide
for the Evaluation of Human Exposure to Whole Body
Vibration. IS0 2631-1974(E). July 1974
 6- 1

Dynamic Behaviour of Motorbikes

F. Bohm and H. P. Willumeit

Department of Traffic Engineering and Applied Mechanics

Technical University Berlin, Institut fur Fahrweugtechnik
Sekr. TIB 13, Gustav-Meyer-Allee 25
13355 Berlin, Germany

Abstract
Dynamic behaviour of motorcycles front wheel are not different to similar wheel suspension
motion of airplanes and road vehicles. In every case self excited oscillations caused by
the tire occur. Only the different design of wheel suspension produces some different
coupling. The deviation of a moving wheel on a horizontal plane from straight rolling is
caused by unstable lateral contact forces. The kinematical generation of such forces can
be imaginated looking on the elastic tire. A first order time-depending behaviour together
with inertia mass and damping of the wheel using the HURWITZ stability criterion gives
important insight into the main parameters of the problem.
Nevertheless the nonlinearities of the system overshadow this clear insight. Cornering force
and cornering moment are nonlinear functions of cornering angle, so the use of the well
known Magic Formula together with the black box method of control theory is a logical
extension. The problem of instability arises again when thh wheel is excited from wavy
surface or similar from mass unbalances of the tire. A direct time depending of the lateral
tire forces can be used to show parametric instability of the system (MATHIEU equation).
The rolling of the wheel on a non-smooth surface needs then to introduce methods of
discrete system approach. The paper presents an overview of the problem of shimmy.

1 Introduction
The pneumatic tire as a rolling suspension is intended, on the one hand, t o ensure that track
unevennesses are traversed with as little vibration as possible, and, on the other, t o guarantee
the driving direction a n d stability of the vehicle. In sharp bends, the pneumatic tire should
also withstand the centrifugal forces, and it should generate the appropriate longitudinal
forces during braking and acceleration. These manifold requirements, however, can only be
met by means of a constructive compromise, wherein a n unavoidable coupling arises, by
virtue of the elasticity of the tire in all three dimensions, between the longitudinal, lateral
a n d vertical forces a n d between their associated dynamic deformations. These couplings lead
t o non-linearities, which complicate mathematical treatment.
A first attempt t o solve the problem starts out from the premise that the track is even a n d
the tire ideally round, a n d also assumes, with regard t o the contact time of a tire particle with
the ground, that the associated slippage between tire and ground occurs in linear fashion,
from the entry point of the contact surface t o the exit point thereof. If one additionally
takes into consideration an average deformation g of the contact surface, due t o the effective
lateral force S, then a linear elastic deformation equation can in principle be written down
for the lateral deformation as a function of lateral force:

s = cs g . (1)

Paper presented at the 81st Meeting of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibration-Free Landing Gear, held in Ban8 Canada from 4-5 October 1995, and published in R-800.
6-2

The associated lateral spring coefficient cs can be determined experimentally. The linearly
increasing deformation from the entry point to the exit point, by contrast, is assigned to
the travelling surface (shear deformation) and is superposed onto the deformation g. This
enables one to write down a kinematic equation for the slip angle of the tire to which the
lateral force S is assigned, of the form

This effective slip angle 6 , j j , which is a function of the elastic lateral deformation of the tire
carcase, is equal to the stationary slip angle for the case S = constant. The function S is non-
linear, since it describes the partial sliding of the tire, and even progressively decreases overall
once it has passed beyond the sticking state. If one only takes into consideration small slip
angles in the region of zero (driving straight ahead, equation (2)), then the increase Po c' in
the cornering force function at the zero point yields the following linear differential equation
of first order, whose eigenvalue is negative and therefore exhibits damping behaviour. Any
disturbance of the stationary rolling state is therefore damped along the direction of rolling,
wherein the damping distance is approximately equal to the tire radius. Since, however, the
stationary slip angle characteristic curve is largely a function of the length of the contact
area, which in turn is a function of the possibly dynamic wheel load, there is automatically
a coupling between the horizontal and vertical dynamics of the tire.

2 Analytical investigation of the stability of shimmy -

. -
Let us now consider a wheel elastically mounted on a large mass in accordance with the
examples of figure 1, and investigate the stability of small vibrations of a wheel that can be
rotated around a vertical axis. In actual technical realisations the pivot point of the steering '

axle is always in front of the contact patch, see figure 2. Let the wheel be elastic mounted to
the straight-ahead position by means of an elastic attachment (torsion spring coefficient C L ) .
There is a principal moment of inertia I about the vertical axis. For the pneumatic tracking
we designate n p = !$. In the case of damping proportional to velocity (steering damping),
a coefficient k~ is used. The characteristic curves for the tire are displayed in figure 3.
Analytically, the use of n p has the advantage of a simpler mathematical description. The
equations of motion are as follows:

1. Angular momentum about the steering axle:

( I + m a') + = -cL cp - kL - S (a +np). (3)

2. Lateral guiding force S for small slip angles

S = S(6ejj) = Po C' 6 e j j = cS. (4)

3. The effective slip angle:

(5)
 6-3

Figure 1: Ezomples of wheel suspension: a) motorcycle, 4) motor car and c) aeroplane (after [Volk93])

S n

200

Figure 2: kinetic and dynamic quantities of the Figure 3 Lateral guiding force and pneumafic
steered wheel irackinglength

After linearisation (np M no - see figure 3), it follows that:

ma') ++kL++cer,p+(a+no) Pod = 0,

Pod (.++) aci, -csii = 0. (7)

With - ex *,that leads to the characteristic determinant solutions ( I . = I + m a2):

= 0. (8)
The expansion of this gives:

2.1 The unconstrained wheel

If, for example, one sets CL = k~ = 0, the characteristic equation obtained is:

In the case of a wheel free to rotate around the steering axle CL = 0, k~ = 0, the ROUTH-
HURWITZstability conditions for the bow wheel of an aeroplane or for the steering of a
motorcycle give
a cs -Po c' > 0. (11)

holds:
pi ""'b
/.
d 5 6 i- K

f(Acos(v t)) cos(v t ) dt = ci(A) ms(v t).

x
In both of the partial domains A < yo and A 1 yo, or for a = with a < 1 and a 1 1,
the following equivalent torsion spring coefficients shall be
a<1: c;=o

a 2 1: c; = CL (1 - k(a))
The shape of the function k ( a ) is given in figure 5. From the polynomial of third degree for
the characteristic value X
 6-5

for c: # 0 but kL = 0, we obtain the HURWITZdeterminant

The determinant vanishes for cr = y,

the relaxation length.
For kL # 0 (and c; # 0), the HURWITZdeterminant is

from which it can be gathered that there is an additional term proportional to the factor c;.
From these formulae it can be seen that
1. without steering damping there can be no stabilisation and
2. if there is steering damping, sufficient stabilisation can also be provided with a dimi-
nished steering stiffness c:.
3. Since in general the geometric tracking length is significantly smaller than the relaxa-
tion distance, the additional terms in equation (9) must compensate for the negative
effects of the stability condition (equation (15)).

Thus steering tolerance is frequently not immediately recognised, since the graph of the
function k(cr) only slightly diminishes the effect of the steering damping. Thus a maintenance
problem arises from the fact that, even with a very small amount of steering tolerance,
vibrations are amplified.

2.3 Taking friction into consideration

In order to avoid the effect of steering tolerance, therefore, friction is introduced into the
steering by means of constructive measures. If one considers the effect of friction (figure 6),

Figure 6: Dependency of
steering restisiance on fric-
tion

this can be represented by an eqivalent linear term of the form

4 MR
';= .IrA(-J
6-6

where MR represents the frictional moment in the steering. For the case where CL = 0, but
ki # 0, the characteristic polynomial is now:

x3 + (-Pov + f )x2 +
cs
c
[(U + no) -
U cs v cs k;
Ia
+ PO Ia
+ + no) I
(U ,- 0.
v cs (18)
The HURWITZdeterminant is given by:

Here one can also see that irrespective of whether A,,, i.e. the HURWITZdeterminant for
the case CL = k~ = 0, is positive or not, a suitably dimensioned frictional moment or ki
compels st ability.
This simple analytic investigation makes it clear that every rotating but elastically mounted
wheel is capable of exhibiting self-excited rotational vibrations about the axis of steering.
Even if wheel load variations are not taken into consideration, and thus an even ground is
assumed, there is a tendency towards so-called wobble vibrations in case of a motorcycle. The
effect of the non-linearity of the system on the amplitudes of vibration alone is also exhibited,
wherein more exhaustive efforts with the aid of harmonic balances also take into consideration
the non-linearity of the cornering curve. Unfortunately this effect is in principle non-linear, I

also with respect to the vertical dynamics of the wheel, leading to states of motion at several
frequencies. The method of harmonic balances is thereby rendered useless for dealing with
the problem. Notably the tire width and curvature of the contact area have a stabilising 4
effect [Boe77].

3 Steer-Shimmy of Motorcycles - i
Instabilities of the steering system are known for several types of vehicles, e. g. the tail- or I
mainwheel of airplanes and the (steer-)shimmy of automobiles. This instability also accurs
at single-track vehicles as bicycles, scooter, and motorcycle. It describes the resonance
movements of the steering-system around the steering axis.
Several authors examined this phenomenon of motorcycles (e. g. [Pac66], [Sha71], [Sha77],
and [Koch80]).
The modelling of the motorcycle by means of a linear differential equation system of 20.
order provides the main (eigen) figures of the driver - single-track vehicle system. [Sha71]
indicated the terms for three eigen-values:
0 capsize mode, as vibration-free movement of tilting (tipping) of the whole motorcycle
to one side, which is stable in the lower velocity range, and constantly instable in the
medium and higher velocity range.
0 wobble mode, which is called shimmy at other vehicles, and is steer motion which
occurs at motorcycles as a frequency between 7 and 10 Hz almost independent from
the vehicle velocity, and is good damped in the lower and medium velocity range.
0 weave mode, as vibration of the whole motorcycle around the yaw axis with an incre-
asing frequency at lower driving speed and in the medium and higher velocity range
movement tending towards a limit value between 2 and 3 Hz.
 6-7

The linear theory, however, does in comparison with measurements on the motorcycle not
come up with real-world condition results for the wobble mode, since the whole calculation
model for the medium velocity range calculates a well damped vibration then, the mea-
surements show that especially here just very low, respectively no damping exists. Other
mechanical couplings are, therefore, effective as they are enclosed in the linear vehicle model.
From the observation of a real vehicle showing that shimmy movements are very much cou-
pled with the balancing condition of the wheel, [KochSO] developed a model, which contains
the steering system of the motorcycle alone, without the rear frame. As a fundamental addi-
tion compared to previous modelling, he encludes the periodic wheel load oscillation, which
comprise a force component around the steering axis, supplying the energy for exciting the
wobble mode. These wheel load oscillations occur through pitching and vertical motions
of the vehicle and front wheel which are connected with the radial deformation of the tire.
These oscillations are produced by road irregularities, as well as by the periodic forces of a
poor balanced wheel.
[Koch80], herewith, got a mechanical backup system, which encludes the radial deformation
of the tire in the wobble mode. Result of this modelling is a non-linear differential equation
system, which can be transformed to a MATHIEU differential equation, so that the kinetic
ranges of instability of the wobble mode can be demonstrated.

1. Solution of the vertical motion

Fig. 7 may be used as model. The following differential equation gives the force-balance

Figure 7: model of the vertical motion

in the vertical direction:

(mo:Mass of wheel, CR: Tire radial stiffness, y: vertical motion of wheel axis, z:
vertical motion of road surface). With y = yo cos(fl t) one gets:
6-8

The reaction force coming from the road excitation results as radial force to

FR= CR (Y - 2)
at the center of the tire contact.
2. Solution of the pitching motion
A road irregularity at the front wheel produces a pitching motion of the vehicle. Fig.
8 may be used as model. The following differential equation for the center of gravity

~~

Figure 8: Model of ihe pitching moiion

of vehicle results from the torque balance

4
(q: pitch angle, cRV/R: front/rear wheel stiffness, z: vertical motion of road surface,
u ~Eigenfrequency
: of pitch motion) so finally the solution is

q = qo cos(n t ) = Qv aG 20 ms(R t ) .
0, (U; - 521)
The following superposed movement results from the pitching and vertical motion of
the whole vehicle for the front wheel

(33)

out of which the radial force can be calculated:

(34)

3. Equation of motion around the steering axis

This torque balance around the steering axis gives the following differential equation,
the model is shown in fig. 9
 69

~ ~~

Figure 9 Model for ihe mofion around !he steering azis

.- ..
*..+ n Fsv + Mzv + FR n
C O S ~ sinX = 0, (35)

8 = Ivz + mv 1, + IGz + mo 2.
2
(36)
Linearization for small steering angles and with

results in
l+2w~D;\+w: (39)

If the radial force from the pitching and vertical motion are introduced into the quations of
motion around the steering axis the following nonlinear equation results out of it:

l + 2 w ~D ;\+U: X (1 + 2 p cx~(S?t ) )= 0, (41)

the MATHIEUdifferential equation.
The interesting part of this equation is that unlimited expanding solutions for particular
~
values of its coefficients are given These values produce ranges of instability.
6-10

Although the solution of the MATHIEU differential equation demonstrates especially at dou-
ble excitation frequency its main range of instability, the chape of the non-linear function,
however, dictates what happens.
The shape of this function is shown in fig. 10 using the data of a motorcycle as function of
the driving speed U, respectively wheel frequency R. The shape of the front axle amplitude

Irl 4

Figure 1 0 Kinetic ranges of instability ofthe sfeenng sysfem vs vehicle speed

(afler [KocheO])

as function of the driving speed, respectively wheel frequency is shown in fig. 11 showing two
resonances of the vertical and pitching eigenfrequency. Since the resonance of the pitching-

I
Figure 11: Amplitude rafe of fhe vertical fmnt-arle track (afler [Koch80])

eigenfrequency (wn = 7,78 Hz) approximately coincides with the medium range of instability,
also instable wobble motion is to be expected within this velocity range.
The measured frequency and velocity during the wobbling of this vehicle agrees with theo-
retically gained values of the nonlinear stability analysis. These results show beautifully
that the wobble mode can also be determined as a characteristic value with the help of the
linear theory, though it does not agree with measurements at the motorcycle, with respect
to either frequency or damping. Only by including in the modelling the vertical dynamics
 6-1 I

of the wheel together with the lateral dynamic characteristics, which leads to a non-linear
differential equation, results in regions of instability that can be confirmed by measurements.
A first- or higher-order theory of tires for the lateral characteristics and a radial tire stiffness
dependent on the tire speed may further refine the results.
Even if [Koch801only used the balanced masses in its model as an excitation, the broad spec-
trum of road unevennesses also provides excitations. This is also confirmed by measurements
wherein greater instability arises with an unbalanced front wheel.
Also G. Nybakken is engaged in work in his dissertation [Nyb73]on investigating the stability
of rheolinear aeroplane landing gear.

4 Tire Magic Formulas

Simulation of systems in general but also of tires can be done for two goals:
1. Description of the physical happening in a system by using physical laws:
2. Imitating the physical behaviour - in the meaning of a black-box view - by appro-
ximating the behaviour.
Engineers use both methods for simulation, whereby the imitation of the physical behaviour
must be done by system identification on the basis of estimated and anticipated describing
rules; e.g. by theoretical control assessments.
Both methods deliver also different statements. Using physical laws, precise predictions
about non existing systems, and their influence by physical parameters are possible. On the
- other hand the computation effort is generally very high. The advantage, and very often, also
- the goal of the second method is the real time ability for a complex dynamic system combined
with the disadvantage not to entirely comprehend the influences of physical parameters of
that system.
The method to imitate the physical behaviour by its approximation on the basis of behaviour
equivalent formulas is called in the last time magic formula.
On one hand in the area of vehicle development but also for driving simulators a real time
ability or at least a very quick statement about the system behaviour of the total vehicle
is necessary without for instance to accomplish a physically exact description of a shock
absorber.
In the following two of these approximating formulas are described, one could be used for
calculation of the tires static and the other for the calculation of the tires dynamic behaviour.

4.1 Static behaviour of tires

Since the static behaviour with respeckt to side force, self aligning torque, camberforce etc.
for braking and tracking forces of all tires are similar the effort to build up a physical model1
and the effort to estimate all necessary physical parameters is enormous, so the idea of using
only some mathematical functions which fits the measured results by changing only some of
function parameters came up.
[Bak87] and (Bak891 presented a paper in which they proposed a formula with coefficients
which describe some of the typifying quantities of a tire, such as slip stiffness at zero slip and
force and torque values. The formula is capable of describing the characteristic of side force,
 ~

6-12

brake force and self aligning torque with great accurac. That mathematical representation
is limited to steady-state conditions.
Figure 12 shows the basic form of each characteristic G :he tire. Bakker et al. took over the

I I I
Figure 12: Steady-state tire c h a r a c i e n x s (after [Bak87])
4
idea by Ruf [Ruf83] to approximate these tire behaviour by special angular functions.
The demands for that formula was expressed - -
0 to be able to describe all steady-state tire characzistics,
0 being easily obtainable from measured dates, - \

,
0 being physically meaningful; its parameters shoii characterize in some way the ty- ,I
pifying quantities of the tire,

0 being compact and easy to use,

0 contributing to a better understanding of tire beAviour, and

0 being accurate.

These demands were found for all tire characteristics i: :he expression

y(x) = D sin {C arctan [B x - E ( I L - arctan(B z))]} , (42)

Y ( X ) = y(.)+S,, (43)
x = X+Sh (44) I

with Y(X)standing for either side force, self aligning tcque or brake force and X denoting
slip angle ( C Y ) or longitudinal slip (6). I

D in equation (42) is the peak value and the produc- D C B equals the slip stiffness at
zero slip. The coefficient C governs the shape of the crve and the coefficient E makes it I
 6-13

I
I possible to accomplish a local extra stretch or compression while stiffness and peak value
remain uneffected.
Due to ply steer, conicity, rolling resistance and camber, the characteristics will be shifted in
horizontal and/or vertical directions. These shifts are represented by Sh and S, respectively.
By these approximating functions Bakker et al. got in parts excellent fitting results of
measured and calculated tire behaviour as shown in Fig. 13 and 14.

B R A K E FORCE -Fa
SIDEFORCE - F t
(NI

8,000

6,000

c.000 UCASUREMEHTS:
= (7, = ODCO.
b 0, - 2 DEO.
2.000

0
0 10 20 30 10 60 80 10 80 90 100
-LONGITUDINAL
SLIP - K 111

c Figure 13: Brake and side force us slip (after [Bak87])

SELF ALlONlNG
TORQUE -Yz
IWm)
2oofMEASUREMENTS

.---

-2001

Figure 14: self aligning torque us break force (after [Bak87])

4.2 Measured Transfer Properties of the Tire at time varying

side slip angle and its approximation by control theory trans-
formations.
~
[Wi11,3], [NastSl] and [Will951reported about the dynamic behaviour of tires due to dynamic
changes of side slip angle and wheel load.
6-14 I

For simulation respectively pre-calculation of the dynamic driving behaviour of single and
double track vehicles, the knowledge of the dynamic behaviour of the tire is necessary,
without time consuming sub-routines for a tire model. [Schu87] reports in detail on the
latter.
Easier and without time consuming appliance the transfer properties on a test stand with
time varying exitation signals can be evaluated. This leads to no explanation, and realization
on the physical processes inside the tire respectively in the contact area though, but enables
the vehicle engineer to easily integrate the tire dynamic into a model of vehicle dynamic or
driving dynamic. When measuring the tire with time varying exitation, the simultaneously
occurring inertia forces and gyroscopic torques, which do not occur at static measurements,
have to be corrected carefully. On this procedure, and the following description of the
transfer function by means of control equations reported [Wi1187], and [NastSl] applied to 4
different tires, two motorcycle, and two passenger car tires. The time varying side slip angle
a(t)was described with a frequency limited white noise up to approximately 17 Hz with a
variance of 0,21" (see fig. 15).

AMPLITUDE (degree'degree)
1.E41

1
1 10

I FREQUENCY (Hz)
~~~~
I
Figure 15: Power spectral density of side slip angle (after [NastSl])

If one applies the corrected dynamic self aligning torque, and the corrected dynamic side
force to the stationary values, that are these values, which occur at side slip angles (f = 0 Hz)
constant with time, so one will come to non-dimensional self aligning torques and side forces.
By computing the power spectral densities, one finally gets the transfer function Malaof the
self aligning torque, and Fa/aof the side force. The fig. 16, and 17 show this exemplarily
as Bodeplot.
 6-15

LMPUTUDE(lldegree)

ia

U. I
1 10 20
FREQUENCY (Hz)

45 -

v3

-45- v*
I - -
-90- vi = 5 m/s
. v2 = 20 m/s
-135 -
- -
. v3 =50m/s
-1 80
1 10 20
FREQUENCY(Hz)

Figure 16: Measured transfer function of self aligning torque relative t o side slip angle
(after [NastSl])
6-16

LMPLITUDE(l/degree)
10

I v,J

1
v*

0.1 1 FREQUENCY(Hz) 10 20

45

0
"3

-45 v*
vl
-90
1 10 20
FREQUENCY (Hz)

Figure 17: Measured transfer function of side slip force relative t o side slip angle (after
[Nast 911)
A PD2-behaviour (proportional plus double derivative action controller) is assumed as ap-
proximation equation for the transfer function, which reproduces both amplitude channels
appropriately. Merely the phase has to be adapted by another approximation term - a
dead time term. That means that the dynamic behaviour of self aligning torque, and side
force has to be adapted by higher terms as 2nd order.
The approximation equation in the time domain for the self aligning torque are

and for the side force

@a ( t - Ts) da (t - Ts)
Ms dt2 +Ds dt
+
cs CY (t - Ts) = F,(t)
Fig. 18 to 20 show each time the mass factors MR and M s , damping factors D R and Ds,
stiffness factors CR and Cs for the 4 measured tires depending on the wheel load, and wheel
speed. The index R refers to the self aligning torque, the index S to the side force. The
types were:
 6-17

etillness factor CR [Nm/deg]

I

QO

4#

Q4

QZ

U 2 2s 3

Whccl load RN]

stirneas factor Cs [N/deg] Damping factor DR [Nms/deg]

.loo
(100
0
1-

dQ0

Y-

I.J 0 S IO I5 I 30 U 40 4J W

Whcel load [kN] vclocity [m/s]

lyre 1; A lyre 2; X tyre 3; M tyre 4

Figure 18: Stiflness factors CR and Cs as function of wheel load; damping factors DR and Ds as func-
tion of wheel speed: 1. Bridgestone 120/80 V 16 (motorcycle); 2. Bridgestone 150/80 V 16 (motorcycle);
3. Conti Contact 155 R 13 (automobile); 4. Conti Contact 175/70 R 13 (automobile) (after [NastSl])
6-18

Massfactor MR [Nms2/degr.]
0,015

0,Ol

0,005

0
0 5 10 15 20 25 30 35 4 45 50
Velocity [m/s]

Velocity [mls]

tyre 1; A tyre 2; X tyre 3; I tyre 4

Figure 19: Mass factor MR and stiffness factor CR as ixction of wheel

speed (after [NastSl])
 0.02

0.01

0
0 5 10 1 5 2 0 2 5 3 4 3 5 4 0 4 5 5 0

verocrrvtw

H tyre 1; A tyre 2; X tyre 3; H tyre 4

Figure 20: Dead time as function of wheel speed
(after [NastSl])

As result there can be concluded:

0 the stiffness factor increases on increasing wheel load for the side force as well as for
the self aligning torque.

0 the damping factors DR and Ds increase on increasing speed.

0 the dead time of the side force reduces on increasing speed.

0 CR is constant for motorcycle tires (cross bias type) over the speed, for passenger car
tires (radial tires) the stiffness factor and CR increases degressively on increasing speed.

0 MR increases progressively over the speed.

0 an influence of the inflation pressure of the tire on the factors could not be measured.

On further measurements at time varying wheel load F. Sun [Sun901 also reports in his
Ph.D. thesis, and reached similar results, in which he additionally determined the dynamic
transfer characteristics for time varying wheel load.
Since the whole transfer behaviour is still very much depending on the wheel velocity, except
the stiffness factor CR for the self aligning torque of the motorcycle tire (cross bias type),
one can assume that non linear effects, i.e. because of gyroscopic effects can not fully be
recorded by means of the linear approximation formulation.
Assuming black-box-formulas are available as break-, sideforces and torques are described in
their dynamic behaviour a straight forward integration procedure is possible which will be
shown here.
6-20

5 Numerical method of investigating wobble

For every non-linear function S(&fr,P ) , an implicit differential equation of first order can
be given-for the non-stationary process:

A stabilisation of the numerical or piecewise linear integration in the case > 0 can be
achieved through an extension of the equation by a damping term

and possibly a mass compensation term. These, however, does not render impossible the
correct assessment of short-wave surface unevennesses, with regard to their effect on P ( t )
and on S(t). As a result, the high-frequency rolling process cannot be treated further by
means of analytical mechanics.
Since further investigations of an analytic kind are not possible, one is compelled to resort to
numerical methods. In these, the contact surface is modelled as discrete, and the sequence
of contact points is investigated in discrete calculating steps. A unilateral contact condition
enables the numerical representation of partial lifting and traversing of track unevennesses.
In the process, the tire structure is resolved into many degrees of freedom, taking into
consideration the tire mass and damping. Such a structure is geometrically non-linear in
principle and has a relatively high frequency. It is guided by means of a rigid body system
of low frequency, consisting of a rim and shock absorber. If, as is customary for motorcycles,
one introduces a head lug angle A, then one obtains the following system of equations of
motion for the steering angle p and the spring deformation s of the wheel axle (see figure 1).

JL J + kL a + CL p = My cos X - M, sin X + a F, (49)

mFd+kFi+CFS = Fy C O S X - F , sinX (50)
y, = Yf+ (s - ),Z COSX (51)

In addition, there are also the lateral motion and longitudinal motion of the motor vehicle
mass itself. The contact forces generated while traversing the ground unevennesses follow
from the equations of motion for the tire structure, which are not given here [Boe93]. The
sample calculation given in figure 21 shows the dynamic effect of wheel load variations on
a stable steering system that is nevertheless capable of damped wobble. The non-linearities
taken into account in the numerical integration are the deformations of finite magnitude, the
unilateral contact, i.e. local lifting and settling of the running surface on rough obstacles,
and the effects of traction and sliding friction. The model does not take into account the
effects of tire width or of cross-sectional bending. Wobble damping, however, is markedly
influenced by enlarging the tire width and increasing the length of the contact surfaces. The
resultant moment of pivoting has the effect of diminishing wobble. Thus, not surprisingly,
for small contact surfaces, which arise,,under low wheel loads, have an especially critical effect
on wobble. A surface-oriented theory of non-stationary rolling contact has been analytically
developed for perfect roadholding, taking into account the bending of the trajectories of the
 6-21

Example: Contact with plane including a rectangular (1x2 cm)

obstacle distance 120 cm

II
1
i

f,<
R

. _ y
= long., v = vertical

I w-'
-_
- -
Breaker Tension Force
1
r
5

2x

Figure 21: Sample calculation for the steering system of the froni wheel of a moiorcycle

contact particles of a flattened tire. Wobble vibrations produced on the testing stand show
that the phase angle between the amplitude of wobble and the restoring torque of the tire can
certainly pass beyond go", which cannot be represented by tire behaviour of the first order.
The tendency of the system to auto-excitation is thereby markedly reduced when there is a
sufficient area of contact between the tire and the ground. If the ground is uneven, however,
I so that the contact surface is at times considerably reduced, one speaks of a transition to
I hard auto-excitation, since the stabilisation is now only produced by the kinematic terms
of f i s t order. At high speeds, therefore, the reduction of load on the front wheel due to
I the air resistance of the vehicle necessarily leads to critical states of wobble, while the axial
load is sufficiently large in the normal driving state. The non-linear coupling between the
horizontal and vertical dynamics of the wheel also calls for numerical treatment here, since
an analytic procedure only permits calculation in the case of perfect roadholding. Such a
6-22

numerical procedure, however, is very costly in term of computing time, and has only been
carried out until now with coarse substitute models with few degrees of freedom. If, however,
one starts out with an analytically based strategy, then the destabilisation effect can also
be demonstrated here by means of numerical examples for integrating the rolling action on
rough ground.

6 Conclusion
Starting from that given overview furture theoretical investigation should include. It is
possible to stabilise the system even in case of small contact area using the main influencing
parameters found by linear theory. If the parameters are near stability border the nonlinear
effects may suddenly disturb the rolling wheel and self-excitation occurs due to vanishing of
vertical spin effects in contact area. In any case also for nonlinear computation the linear
theory can be used as a guideline. All possible non-smooth surfaces could be taken into
account using discret system modeling for the tire.

Lit eratur
[Bak87] E. Bakker, L. Nyborg, and H. B. Pacejka:
Tyre Modelling for use in vehicle dynamics studies.
SAEP 87 04 21, Society of Automotive Studies, Inc., 1987.
[Bak89] E. Bakker, H. B. Pacejka: ..
A new tire model wih an application in vehicle dynamics studies.
SAEP 89 00 87, Society of Automotive Studies, Inc., 1989.
[Boe77] F. Biihm:
Computing and measurements of the handling qualities of the belted Tyre.
Proc. 5th VSD 2nd INTAM Symposium, Vienna, pp 85-103, Swets and Zeitlin-
gen, Amsterdam and Lisse, 1978.
[Boe93] F. Biihm:
Reifenmodell f i r hochfrequente Rollvorgfmge auf kurzwelligen Fahrbahnen.
(Tire model for high-frequency rolling processes on short-wave tracks.)
(Bericht Sonderforschungsbereich 181, Teilprojekt B1)
VDI-Berichte Nr. 1088, pp 65-81, VDI-Verlag, Diisseldorf/Germany 1991
[KochSO] J. Koch
Experimentelle und analytische Untersuchungen des Motorrad-Fahrer-Systems
(Experimental and analytical assessment of the Motorcycle-Driver-System;PhD-
Thesis 1978).
Fortschr.-Ber. VDI-Z., Reihe 12, Nr. 40, 1980.
 6-23

[Nastg1) R. Nast, Chr. Teubert and H.-P. Willumeit:

Messungen der Ubertragungseigenschaftenvon Luftreifen bei zeitlich verkderten
Schriiglaufwinkeln und anschlieflende Nachbildung dieser Gr6flen durch Approxi-
mationsgleichungen
(Measurement of transfer properties of pneumatic tires under time varying side
slip angle and followed by modelling using approximation equations.)
VDI-Berichte Nr. 916, pp 321-344, VDI-Verlag, Dusseldorf/Germany, 1991.

[NYb731 G. Nybakken:
Investigation of Tire Parameter Variations in Wheel Shimmy.
PhD-Thesis, The University of Michigan, 1973.
[Pac66] H. B. Pacejka:
The wheel-Shimmy phenomenon.
PhD-Thesis, Technische Hogeschool, Delft, 1966
[Ruf831 E. Ruf
Theoretische Untersuchungen des Federungsverhaltens von Vierrad-StraBenfahr-
zeugen.
(Theoretical investigations into the suspension behaviour of four-wheel road ve-
hicles. )
Fortschr.-Ber. VDI-Z., Reihe 12, Nr. 44, S. 104-105, Dusseldorf, 1983.
[Schu87] D. H. Schulze:
Instationare Modelle des Luftreifens als Bindungselement in Mehrkorpersystemen
fur fahrdynamische Untersuchungen.
(Non-stationary models of the pneumatic tire as a linkage element in many-body
systems for investigations into the dynamics of driving.)
VDI-Fortschritt-Berichte, Reihe 12, Nr. 88, VDI-Verlag, Dusseldorf, 1987.
[Sha71 R. S. Sharp:
The Stability and Control of Motorcycles.
J. Mech. Engin. Sci., 13 No. 5, pp 316 - 329, 1971.
[Sha771 R. S. Sharp and C. J. Jones:
The straight running stability of single track vehicles.
Proc. 5th VSD 2nd INTAM Symposium, Vienna, pp 334-342, Swets and Zeitlin-
gen, Amsterdam and Lisse, 1978.
[Sun901 F. Sun:
Dynamic Properties of Motorcycle Tires.
Dissertation, Beijing Institute of Technology, 1990.
[Valk93] R. van der Valk, H. B. Pacejka:
An analysis of a civil aircraft main gear shimmy failure.
Vehicle System Dynamics, 22, Swets and Zeitlingen, BV Lisse, Netherlands, 1993.
6-24

[Will871 H.-P. Willumeit und Ch. Teubert:

Dynamische Eigenschaften von Motorradreifen bei zeitlich veriinderlichem
Schriiglaufwinkel.
(Dynamic properties of motorcycle tires at slip angles varying with time.)
VDI-Berichte Nr. 657, pp 127-136, VDI-Verlag, Dusseldorf, 1987.
[Will891 H.-P. Willumeit and F. Sun:
Dynamische Eigenschaften von Motorradreifen bei zeitlich veranderlicher Radlast
(Dynamic properties of motorcycle tires under time varying, wheel-load).
VDI Berichte Nr. 779, pp 389401. VDI-Verlag, Diisseldorf/Germany, 1989.
(Will951 H.-P. Willumeit and F. BGhm:
Wheel Vibrations and Transient Tire Forces.
Vehicle System Dynamics, 24, Swets and Zeitlingen, BV Lisse, Netherlands, 1995.

1
 7-1

LANDING GEAR SHIMMY - DE HAVILLANDS EXPERIENCE

John Glaser and George Hrycko

Structural Dynamics Group
de Havilland Inc.
123 Garratt Blvd.
Downsview. Ontario, M3K IY5
Canada
1. SUMMARY 2 tire lateral deflection at aft end of contact patch
Landing gear shimmy is an unacceptable situation which, at the A tire lateral deflection (elastic)
very least, leads to increased maintenance costs and at the worst AM0 incremental roll moment
could result in catastrophic failure of the gear andor the attach- AMv incremental yaw moment
ing structure. Two landing gear shimmy problems experienced Ep pneumatic trail parallel to ground
by de Havilland are presented. one in service and the other pre- & gear tilt angle
dicted during design. While these problems occurred over IO h~ yawed-rolling relaxation length
years ago, it is the authors contention that little progress has p friction coefficient
been made in this design discipline in the intervening years. It is 5 critical damping ratio
recommended that advisory material is needed in this important 6, structural damping ratio
but often neglected subject. r, (Moreland) tire time constant
r t equals T, at 100 kts (Reference 3)
@ roll displacement
2. LIST OF SYMBOLS
y yaw displacement
tire lateral damping coefficient
tire torsional damping coefficient
y, tire elastic twist deflection in yaw
CO circular frequency
tire cornering power
damping matrix
tire cornering force 3. INTRODUCTION
tire side force Landing gears that shimmy are unacceptable, causing consider-
, - - side force able discomfort and annoyance to passengers and crew. In fact, a
landing gear vertical force (static) severe occurrence of shimmy can damage the landing gear and
vertical force on left tire its attaching structure, resulting in significant repair costs and air-
vertical force on right tire plane down time. The prevention of shimmy on an airplane with
gyroscopic matrix a shimmy-prone gear increases maintenance costs and may bur-
half footprint length den the pilot with self-established shimmy avoidance procedures.
inertia matrix Some assurance is therefore needed that landing gear designs
polar moment of inertia of wheel assembly will be free from shimmy under all operating conditions taking
stiffness matrix into account normal wear and tear experienced in service.
tire lateral spring constant
tire torsional spring constant Two examples of landing gear shimmy are presented in this
tire vertical stiffness report; the first involving the DASH 7 nose landing gear and the
structural torsion stiffness second the DASH 8 main landing gear. The DASH 7 airplane is a
mechanical trail 4-engined turboprop, seating 54 passengers, that is no longer in
geometric trail production but is still in service. Nose landing gear shimmy
half axle length appeared early in the flight development phase of this aircraft
moment due to tire torsional stiffness and re-appeared again after its introduction into service.
roll moment
yaw moment The DASH 8 airplane is a twin turboprop that seats 36 passen-
tire (deflected) rolling radius gers in its earliest version, the series 100. Landing gear shimmy
undeflected tire radius did not occur physically in this case, but was predicted to occur
time during the design phase. What is troubling about this example is
displacement vector [ y, Q, yIT that it took an inordinate amount of time for de Havilland and the
ground speed landing gear vendor to settle on a design considered to be
lateral displacement shimmy free. Obviously, the design process was costly and,
tire equator coordinate at forward end of contact despite the time spent, the structural efficiency of the design
patch remains an open question: ie.. the landing gear may be heavier
tire equator coordinate at aft end of contacr parch than necessary.
tire lateral defection at forward end of contacr patch

Paper presented at the 81st Meeting of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibration-Free Landing Gear, held in Ban& Canada from 4-5 October 1995, and published in R-800.
1-2

These examples occurred over I O years ago. While this may be actuator and the second, the more classical process addressed
considered ancient history by some standards - the computer during airplane development.
industry, for example - it is the authors opinion that there has
been little progress in the state of the art of shimmy design in the In the first case, shimmy occurred at relatively low speed and
intervening years. produced alarmingly large lateral oscillations of the airplane
nose. One report tells of tire marks on the runway indicating
The authors havereason to believe that the shimmy problem is angular deflections of up to 40 from centre in one direction and
perceived by many to be a black art with no basis in sound up to 35degrees in the other. This rather violent form of
engineering methods. The authors disagree and are concerned shimmy was caused by backlash at the tiller inpur potentiometer
that this black art perception has hindered progress in shimmy to the steering actuator and a modification was introduced to
design unnecessarily. The shimmy problem is no more myste- eliminated the backlash. This type of system induced shimmy
rious than, say, aircraft flutter with which the authors have expe- problem will not be discussed further.
rience. What is lacking is the accumulation, integration and
dissemination of currently available analytical and experimental The second type of shimmy, unlike the first, occurred at rela-
experience to support the development of practical design pro- tively high speed and with fairly modest amplitude oscillations
cesses. These, in turn, would yield structurally efficient, shimmy- of the gear assembly at a relatively high frequency (approxi-
free designs. The authors are optimistic that practical design pro- mately 20 Hz). It usually occurred with the steering system ON
cesses are feasible and believe that the aerospace industry would (ie., the mode with lower damping), but there were reports of
benefit if those processes were formulated and made generally shimmy with the steering system OFF as well.
available.
This latter shimmy problem stimulated a broad range of immedi-
Presented in this report are reviews of the DASH 7 and DASH 8 ate and longer term actions, all intended to achieve an expedient
shimmy problems experienced by de Havilland. The report out- and lasting solution. The actions taken were:
lines the analysis models and methods used and presents the sta-
bility results obtained from those methods. The report concludes (1) Introduce (large) unpaved runway tires across the
with a recommendation for the development of advisory material fleet because there were fewer reported incidents of
on shimmy design. shimmy involving these tires.

(2) Recommend operational procedures for curtailing

4. DASH 7 NOSE LANDING GEAR SHIMMY EXPERI- shimmy onset until a solution was found; specifi-
ENCE cally, that steering be left off during landing and that
4

4.1 The Occurrence of Shimmy steering engagement be avoided at high speed.

The nose landing gear on the DASH 7 airplane (see Figure I ) has (3) Undertake inspections of the nose landing gear to
two modes of operation, a castering mode with high angular ensure that its modification status is current and its
damping - sometimes referred to as the high damped mode - maintenance condition is to specification (e.g.,
and a steering mode with relatively low angular damping. Set for wheels within balance, tires in good condition, etc.).
castering at touchdown, the nose gear automatically transitions to
the steering mode by a weight-on-wheels switch. (4) Institute a questionnaire style reporting procedure to
accumulate information on shimmy occurrences in
In 1976, during the flight development trials, nose landing gear service.
shimmy was experienced in the lower-damped steering mode.
This resulted in two modifications to the design: (5) Undertake taxi trials using a typically worn landing
gear to obtain quantitative information on nose gear
(1) the introduction of a 0.75 second time delay between shimmy.
the nose wheel weight switch closure and power
steering activation to allow the spin-up transient to (6) Measure the frequency response characteristics of
subside before switching into the lower damped the steering actuator in an isolated test rig to estab-
steering mode. lish its (dynamic) stiffness and damping properties
as a function of frequency.
(2) the introduction of a 0.01 inch diameter cross bleed
damper in the steering mode to increase the angular (7) Develop an in-house analytical methodology cali-
damping in that mode. Further additional damping brated to the results obtained from the above tests.
was prohibited by steering torque requirements. This analytical capability could then be used to eval-
uate the benefit of design treatments; such as. a
The shimmy problem was thought to have been solved by these torque link damper, torsionally coupled nose wheels
modifications as there were no further occurrences during the (live axle), changes in trail, mass balance, etc.
flight trials. However, by the Fall of 1980, reports of nose wheei
shimmy were accumulating from aircraft in service. These prob- It is not the purpose of this report to review each and every action
lems took two forms, one involving feedback from the steering taken to resolve the DASH 7 nose gear shimmy problem: rather.
 1-3

the purpose for presenting the above list is to indicate the effort As indicated in that figure, effective values for rotation arm
required to resolve a shimmy problem when it is experienced in length, inertia, damping and frequency are dependent on the
service. Following are some experimental and analytical results length of the shock strut. Coulomb damping or free play is not
which were important in helping us develop a physical under- included in the side displacement model.
standing of our nose gear shimmy process and in establishing our
analysis methods. The equations of motion for shimmy analysis, including the von
Schlippe and Moreland tire models, are presented in Appendix
4.2 Shimmy Test Results A. The tire coefficients used in the equations derive directly from
The test results presented in Figures 2, 3 and 4 are instructive. In Reference 2, except for Morelands tire time constant, 7, , which
all cases, the ground speed is about 55 knots. Figures 2 and 3 was selected to conform with test results. The structural damping
apply to a tight gear for which free play in the scissors - the ratio, 5 , . was assumed to lie in the range zero to 0.03.
only significant source of free play - was removed using shims.
In the case of Figure 2, the weight-on-wheels switch was deacti- 4.4 Results
vated to allow excitation by the spin-up transient and steering Stability analysis results for both the von Schlippe and Moreland
was ON at the outset. The result clearly shows the landing gear is tire models are presented in Figure 6 for the case of steering sys-
fundamentally unstable at 55 kts in this low-damped configura- tem ON and no free play. As noted, all curves except one apply
tion. for a single point model of the nose gear. Also plotted in Figure 6
is the damping result obtained from the test data given in
In Figure 3, the landing was made with the time delay circuitry Figure 2, a single point at 55 knots ground speed with a damping
operating in order to eliminate excitation by the spin-up tran- value of 0.007.
sient. With steering OFF at touchdown, the large transient oscil-
lations due to spin-up are observed to attenuate quickly, due to The predicted frequencies are not shown but lie in the 26 Hz
the relatively high damping in the castering mode. While small range. This is about 20% higher than the value measured and is
random vibrations persist, there is no evidence of shimmy after attributed to mass and stiffness modelling inaccuracies. The
the time delay period when steering is ON, showing that the time effect of this higher torsion frequency was determined by analy-
delay was effective in circumventing the excitation of shimmy sis to increase the predicted stability by 0.01; i.e., to lower the
by the wheel spin-up transient. curves in Figure 6 by that amount.

Figure 4a shows results for which the shims in the torque scissors With respect to damping, three curves are indicated to give rea-
were removed to produce approximately 1/8 inch of free play in sonably good agreement to the test result in the 55 knot speed
the scissor apex. Again, steering was OFF at touchdown and the range: the von Schlippe single point result for 6, = 0 , the von
transient oscillations are observed to decay to limit cycle oscilla- Schlippe distributed mass model result for 6, = 0.03 and the
tions within the free play range wherein, steering actuator damp- Moreland single point result for 5, = 0.03 and T~* - 0.002.
ing has negligible effect. With steering ON - activated after the Clearly, much better agreement with this single test point could
0.75 second time delay - and with torsion damping thus have been obtained both in frequency and damping by applying
decreased, the oscillations within the free-play band cause the arbitrary adjustments to the parameter values initially chosen.
gear to go unstable, that is, to shimmy (Figure 4b). However, this was not pursued because the problem remained
that no conclusion could be drawn as to which structural and tire
4.3 Analysis Model models would give the most reliable result due to the lack of test
The analysis model used to investigate shimmy follows the data.
approach used by Lockheed in Reference 1. It consists of a tor-
sion model defined relative to the inclined leg axis and a side dis- The analysis was extended to investigate the effect of free play
placement model (Figure 5). The torsion model is made up of 4 using the distributed mass model (Figure5a) and the von
concentrated inertias to give the freedoms required to represent Schlippe tire model. These results demonstrate why free play
free play (backlash), bearing friction (coulomb damping) and must be avoided in landing gear design. Its presence can result in
torque scissor stiffness and free play (both functions of stroke). limit cycle vibrations at speeds lower than the classical shimmy
Fuselage torsion stiffness and trunnion attachment free play are speed - the speed obtained from the solution for a tight gear.
also included in this model. Initially, the steering actuator stiff-
ness was defined solely by the value of bulk modulus for hydrau- The addition of mass balance to the shock strut, was investi-
lic fluid which was assumed to have a value of approximately gated as a means for improving nose gear stability. The results
100,000 psi (with typical air entrainment). This value was subse- presented in Figure 8 show that the forward placement of weight
quently reduced to about 34,000 psi to account for the flexibility has a stabilizing effect and that the most effective arm length, at
of the steering actuator structure determined from an isolated least for this landing gear, is equal to the tire radius. This
shake test of that unit. approach ultimately led to a modification - using a 25 pound bob
weight - which was made available to operators as an option.
The side displacement model used eliminates the y independent While the increase in stability is predicted to be only 0.01, it is
degree of freedom by assuming an infinitely stiff leg free to roll the authors opinion, based on reported performance, that the
( @) about an effective rotation point as illustrated in Figure 5b. mass balance modification is more effective than predicted.
5. DASH 8 MAIN LANDING GEAR SHIMMY EXPERI- Increasing the side bending stiffness was also determined to be I
ENCE important but not as important as increasing the torsion stiffness.
5.1 The Occurrence of Shimmy It was also determined that shimmy stability could be improved
by increasing the mechanical trail so the design was changed to
Early in the design phase, the vendor for the main landing gear,
increase the mechanical trail to the maximum extent possible.
using analysis methods which they considered reliable, reported
that shimmy would occur within the operating speed range of the The results from more elaborate analysis models indicated that a
airplane and that torsion and side bending stiffnesses would have
torsion stiffness of 2 . 2 ~ 1 Ib-idrad
0~ and a side bending stiffness
to be increased to achieve stability. Because of the weight
in the range of 6,000 to 8,000 Ibhn would be required to preclude
increase involved, many months of intense analysis effort and shimmy. By mid 1981, these stiffness requirements grew to a
debate were required before de Havilland and the landing gear
high of 2 . 4 ~ 1 0Ib-idrad
~ and 14,000 I b h . As noted above, it
vendor agreed on final design values. was only after a long, protracted period that the final design val-
ues of approximately 1 . 8 ~ 1 0Ib-idrad
~ and 7,600 I b h . with
5.2 Analysis Models
nacelle flexibility included, were agreed to by both the vendor
The main landing gear for the DASH 8 airplane is shown in Fig- and de Havilland.
ure 9. The DASH 8 is a high wing aircraft with the main landing
gear supported from within the nacelle. As a result, the landing Final analysis results are presented in Figures 1 1 and 12. Stabil-
gear is relatively long and flexible and therefore may be more ity in the side bending mode is of little interest because the
prone to shimmy. damping values in this mode are high and increasing at speeds
above the no-slip speed. Regarding torsion stability, Figure 11
During the course of the shimmy investigation, a variety of gear shows the rigid wing giving generally conservative results com-
structural models were developed ranging from simple stick pared to the full finite element analysis. Figure 12 shows stability
(Figure IO), to full finite element. In all cases, except for the in the torsion mode increases with stroke. The results indicate the
modal analysis discussed below, these structural models were landing gear to be stable for the complete operating speed range
used to define a 3x3 stiffness (or flexibility) matrix defined at the of interest with a damping margin of 0.03.
axle centreline in the single point analysis method.

In the modal analysis method, the finite element model for the
landing gear and nacelle structure was extended to include a stick
model for the wing with distributed mass and stiffness properties.
The mode shapes and frequencies forthis system were calculated
and the shimmy equations were then set up and solved in modal
form.
Summarized in Figure 13 are the results of a non-linear analysis
to establish the influence of bearing friction and free play on
shimmy. These results contributed to the specification of free
play limits in the main landing gear design.

6. DISCUSSION
I
I

In the case of the DASH 7, it appears that the nose landing gear - -
The single point analysis method was extended to include free problem has been contained provided the tires are matched and
play. Its effect was simulated (in a global sense) by replacing the in good condition, the wheels are kept in good dynamic balance,
linear stiffness matrix defined at the wheel axle centreline with a free play is limited and nose gear touchdown and taxi speeds are -
non-linear stiffness matrix having low angular stiffness, kyw , kept low. Nose gear mass balance is generally considered to be
within the free play range and full predicted stiffness outside that beneficial by those operators who elected to have that option.
range. These equations were solved in the time domain. Obviously, the containment of nose gear shimmy has increased
the maintenance cost, pilot workload and in some instances, the
Both Moreland and von Schlippe tire models were used to repre- airplane weight.
sent tire dynamic properties. However, the Moreland model was
considered less reliable for absolute stability results because In the case of the DASH 8, with over 400 airplanes in service, all
there was no empirical method for defining the tire time constant, shimmy free, it is safe to say that the main landing gear design
T,, (see comments in Appendix A). has now been proven. However, the main concerns in this case
are the time and effort required to establish the final design
For structural damping, a typical value of 6, = 0.03 was requirements and the degree of over-design (weight penalty) that
assumed in the computations. However, no allowance for this may have been included in the final product.
damping was given in defining stability; that is, stability curves
had to have a margin of 0.03 to be considered shimmy free.
7. RECOMMENDATIONS
5.3 Analysis Results At the time the DASH 8 was being designed, there was no well
established consensus on what constituted a reliable shimmy
In January 1981, using the simpler stick models shown in
design methodology. This still appears to be the case today. For
Figure IO, de havilland confirmed that the landing gear would be
this reason. it is recommended that some guidance information
unstable within the operating speed range and that the most
be provided - similar to an advisory circular issued by aircraft
effective way to increase the shimmy speed was to increase the
airworthiness authorities. This advisory material should:
torsion stiffness of the gear (at the time 1.13~10~ Ib-in/rad).
 7-5

Establish standardized terminology lined herein. At the very least, their resolution would have been
more expeditious.
Give guidelines on structural modeling
REFERENCES
Define tire models, including the assumptions and
limitations (1)Active Shimmy Control System, AFFDL-TR-75-136.

. Review and update tire properties and ranges of

parameter variation
Lockheed-California Company, Burbank California,
December, 1975.

(2) Smiley, R. F., Home, W. B. Mechanical Properties of

Provide expressions for the tire terms appearing in Pneumatic Tires with Special Reference to Modem Aircraft
the equations of motion for a variety of wheel con- Tires, NASA TR-R-64, 1960
figurations
(3) Leve, H. L., Designing Stable Dual Wheel Gears, AlAA
Recommend design analysis guidelines taking into 69-769, July 1969
account free play, recommended stability margins,
(4) Moreland, W.J., The Story of Shimmy, Journal of rhe
structural and tire parameter variability, etc. (Table 1
Aeronautical Sciences, pp. 793 - 808 ( 1954).
suggests variances for some parameters appearing in
the equations of motion and indicates whether they ( 5 ) Collins, R. L., Frequency Response of Tires using the
have a stabilizing or destabilizing effect on Point Contact Theory, J. Aircraf, June 1972.
shimmy.)
(6) Clark, S . K., Dynamic Properties of Aircraft Tires, J. Air-
Provide guidelines on component and vehicle test craji, March 1974.
procedures:
tire properties (7) Collins, R. L. and Black, R. J., T i r e Parameters for Land-
structural properties (static and dynamic) ing-Gear Shimmy Studies, J. Aircraft, May-June 1969
complete installed system
(8) Grossman, F-15 Nose Landing Gear Shimmy, Taxi Test
Provide practical design information on anti- and correlative Analyses, SAE Paper, No. 801239.
shimmy tires, live axle, etc.

. Provide standardized analysis cases for a variety of

landing gear configurations
(9) Shimmy Analysis Addendum to QTR No. ATR 923, 13.5 x
6.00 - 4, DR17922T 14 PR Outrigger Tire, Dunlop Manu-
facturing, England, September 1981.

Provide information on related vibration processes (10) von Schlippe, B., Shimmy of a Pneumatic Wheel, NACA
and problems - sometimes confused with shimmy - TM-1365,August 1954
such as wheel unbalance, tire out-of-round. judder.
anti-skid processes, steering feedback processes,
APPENDIX A. SHIMMY ANALYSIS MODELS
gear walk, etc. Some of these processes may warrant
separate advisory material of their own.
This Appendix describes the shimmy analysis methods devel-
oped at de Havilland.
Provide landing gear maintenance guidelines, based
on analytical and experimental data, for minimizing
A.l Equations of Motion
landing gear vibrations
A much simplified model for a single or dual wheel landing gear
Provide a bibliography is illustrated in Figure AI which shows an inclined shock strut
with mechanical trail, L, measured parallel to the torsion axis and
geometric trail, LG = R x sineg, where Os is the inclination
8. CONCLUDING REMARKS angle and R is the deflected tire radius.
As a consequence of the DASH 7 and DASH 8 shimmy prob-
lems experienced at de Havilland, the authors were required to Also shown in Figure AI are two equally valid coordinate analy-
develop analytical tools for problem solving and design. This sis systems, one located at the apex of the shock strut and the
work was accomplished successfully despite the lack of a stan- mechanical trail arm and aligned with the strut, and the other
dardized industry accepted methodology at the time. It is the located at the wheel axle with axes orthogonal to the ground
authors contention that reliable shimmy design procedures are plane. The former system simplifies the mechanical terms in the
required and that it is feasible to do so. Had advisory material equations of motion while the latter simplifies the tire force and
been available during the design of the DASH 7 and DASH 8 moment terms. In order to expose more clearly the less familiar
landing gears, de Havilland may not have had the problems out- tire terms, the latter coordinate system is used in this report to
develop the equations of motion. These equations are written in A.3 TireTerms
terms of the lateral ( y), roll ( 4 ) and yaw (y)degrees of free- During shimmy, the motions of the gear axle are coupled not
dom defined at the wheel axle; only through the gear structure as expressed by the terms on the
left hand side of equation (AI) but also through the interaction
between the tire and ground, as represented by the tire forces and
moments appearing on the right hand side of that equation.

Two tire models were considered, the Moreland Point Contact

model and the von Schlippe Stretched String model. Both
models are presented below in a coordinate system orthogonal to

where U = [i] the ground plane. This was done in order to develop the simplest
possible expressions for the tire forces and moments. Formula-
tions for single and dual wheel assemblies are included.

A.4 The Moreland Point Contact Model (Single Wheel)

and where I = inertia matrix
The Moreland point contact model (Reference 4) expresses the
D = damping matrix
tire contact forces and moments in terms of the tire cornering
G =gyroscopic matrix

E;]
force, Fc , and pneumatic trail, E ~ as, follows (Figure A2):
K = stiffness matrix
A.2 Structural Terms
Formulations for the left hand side terms are not given in this = (A3)
report but they can be found in References 1 and 3. It is impera-
tive that utmost care be taken in defining and verifying the
assumptions made in establishing the analysis model and in where
developing the data used. For instance, if a landing gear is Fc = K T ~ A +C T ~ A (A41
assumed to be a uniform structure, its mass and inertia effects
can be lumped at the wheel axle and a single point model ana- and where K T =
~ tire lateral spring constant
lyzed. In this case the mass and inertias of the wheel rolling C T=~ tire lateral damping coefficient
assembly should be supplemented with appropriate effective = pneumatic trail parallel to ground
masses and inertias of the supporting gear structure. The 3x3 A = tire lateral deflection (elastic)
mass matrix, M, will contain (off-diagonal) terms that couple the R = tire (deflected) rolling radius
three degrees of freedom, y, 4 and y. If it is important to
include detailed representations of bearing friction or free play To enable a solution of the shimmy equations, Moreland postu- -
(backlash), then a distributed mass model will be required to pro- lates an alternate expression for the tire cornering force in terms
vide the additional degrees of freedom needed to represent these of a tire cornering power, C y , and time constant, T,, , as follows:
characteristics.
Fc = CwWt + CtyTu$t (45)
The damping matrix, D, should include both structural and bear-
ing friction damping terms, if appropriate. As indicated ahove, where y, = the tire elastic twist deflection
this may require the development of a more complex gear
dynamic model than can be expressed by a single point formula- Finally, Moreland assumes the absence of slipping or scrubbing I

tion with 3x3 matrices. between the tire and ground by imposing the kinematic relation-

The gyroscopic matrix contains roll-yaw coupling terms with

coefficients I O X ( V/RT) , where 10 is the polar moment of
ship;
)i+A+V(W+Wt) = 0 (A6) I
inertia of the wheel rolling assembly, V is the ground speed and Embodied in the above kinematic relation is the assumption of
RT is the tire undeflected radius. no slipping or scrubbing. According to the empirical data pre-
sented in References 5 and 6, this assumption is invalid and
As in the case of the mass matrix, if the gear structure is assumed therefore, classical shimmy is not likely to occur at ground
to be uniform, its stiffness properties can be represented by a 3x3 speeds
wh
stiffness matrix, K,defined at the wheel axle. Once again, if it is v<o.8 (A71
important to include detailed representation of bearing friction or
free play, then additional degrees of freedom and stiffness terms where w is the shimmy frequency and h is half the tire footprint
will be required. length. (The von Schlippe model described below also assumes
that no slipping or scrubbing occurs between tire and ground and
Finally, the mass, stiffness and damping effects of the landing therefore it too is subject to the same speed limitation.)
gears supporting structure should be included in the equations of
motion because they can have a strong effect on the stability Empirical values for the tire parameters appearing in the above
results. equations can be obtained from Reference 2 with one notable
 7-1

exception - the tire time constantrv. Choosing a reliable value in terms of the yawed-rolling relaxation length, 2.0. This geo-
for rY is generally a problem because no consistency can be metric relationship can also be deduced from Figure A4.
found in the literature for this parameter as illustrated by the
compiled data of Figure A3. Since r, has a strong influence on
the results, it is recommended that the value used should be vali-
dated in some way if the results are to be accepted in absolute
terms.
Empirical values for all of the parameters required by the von
Also. while empirical values for the pneumatic trail, Ep. are Schlipp tire model can be obtained from Reference 2. It might
available. some analysts prefer to base the tire torsion moment, be noted that some analysts also extend the expression for M Iby
.
Mr directly on the tire torsional stiffness K T ~viz;
, including the tire torsion damping term, C T ~viz;
,

Mt = KT,W (A8) + ~C T ~ ' &

Mt = K T ~ W (AI3
instead of usrng the expression proposed by Moreland Thts vari-
mon has not been used in the present report. This variation has not been used in the present repon.

A 5 The von SchUppe Stretched String Model (SingleWheel) A.6 Dual Wheel G e m
The von Schlippe tire model (References IO and I) assumes that The Moreland and von Schlippe tire models given above apply
the tire equator ISrepresented by a "stretched string" that takes for a landing gcar with a single wheel. The following steps
on the deflected shape shown in Figure A4 due to out-of-plane extend those formulationsto uual wheel landing gears:

E] E]
mouons In this model the tire forces and moments are:
(1) Factor the force and moment expressions by 2

(2) Add the following rolling moment term to aecount

= (AS) forthe vertical &round reaction forces. Assuming the
total static force. Fz, is shared qually, the ground
where reaction force on each tire is comprised of a static
F, = K T ~ A + C T ~ A force and a dynamic force due to roll displacement
$.Thus.
Mt = KT.N
AM$ = L ~ ( F ~ - F , R ) (AW
and where the tire elastic lateral and torsional deflections are
given by: where the (upward acting) ground reaction forces on the
A = -z + 2 right and left tires are. respectively;
2

F,R = F.
2 +L ~ @ K ,
(A171
F
As before, h equals half the footprint length. F.L = ;-L~@K.
Z and 2 are defined in Figure A4 from which the following geo-
metric relationships can be deduced. and where LR = half axle length
Kz = the tire vertical stiffness
Z = Y-y-R@-hW
(3) Add the following yawing moment term to account
2 = ?-y-R$+hW for the ground friction force;

AM^ = L~~(F,L-F,R)
Finally, the kinematic relationship that assumes no slippage
between the tire and ground is given by the following expression where p = tire rolling friction coefficient
E9
7-9
7-10

.
 7-1 I

.. .. ..
: : : :
:i: - I
; : t i ; -

: r
7-12

I 1%

3>
6 t
!Ii aa
5?
Y
 8- 1

Unsteady Tire Dynamics and the Application Thereof to Shimmy and Landing
Load Computations
Klaus Koenig
Daimler-Benz Aerospace Airbus GmbH
HunefeldstraBe 1-5
28199 Bremen, Germany

SUMMARY some amendments from other sources a complete set of

A complete set of nonlinear equations of unsteady tire dy- equations for tire dynamics is given. Eventually, its ap-
namics for all the 6 degrees of freedom of the wheel move- plication to shimmy and landing loads computation is stu-
ment is presented. The coefficients of these equations are died.
based on available publications and may need some im-
provements. 2. GENERAL TIRE DYNAMICS
The influence of some of the coefficients which are new The most simple way to establish reliable tire dynamics at
or controversely discussed is shown revealing areas of least applicable to shimmy or linear stability studies would
caution and stressing the necessity of adequate tire tests. be to put a rolling wheel with tire on a drum to excite the
Nevertheless it could be demonstrated that reasonable wheel at its centre by frequency sweeps of deflections and
shimmy analyses are possible with these equations and to measure the load response. If this is done at a well de-
that the computation of realistic unsymmetric landing im- fined working point of the tire and subsequently for all the
pacts require their inclusion as a must. 6 components of the deflection applicable at the wheel
centre 36 transfer functions (frequency response func-
1. INTRODUCTION tions) for all the 6 possible perturbations of the steady state
The wheel - without example in nature - is one of the most movement of the wheel centre will be available. Each
efficient inventions of human spirit and a fundamental ele- transfer function can be linearly approximated by a frac-
ment of technical constructions. The inflated tire is a rela- tion of polynominals of the imaginary Fourier frequency.
tively new additive to the wheel but most probably its best
- improvement. If one uses as a first approximation the same denominator
Modern civilization is inconceiveable without wheels and polynomial for the same load component and accepts that
. for linear systems superposition is possible, then Fourier
tires. Millions of them are rolling on the streets every day
and every minute . But in spite of the huge application of transformation allows to establish with the polynomial co-
tires no comprehensive mathematical description of their efficients the simple matrix equation*:
characteristics is available.
It is nearly unbelievable but it seems that all the progress
of modern transportation technique was possible without n m

a qualified theory of tire dynamics. There are early publi-

cations with rather good results already 50 years ago but where PtR = LpxtR, P Y t R ? PZtR, Pa)tR,PetR, P V t R 1
there is no reliable up-to-date theory available (s. ref. 8). is the perturbation vector of loads and
What are the reasons? Is there really no need for a better F:Rits derivative to the time
theory? Was it too expensive to establish one? Was it just
not published or available or was it too difficult for the en- rIA = [XtA, ztA, @all? V m t ]
gineers? is the perturbation vector of wheel
centre deflections
Tire dynamics is characterized as the dynamics of any or-
dinary vibrator by the interaction of elasticities and Bn,Wm are coefficient matrices.The diagonal
masses. But superimposed in this special case is the creep- Bn contains the denominator and the
ing process within the contact area between tire and complete Wm the nominator
ground - the so called relaxation mechanism. coefficients of the nth or mth order
One of the earliest studies of tire behavior was performed polynomials
by Fromm at the German university of Danzig. The first That is all, at least for the small perturbations of shimmy.
real shimmy theory was published by von Schlippe and
Dietrich (s. ref. 1). The last fundamental collection of ap- *The more general solution would result in 36 scalar dif-
plicable data was presented by Smiley and Horne (s. ref. ferential equations which must be solvedfirst to allow a
2) in 1960. summation of the 6 load portions of each load component
So what is available in 1995 can only be called a substitute caused by the 6 different excitations.
or stopgap but not an applicable theory of modern tires By the way, f o r second order polynomials one would get
for reliable dynamic studies of dynamic structures. 2 times 36 eigenvalues. That would allow 36 possible tire
Therefore in the following first some thoughts are spent on load oscillations without wheel movement perturbations
how to get qualified tire data. After that published shimmy and 36 wheel movement oscillations without tire load per-
theories are compared. Based on this and together with turbations.

Paper presented at the 81st Meeting of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibration-Free Landing Gear, held in Banff, Canada from 4-5 October 1995, and published in R-800.
 8-2

For larger deformations as they occur at landing impact theories use a relaxation length [m]. This leads to the
further investigations with larger amplitudes and with and fact that e.g. B2.. depend only on the rolling velo-
without rolling velocity are necessary. city v in the Mooreland theory but on v2 in the other
But this should be a second step to be defined in detail after theories.
successful execution of the first one with small perturba- - Moreland originally only studied I$. The extension
tions. by y was introduced by Rogers and Brewer, and the
Partially such studies have already been performed (s. ref. extension by $ by Edman (9).
5 ) but no complete test result is known to the author.
Nevertheless, since no principle obstacle to such an ap- 4. A COMPLETE SET OF TIRE EQUATIONS
proach to establish reliable equations of tire dynamic is ev- For the following, Smileys equations as given in chapter
ident, it can be hoped that perhaps in another 50 years such 3 will be used as a basis. They are modified by neglection
data will be available for the benefit of future engineers of minor unknown elements (t=O) inclusion of informa-
and of technical progress - if further progress is ever tion from ref. 2 (WO,,, WO,, WO,,), assumption of sym-
wanted at that time. metry (Wl,,, W2,,) and some other simplifications (WO?,
: analog to Wl,, and Wl,,, W2+, : reduction to the main
3. COMPARISON OF SOME PUBLISHED TIRE element). But this is not sufficient. Even for tire rolling in
EQUATIONS steady state one needs information for the x,z and 6 com-
The following publications will be compared: ponents. They are taken from ref. 2 and 6 and mainly based
- von Schlippe, and Dietrich, (s. ref. 1) on radial tire stiffness and respectively the assumption that
I - Smiley, (s. ref. 3) x loads depend on a friction coefficient and that this coeffi-
, - Pacejka, (s. ref. 4) cient itself depends on the slip ratio as shown in Fig. 4-1.
- Rogers, and Brewer, (s. ref. 5 ) with the Moreland
theory. For the not free rolling tire in the spin up phase of landing
If the equations of movement as given in these publica- the situation is more complicated. Nearly nothing is
tions are transformed to the coordinate system of Fig. 3-1 known. The only fact is, that for the pure sliding, all hori-
and linearized by series developments up to elements of zontal ground loads depend on the vertical load and a fric-
second order coefficients for matrices B and W as defined tion coefficient while their direction is opposite to the
in chapter 2 can be derived as given in table 3-1. movement.
I These equations are given in their original symbols with-
out further explanation of the symbols. To have at least something in hand a very crude assump-
Nevertheless some basic conclusions can be drawn: tion was made. It defines that during the spin up phase both
- There are really only diagonal elements in the B types of loads are acting but the one for pure rolling with
matrices. a factor between 0 and 1 increasing with the slip ratio and
- B.,, = B.,, = B.,, are equal for all theories . This the other one for pure sliding decreasing in the same way
may mean that the tire can vibrate in itself with a down from 1 to 0 until the slip ratio reaches its zero for the
single mode, showing at least 3 different load compo- first time.
nents More details of the derivation should not be given here to
I
- The different theories dont place coefficients at the avoid any boredom. But the final result is presented in
same places. table 4-1.
- A larger portion of the matrices is non-symmetric. These are rather complicated non linear equations but they
, - Von Schlippe and Dietrich do not study the tilt. should be valid for the whole landing process of horizontal
Pacejka does not give tilt moments . landings.
- Remarkable are the missing coefficients Wl,,
W2,, in the theories of Smiley and Pacejka, the mis- 5. SHIMMY ANALYSES
sing coefficient W2w, WO,,, WO,,, WO,, in More- 5.1. Tire Equations
lands theory and the missing coefficient W q , in First linearized tire equations should be derived from the
I
Pacejkas theory. nonlinear set given in table 4-1.
- The absence of the elements WO,,, WO,, WO,, is Details are given in table 5.1-1. This leads to extended
caused by the relaxation mechanism. Loadings equations of unsteady tire loads for all the 6 degrees of
caused by sideward deflections will disappear with wheel freedom.
time.
- The theories of von Schlippe and Dietrich, Smiley and 5.2. Influence of Tires on Shimmy
Pacejka look rather similar, they are based on equal To study this it is reasonable to look at a rigid undercar-
basic thougths. The Moreland theory is funda- riage to avoid coupling of structure and tire elasticity. For
mentally different the most simple rigid model of Fig. 5.2-1, after some rear-
- Moreland introduces a time constant C1 [sec] for de- rangements, one obtains the equations of motion
scription of the relaxation mechanism while the other
 8-3

approach to Variant S.
Used are (Variant E) :
Type I11
o = z0 wi,, -
(2 d, q,, + rr e)-Ip e PR(O)= 1765.8 N
zp = 0.02 m
ro =0.13m
~0 = 0.12 m
+ is,(Wlqq-LR
* * (Wlqy + W1,) + L i * Wl,,) + dR = 0.085 m
pi = 225630 N/m2
+ qst (w2qq-LR (w2qy + w2,) + L i w2,,) Pr =Pi
with
=o
pk =0.8
* p = pq(R1) + pq(R2) - LR (py(R1) + py(R2)) sk Z0.2
m, = 2.4525 kg
8 = 8,(R2) - 8,(R1) IR = 0.05 1993 kg m2
Iw = I, +
m, (di Lk) + I, = 0.0981 kg m2
m,=O
I, = radial moment of inertia of wheel including tire
I, = polar moment of inertia of wheel including tire For the single wheel case the estimation did not lead to ex-
actly the same stability areas as Variant S.
d, = swivel damper constant For the double wheel first only the wheel distance dRwas
introduced to show the influence of the increased swivel
An eigenvalue analysis of these equations delivers fre-
inertia 2.rnw.d~z.Then the polar inertia I, of the wheel was
quencies and modal dampings. Different cases were ana-
introduced and the influence of the swivel damper shown.
lyzed and compared.
~
Finally the tire coefficients W(x,e) are added.
As one can see there are some influences. The shimmy ve-
Variant S :
I Ki = 44145 N/m
K, = 298.224 Nm/rad
LI = 0.142222 m
locity of the double wheel is smaller than that of the single
wheel. Both the swivel and polar inertia contribute to this
effect.

I-: L2=5.1136. lO-3m2

I R = 0.051993 kg.m2
For smaller trails there is no influence of the tire coeffi-
cients W(x,e) while for larger trails the influence is strong

i -
I

!
-
m, = 2.4525 kg
K,= 0
WI,, = 0
and increases the shimmy velocity.

Fig. 5.2-5 shows the influence of some other coefficients.

Basis is again Variant S .
.
These are the original data from ref. 3 case I approxima- First W1, is studied since it was W1, #O in the More-
tion B. They are based on a test performed by v. Schlippe land theory. W1, = Wlqy was tried (case a) but a corn-
and Dietrich. pletely different stability boundary occurred, absolutely
Fig. 5.2-2 shows eigenfrequencies and modal dampings not in agreement with the test results. Next by analogy
over the rolling velocity for the single wheel case dR = 0, with W2,,/ Wl,, it was set W1, = K1.L2/v (case b) but
LR = 0,08m, d, = 0,06 Nm/(rad/sec).The eigenvalues of with the same result. Even the inclusion of a swivel dam-
the decoupled tire mode are included. per did not help.
One can discern that the frequencies of the tire modes in- Another critical element is W2,,,,+,which is not included in
crease with velocity, while the frequencies of the shimmy Morelands theory. But with W h q = 0 the whole stability
mode remains about constant. area for smaller trails disappears.
The dampings of the tire modes are rather large. The Eventually a different influence of v in W2,, and W2,,
damping of the shimmy mode crosses zero at about was studied (case c) to allow better comparison with the
3,7 m / s . At larger velocities the system is instable. Moreland theory. It was assumed that W2,, and W&
Fig. 5.2-3 shows the stability areas and the shimmy fre- were measured at 3 m/s and that these coefficients depend
quencies on the stability boundary for different swivel only on I/v but not on l/v2.
damping values d, together with the test results of The result was not very different as one can see in
v. Schlippe and Dietrich and theoretical results from Fig.5.2-5.
Smiley ref. 3. So the final conclusion is that detailed tests are necessary
Smileys results were computed without swivel damper. to check W1, and W2,,.
The agreement is rather good. With d, = 0,6 Nm /(rad/see) Caution is necessary and if nothing better is known the co-
the computation fits to the test results. efficients given by Smiley should be chosen.

In Fig. 5.2-4 the influence of x and 8 movement is studied.

Unfortunately no measured tire data are available to do
this.
So an estimation, Variant E was necessary which should
8-4

6. COMPUTATION OF THE LANDING IMPACT- REFERENCES

For the landing impact often just the vertical stiffness of 1. Von Schlippe, B. and Dietrich, R.
the tire and a friction coefficient are considered. But that Zur Mechanik des Luftreifens
is not adequate in our times. With modern computers ZWB Verlag Oldenbourg, Miinchen, 1942
equations as given in table 4-1 can be handled without
problems. 2. Smiley, R.F. and Horne, W.B.
Of course the inclusion of such tires is only reasonable if MECHANICAL PROPERTIES OF PNEUMATIC
also the elasticity of the undercarriage structure is ade- TIRES WITH SPECIAL REFERENCE TO MOD-
quately modelled. ERN TIRES
At least 6 degrees of freedom for each wheel centre should NASA TR-64,1960 I
be available and the variation with shock absorber deflec- I
tion should be included. Ref. 7 gives some indications how 3. Smiley, R.F.
to do this. CORRELATION, EVALUATION AND EXTEN-
Fig. 6-1 gives an example of a computation. Loads at the SION OF LINEARIZED THEORIES FOR TIRE
right and left main undercarriage together with shock ab- MOTION AND WHEEL SHIMMY
I
sorber deflection (U)and parameters of the aircraft move- NACA-TN-3632,1956
ment are shown. The results belong to a smaller aircraft
with one wheel per undercarriage. The landing conditions 4. Pacejka, H.B.
are 2.44 m/s sink speed (its), 69.5 m / s horizontal speed THE WHEEL SHIMMY PHENOMENON
(is), 9 degrees pitch (fif,) and 8 degrees yaw (Vfl) angle. Dissertation TH Delft, 1966
$flis the bank angle and yts the lateral velocity of the air-

craft centre of gravity. pLis the total lift. The tire loads are 5. Rogers, L.C. and Brewer, H.K.
PX,R,p y l ~P, Z ~ Rand the undercarriage loads at the wheel Synthesis of Tire Equations for Use in Shimmy and
centre are pxuu,pyUu,pzuu.The P.~Rare measured parallel Other Dynamic Studies
to the ground, the peUu are measured in a coordinate system Journal of Aircraft VoI.8 No.9, Sept.1971, p.689-697
of the undercarriage. The undercarriage loads include the
tire loads and the wheel inertia loads.

In Fig. 6-1 one can see first of all how different the loads
6. Horne, W.B. and Leyland, T.J.W.
INFLUENCE OF TIRE TREAD PATTERN AND
RUNWAY SURFACE CONDITION ON BRAKING
c
of the two undercarriages are. FRICTION AND ROLLING RESISTANCE OF
Further, one can clearly discern that the loads oscillate and MODERN AIRCRAFT TIRE
that this oscillation was obviously induced by the wheel NASA-TN-D 1376
spin up / spring back effect.
The large sideloads are remarkable. They reach their max- 7. Koenig, K.
imum after spin up - on the rolling tire. They are nearly Die Lasten des Landesstoaes
doubled on the undercarriage if compared with the loads ZFW 3 (1979) H6, p.344-360
on the tire and they dont always decay with the vertical
loads. They are mainly caused by the tire dynamics and 8. Krabacher, W.E.
can not be predicted without such a complete set of equa- Aircraft Landing Gear Dynamics Present and
tions as presented above in table 4-1. Since the whole Future
movement of the aircraft in this unsymmetric landing case SAE-paper 93 1400 (1993, Dayton Conference)
is determined by the sideloads one can conclude that no
prediction of the unsymmetric impact loads which is in the 9. Edman,J.L.
least reliable can be computed without inclusion of the Experimental Study of Morelands Theory of
complete tire dynamics of the rolling tire under varying Shimmy
vertical deflections. WADC Techn. Rep. 56-197 (July 1958)
A measurement on the flying aircraft instead comes com-
paratively late and is tremendously more expensive.

Finally it should be said that also a shimmy analysis to-

gether with the landing impact may be more realistic and
would cover a larger area of possible instabilities includ-
ing nonlinear effects.
 8-5

cn
W
z
0
W
r
k
n
w
I
cn
i i
m
I- 3
e
U,
0
2
0
cn
z
8-6

total tire loads at the centre of the wheel Wl,, = K, z,/X, -

W2,, = K, * L1 * Z,
WO,, = - K,
Wl,, = - K, * L1 Xu\
2
W2,, = - K, * L2 /Xw
Stiffnesses
-
K, = 0.8 2 ro * (pi + 4 p,) - -i m+
+ Kxmin (fore/aft)
K,, = 2.4 * (pi 0.08 pr) . + -4
(radial)
pVtR = Pll, (1 - IsI * a) K, = (pi + 0.8 * pr) * di * b, * b, Kamin (torsion)+
if zp > 0; but all ptR= 0 Ky = 0957 * PR
if z,, I0 K, = 2 m, ro (L,, L, - + -jm/
radial deflection load /(3 n * (Lf rt))- +
(gyroscopic)
P R = 0 ; if zpe I 0 KL = bb do * (pi + 0.24 . pr) (1 - 0.7 * Zpc/do)
* b, (lateral)
lengths
L1 = Lr Lh + (damping length)
- +
L2 = (( 2 L, Lh ) . Lh / 2) (inertialength)
where PRK = PR(zpc = Zpk) L, = (2.8 - 0.5 pi/pr) * (1 - 4.5 zp,/(2 ro)) do
load portions of the rolling tire relaxation
- +
B2 py B1 * py py = Wl,,; y, + W2,, ;, +
y,- +
WO,, Qam Wl,, * Qm w2,, * Qm+ + + Lh = 0.85 2 * ro * Jzpe/(2 ro) - 2;,/(2 * ro)2
(U2 foot print)
+ WO,, * V m t + W1, * 3,, + W2y, Gmt general tire factors
+
B2 p$ B1 p 4 pQ = Wl,, jr, - + W2,,; + +
b, = 1 m, Xi/(4 * x ri * do pi) centrifugal
* Y, + WO,, * 4, + W,, * ?am + W2,, * ?am + -
b, = 250 zi,/(2 ro) (Type III and VII) torsion
+ WO, * V m t + W1, * V,, + W2, * V,, if Zpe / (2 ro) I 0.03
+
B2 * p$ B1 . pV + = Wl,, * y, W2,, * + = 15 (zp,/(2 ro) - 0.015) -
y, + WO,, Qam Wl,, + -
, + w2,, Gam+
Q if zpe/ (2 . ro) > 0.03
+ WO,, * V m t + W1,, * qmt+ W2,, 3,, = 475 zi,/(2 * ro) ( Type I)
but always if Zpe / (2 ro) 5 0.02
= 19 * (zp,/(2 * ro) - 0.01)
by1 II pk pztR I
IpQl I I Z* pk * PZ,R I if zpe /(2 ro) > 0.02
by = 0.21 (Type I);
-
Ipql II pk 0.44 ro pztR/6 I = 0.15 (Type I11 and VII) side
so that for a new start :
p. = 1p.l sgn(p.(t - At)) b, = 0.02 (Type I);
= 0.03 (Type 111 and VII) vertical
p. = 0 ; if sgn(p.) = sgn (p.)
t+At
by = 0.48 * do tilt
bb = 3 (Type I); = 2 (Type 111 and VII) side
p. = - p. . dt ; if sgn(p.) # sgn (p.)
deflections
coefficients of rolling tire zpc = zp - zpx- zpy- zp, effective
B2 = L2/X,2 2 0
B1 = L,/X, + +
zp = ro re z,
2 0
visible
Wl,, = -Kb * Ll/X, .2
W2,, = -Kb L2/X
2
- re = 0.01 m, . B,/pi centrifugal
= PR zpx = 0.1 I px,/Kx I x loading
Wl,, = Wl,, = (Ky - K,. z,) * Ll/X, zpy = by I PY, bm/K, I y loading
W2,, = W2,, = (K, - K, * z,) . L2/X2, -
ZP, = 0.1 I pq, * b, Lh/K, I - loading
WO, = K b *L, friction coefficient
w1, = w2,, = Wl,, = w2, = 0 p = c1.r+ (pk-pr) sk ; if-sk 5 s l Sk
WO,, = - PR by
= pk + (ps - pk) * (s - %)/(I - Sk) if not
Wl,, = - K, * z, L,/X,
W244 = - K, z, L,/X2, slip ratio
WO, = - (K, - Kb * zW) L , +
s = 1 6, (r0-zp/3)/i, but always Is1 5 1
Wl,, = K,/X, indicator of spin up period
W2,, = K, * L, a = 1 at the beginning of a horizontal landing
WO,, = 0.01 d i * Kx = 0 all time after s passed first time s = 0

Table 4 - 1 Complete Tire Equations

 8-7

wheel centre plane

wheel centre
A (axle, non rotating)
R (wheel, rotating)
definition of working point:
Pr = O (free rolling)
a =O (free rolling)
T- s(0) = 0 (free rolling)
y,(O) = Z,(O) = $JO) = $,,(O) = 0 (straight on)
zt 20 zrn
conditions of linearization:
< zpc < Zpk (no bottoming)
V
ro = zpx= zpy= zpv = 0 (simplification)
plane with ground plane
b, = 0 (linear vertical stiffness)
w.... = w....(0) (as would be measured)
B .... = B ....(0) (as would be measured)
l/X, E (1 - AX,/V)/V
abbreviations:
XO
(, ) = v
ztA(0) - 0,25 * pR(0)/Kx(O) = zo
bottom dew
rr = ro - zp(0)/3
additional tire coefficients for the extended
unsteady tire loads
Fig. 3-1 WHEEL COORDINATES
P
wlxx = - Pk pR(o)/(sk *
'

A = - pk PR(o)/(3 sk * rr)
wlxO = - pk ' pR(O) * rr/(sk
WO,, = - K,,(O)
WI,, = - 0,1 ,/E
wlOx = - V k pR(O) ' z(o)/(sk *
= - Pk PR(O) 2(0)/(3 sk * rr)
wlOO = - P k * pR(O) * rr zo/(sk *

slip ratio
1 Bl(x,z,8) = B2(x,z,8) = 0

Fig. 4-1 FRICTION COEFFICIENT Table 5.1-1 Definitions and Coefficients

DEPENDING ON SLIP RATIO for Linearized Shimmy
Analyses
Constants (input data)
TYP tire type
do tire width nominal (undeflected)
Kz2 radial tire stiffness after bottoming
mt tire mass
".
y'A1
m, wheel mass including tire S

Pi tire pressure
P' rated pressure

r0 tire radius nominal (undeflected)

sk slip ratio with maximal friction
Zpk bottoming deflection

Pk
ClS
maximal friction coefficient
full gliding friction coefficient
Y \

Pr full rolling friction coefficient

Fig. 5.2-1 SIMPLIFIED SHIMMY MODEL
Kxminf 0 minimal fore/aft stiffness
&,,,in f 0 minimal torsional stiffness

Table 4 - 1 Complete Tire Equations

(continued)
8-8

0.20
I
-E
sE 30
3
-
g 20 -._
2 0.16
10

0.12
unstable . .'
. .

0.08
I 1, 51 shimmy mode (flutter critical yaw)
+ f2' 52 stable yaw (coupled tire mode)
: . -.____ _ _ _ - _-_
e l3 EJ __ decoupled tire mode
~.

0.04

0
1 2 3 4 5 6 7 8 9
velocity v [m/sec]
Fig. 5.2-4
INFLUENCE OF x AND 6 MOVEMENT
Fig. 5.2-2 variations :
0 single wheel W(x.8) = dR = ds = 0
EIGENFREQUENCIES AND 0 double wheel W(x.8) = Ip = d. = 0
I '8 W(x.8) = ds= 0
MODAL DAMPINGS +A 0)
"
'8
W(x,8) = Ip = 0; ds= 0.06
variant E ; ds=O.oS
single wheel ; rigid u/c
Smiley case I approx. B
d, = O.OGNm/(rad/sec) ; LR= 0.08 m

I I - .
unstable stable
- area 11 I I area I

= s s s s s s s
0;s a b b b 0;s 0;s

Fig. 5.2-5
INFLUENCE OF y AND q~COEFFICIENTS
Fig. 5.2-3 STABILITY DIAGRAM single wheel o test results
single wheel - d, = 0.6 Nm/(rad/sec)
I-
 8-9

I)R (nose to the right)

V
Fig. 6-la LANDING IMPACT, LEFT MAIN UNDERCARRIAGE

I.

Fig. 6-lb LANDING IMPACT, RIGHT MAIN UNDERCARRIAGE

 9- 1

INFLUENCE OF NONLINEARITY ON THE SHIMMY BEHAVIOUR OF LANDING GEAR

P. WOERNER, 0. NOEL
Messier-Dowty S.A.
Zone ACronautique Louis Breguet
BP 10, 78142 VClizy Cedex
France

ABSTRACT Messier-Dowty SA. has performed many analyses on a

One of the most critical vibratory phenomena which may landing gear which showed some instabilities under
occur on a landing gear is ca1ledshimmy.It is typically certain conditions, due to the appearance of coupling
unstable oscillations that could reach such a level of between torsion and lateral flexion. These analyses have
I
instability that major landing gear structural components permitted the identification of parameters that influence
may be damaged or destroyed. the phenomenon, in particular aircraft speeds and loads,
and have validated the improvement achieved by
For many years, Messier-Dowty SA. has studied this increasing the stiffness of this landing gears torque link,
, phenomenon in order to perform more accurate to increase torsion frequency and move it away from
predictions of the dynamic behaviour of landing gear. lateral flexion frequency, thus reducing coupling
Several models have been developed taking into account tendancy.
many parameters, and especially the non-linear ones, in
agreement with test measurements.Simulations have been These analyses are still in progress in Messier-Dowty
performed, which show the sensitivity of shimmy stability SA., in order to take more precisely into account the
I
to variations of these parameters, and therefore the non-linearities of the landing gear and to improve the
importance of taking account non-linearities in shimmy knowledge of their effects, which will permit to anticipate
landing gear models. early in the design the landing gears shimmy stability.
This paper describes the methods developed in order to
1. INTRODUCTION build such a model, namely an analytical way to describe
I - @lane landing gear shimmy be defined as a
these nonlinear effects, as well as an experimental
determinationofitsnumericcoefficients,andsimulations
self-excited instability during take-off, landing or and conclusions.
taxiing, involving up to three vibration motions : angular
wheel motion about a vertical axis (yaw), angular motion
about a fore and aft axis (roll), and lateral displacement of 2. GENERALITIES ABOUT MODAL ANALYSIS
1
I the wheel. AND TESTS
II
The basic cause of shimmy is an energy transfer from the 2.1 Concept of eigenmode for linear models
contact force between the tires and the ground to the In theory, the notion of mode of a linear structure is
vibratory modes of the landing gear system. The vibration introduced from the equation of motion :
stability depends on the dynamic characteristics of the
gear, tires and fuselage, as well as the degree of coupling MX + CX + K X = F (1)
that exists between the modes of these components. At - X , X , X : vectors of displacements, velocities,
certain speeds, small motions may be unstable, and in accelerations
most severe cases, the pilot can not take corrective action, - M I C ,K :matrices of mass, damping and stiffness of
and failure of the gear structure may occur. the structure
-F :vector of excitation forces
One of the difficulties of a shimmy analysis is that real
landing gear systems exhibit many non-ljnear Eigenmodes of the structure are defined as the roots of
characteristics : swivel friction from the action of the equation (1) without righthaad term, which caracterizes
upper and lower bearings of the shock strut, steering collar the free structure :
friction,non-linear viscous damping from the steer-damp Mx + C X + K x = o (2)
unit, mechanical deadband associated with the strut lateral
motion, and slippage of the tires : some non-linear The determination of these modes constitutes an
characteristics are &pendent on landing gear wear. intermediate calculation to solve equation (l),whose
Nonlinear behaviour of landing gear makes the shimmy general solution is the sum of general solution (2) and of
phenomenon less predictable : if a test is performed to a particular solution dependent on F.

It is therefore necessary to evaluate and model landing as damped vibratory : because of damping influence, an
gear non-linearities in order to improve the pmiictability imaginary quantity appears, from which the denomination
of the shimmy phenomenon; an accurate shimmy analysis of complex modes. This damping term can be
must take into account a range of non-linearities values as suppressed in equation (2) to obtain the equation of the
wide as necessary, in order to reveal the most unstable conservating system associated :
configurations. Mx +m=o(3)
Paper presented at the 81st Meeting of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibration-Free Landing Gear, held in Banfi Canada from 4-5 October 1995, and published in R-800.
 9-2

Solutions of (3) correspond to pure sinusoidal motions, (s2 + 2 ti ui s + w2) qi = j

and are denominated real modes : all the nodes of the
struaure are in phase or opposite phase, whereas, for a After substituting s by (for frequency response) :
complex mode, vibration nodes move independentlyfrom
each other, and make the observation of motion more (0; -02 + j 2 ti U; w ) qi = j
difficult.
We can return to a matricial expression between q andf:
Equation (3), which govems real modes, has a solution of
the form : x = X e+, which leads to find eigenvalues q = diag ( [Wi2 -a2 + j 2 ti U; U ] - ) f = A f
of: ( - J M + K ) ~ = o We can express X as a function of excitation F :
- eigenvalues wi2 are the squares of mode pulsations ui X = # q = $ A f = $ A$T F =H F
-eigenvectors $; represent eigenshapes or mode deform
H matrix has the signification of a transfer matrix between
With the variable change : X = $ q (4) input F and output X,whose expression is : I
- $ matrice of eigenshapes @,- H(w)= $ A $T
i
- q generalized coordinates in modal basis
H(w)= # diag ( [wi2 - w2 + j 2 6 w; w ] - I ) $T I
we obtain, after pre+multiplicationby #T :
iPTM# q +$TC# q + # T K # q = #TF If HQ andhi are the i- term of matrices H and$, the I
product of above matrices, taking into account the
M and K are symetricmatrices, and modes are orthogonal
relating to them ;we obtain :
presence of a diagonal matrix, leads to : I
mq+cq+kq= #TF (3
- m :diagonal matrix of generalised masses mi
- R :diagonal matrix of generalised stiffnesseski with If the modes of the system are well spaced, thisexpression
= m i 2 m; evaluated at a natural frequency or will be dominated by
- c : damping matrix, symetric,non diagonal in general the term corresponding to that frequency, with an
I
I amplitude depending on its modal damping : I
We can assume c to be diagonal (terms e;)if the structure
is weakly damped (Basile hypothesis) ; we can define a
reduced damping 5;. ,without dimension, as:

= I 2 mj wi This expression in modal basis (after variable change)

leads to less voluminous calculations than a direct -
Equation (3clearly shows the parameters to determine method, because of the limited number of parameters : we
during a modal test : can choose the number of modes N to be retained, which
is the dimension of matrices m ,c and R.
- eigen pulsations w;
- eigenshapes $i 2.2 Modal tests classification in linear configuration
- generalised masses m; and generalised stifiesses R; We first assume that the structure is linear, or weakly
- damping coefficients of matrix c, assumed to be non-linear.
symetric
When those terms are known, equation (5) allows to There are two main categories of modal tests :
calculate q as a function of F,then equation (4) allows to - tests with appropriation of the excitation, which
calculate motion of struuure X sollicitated by forces F:X allow the isolation of each mode and the direct
is obtained by taking all modes into account. measurement of their characteristics.
- tests with transfer functions measurements, to
Equation (5) is mahicial, but can be expressed as N scalar sollicite a group of modes which are determined by
equations upon each eigenmode ; for the i-th mode : means of post-processingmeasurement results.
4 q; + q 4; + 4 qi = f;. , with the notation : f = $TF
2.2.1 Tests with appropriation of the excitation
Generalised masses mj can be normalised, and be chosen
to equal I : generalised stiffnesses 4 are therefore equal The structure is loaded by the means of a sinusoidal
to w,2 excitation sent to an excitator, whose level is tuned in
order to isolate each mode. The domain of excitation
frequencies is known by the means of a previous
frequency sweep to locate the interesting modes.
The theoretical e x p W o n of this method follows from
solving equation (1) with a sinusoidal sollicitatioa :
It can be expressed as Laplacian variables : W + C X + K X = F = ft@
 9-3

The solution is of the form :X = x e@, which leads to :

(K-uM+ j w c ) x =f
The advantage of this method, as opposed to the
Isolating an eigenmode by the means of an appropriate appropriation method, is the comparative simplicity of
excitation CoOSists of redkbg 0 P 6&k ,where 6&k is the experimental proceedings, which imposes few restraints
kth eigen pulsation, h m tbe excitation amplitudefi. This on choice of excitation; however, this method requires a
leadsto: ( K - o h ) k 2 M + j o h ) k C ) . q =ji more complex method to exploit measufed results, which
are based on a mathematical, rather than physical,
6&k and q are roots of the associate conservative system, approach, moreover, accelerometers must be located near
&US: (K-@0k2M)q =O mode antinodes, in order to achieve better measurement
precision.
We deduce the expression of loading which excites the
wok mode, fi, = j w k C Xk
3. LANDING GEAR NON-LINEAWIIES
With an appropriate excitation, the loading force THEORETICAL DESCRIPTION
compensates internal dissipation forces, and has a The shimmy analyses in progress in Messier-Dowry SA.
!phase with the response : the criterion for show that most important non-barities are friction and
appropriationis a phase criterion. For an isolated mode, all freeplay along torque links, with noll-lioear effects in
responses are in phase quadrature with excitation, which torsion motion. This chapter presents a theoretical
can be easily shown during testing by examination of description of those effects, and their experimental
Lissajouss curves. determination.
The advantage of this method is the ease of exploitation 3.1 Theoretical descriptioa of frictioo
I of results, which are based on a physical approach,
however, experimentalprocedure may be delicate : phase Coulomb friction or stickdip friction is a nonlinear
criteria for all measures may be difficult to adjust for some function of relative velocity between the two bodies in
modes. contact inside the landing gear. Such contact surfaces are
never completely smooth :when two peaks come into
- 22.2 Tests with transferfunctions measurements contact, the corresponding force is similar to a rigid

11 -
-
The structure is loaded on one or many excitation nodes.
Tbe input signal is sinusoidal,transient, or random, and is
mass-spring system, a funuion of position but not of
velocity; on the other hand, with no contact between such
peaks,slipping occurs, and this force depends only on the
rnulhfxquency : all the modes can be simultaneously velocity. A good model must take both aspects into
- excited. account :cf reference [l].
I

I The modal analysis is performed by aaos&r function Elementary representation of friction as a function of
measurement and processing : the theoretical explanation relative velocity alone is shown below.
of this method follows from solving equation (4) and (5)
with whatever load form :
X = # q (4)
m q + c q + k q = f = # T F (5)
According to 2.1,this can be explained as a transfer matrix
FnFl-
between input F and output X,whose terms are of the form 4
N
# 4%
~ i (0)
j = C,:-
r-1
0 2 +j x, U 0,
Standard stick-slipfriction modcl
There is a strong nonlinearity around &=O, with a
multiple evaluation of friction between -Fh and Fh : this
The term Hij(U) links the dynamic -LUX in the is unacceptable for numeric computations,which impose
accelerometer localizsd in d.0.f. i , to the excitation the linearizationof such a zone.
applied in d0.f. j .
When the transfer maaix between input and output is
measured, we have to find an analytical expression such
as mentioned above, which best fits the experimental
result (identification process), then to deduce eigen
pulsations q and damping coefficients tr . The
determination of #, t e r n can be ma& easier by
measuring responses on tfK same points as excitation :we
have i=j, thus :
9-4

3.2 Numerical simulation of friction

The friction nonlinear model must have a good physical Fex t(t )
interpretation of its parameters, as well as an easy
implementation in numerical software : a satisfactory e
representation for friction is the DUM model (1%8),
described in reference [2], in which friction force is the
solution of following first order differential equation : I f m is the inertia moment relative to torsion, described by
wheel r n m s with respect to vertical, we have a system
of 2 equations with 2 unknown values F and 8, with 2
valuesf, and Kj to identify :

Such a system is in fact a combination of a stif6ness and

a Coulomb friction, whose parameters are Kj and fo :
c(8) and 48) depend on t
k angular limit of freeplay j. I

Such behaviour can be described with a non-linear model

with torsion &play and friction, whose parameters are
in agreement with experience, taking into account that
stiffness and damping depend on torsion angle 8.
The significance of these parameters appears when
solving the above differential equation : we obtain an As far as I 8 I < j/2 , the system is in h e motion, with
hysteresis c w e F as a function of 9 : c(e)=k(e)=o, SO only friction force F OCCU~S.

For greater amplitudes, freeplay limits are reached, with

F
non-zm stiffness k and damping c, coming principally
ti Kf from torque link flexion characteristics.
- - ,i
3.4 Numerical simulation of nodinemities with
sweeping frequency. I

To characterize noblioear parameters, tests were

performed by Messier-Dowfy SA. with sinusoidal
Such a system behaves as stick-slip friction, with 2 excitations. Before describing these test proceedings, we
parameters Kj and fo : show a few theoretical results from the numerical
resolution of equations (6) and (7)with F-(t) = F sin wt.
- Kf is the systems apparent stifhess, which occurs whicb shows a physical interpretation of such tests when
during friction initiation, which is a sticking phase (9 0varies.
is small) ; the friction equation becomes - dF = Kf
de To make it easier, we solve separately two straightforward
around F=O: Kf is the zero point slope of above curve. 1 d.0.f. models,with friction only and with freeplay only.

- fo is the level of sliding friction ; when F= ffo,we 3.41 Nonlinearfriction d e l

obtain dF = 0 :friction force depends on 8 only, not
de
on 9,during this slipping phase.

This model requires identification of parameterfo and 4

from test results.

3 3 Numerical simulation of landing gear torsion

freeplay and friction

Freeplay nonlinearity is classically described as a variable A + & + A B +F = ~ s i n ~ t

stiffness,whicb poses no particular problem to model. m = 0.l ;c = OJ ;k 1 3 ;Fo = 2 ;kj= 10
According to above, we can model a nonlinear torsion link
of a landing gear with a straightforward 1 d.0.f. model : We obtain the following d t with this model :
 9-5

I
OO 0.1 1 1.6 2 28

Frequency rexponu of fieploy mods1for diflsrunt sxcitation

Frequency responses o f w t i o n modelfor diflerud e x a t d o n h l r (between 0,S and 100 Nm)
levels (between 0,S and lo0 N m )
This system also has two charaaeristic frequencies :
This system has two characteristic frequencies : - the first ow, associated with lower excitation levels,
- the first one,associated with lower excitation levels when the threshold freeplay is not reacbed Tbis .
(cf curve with 0,5 Nm), when the threshold force
before sliding is not reached: the friction model is in frequency is :f i = & $ ,therefore : fi = 0.87 HZ
fact a stiffness model, with value q., which is the zero - the second one, associated with higher excitation
point slope of the above curve. ( 6 3 3 ) .This frequency levels, when the Wshold heplay is reacbed. This

is:fi=kJG K +k
,therefore: fi=1.81Hz frequency is : fa = &&
K +k
,therefore : 4 = 1.81
HZ
- the second one, associated with higher excitation
levels (cf c w e with 100 Nm), when sliding can appear The system gets more stiff when excitation level grows;
; the stiffness Kj does not occur any more. This the frequency suddenly moves fromfi t o 6 when the
excitation level grows (4 > fi ) , which is often called the
frequencyis: fa = &&,therefore: A =0.87Hz "jmp"phenomenon
The system gets less stiff when excitation level grows: it This freeplay behaviour is not the same as friction
progressively slides between fr ami when excitation behaviour, whicb helps to characterize these two kinds of
level grows (12 fr ) . non-linearity during sweeping frequencies tests.

3.4.2 Nonlinearfreeplay model 4. MEASUREMENlS AND IDENTIFICATION OF

LANDING GEAR NON-LINEARITIES
4.1 Landing gear test descriptioa
The shimmy analysis, mentioned in the introduction, has
required several tests to measure dynamic characteristics
F sin (ut) of this main landing gear, in linear and noo-linear
i
domains.
A prelimhary test campaign was performed to verify the
closeness of the two frequencies set in evidence with the
initial con6guration, and their spreading out with the new
one (cf introduction); insofar as the order of size of such
frequencies was known, tests with appropriation were
We obtain the following result with this model : suitable.
9-6

A second test campaign was performed to obtain a more - tbe second zone is governed by freeplay : dqlacement
complete dynamic characterization of the landing gear, in increases a lot, which shows a strong non-linearity ;
order to validate the numerical model on many modes and friction influence is less sensitive : frequency drop noted
to make it more accurate. Tests with transfer function in first zone fades.
measurements were well-suited, to measure many - the third zone approaches linear behaviour :
frequencies initially unknown. displacement increases lineary, and frequency reaches an
asymptotic value.
4.2 Identification of non hearities with respect to These tests with appropriation have shown that,when the
tests with appropriation studied landing gear has this linear behaviour, torsional
frequency is about 16 Hz and lateral flexion frequency
4.2.1 Experimental characterization of landing gear with about 14 Hz, which are close values. This result will be
appropriation recalled further in non-liaear simulation analyses (cf 5.3)
Furthermore, this test permits the deduction of damping C
Messier-Dowty SA. has performed a modal analysis on a
clamped main landing gear with no contaa between
in linear behaviour from force and displacement
measurements,according to relation shown in 23.1 when
wheels and ground, by means of appropriation of the
thewkmodeisexcited: = j q ~ k Ca
excitation on a wheel center, with 10 levels of excitation
force. Landing gear experimental curves show a 4.2.2 Friction parameter identification
non-linear behaviour in torsion, with dependance of
frequencies and displacement on excitation amplitude :in In addition to linear characterizationof landing gear, this
particular, for low values of excitation, frequency kind of test permits the identification of n o n - b a r
decreases, but reaches an asymptote for higher values of parameters to be introduced in a numeric model to fit the
excitation, for which the phenomenon becomes three zones previously identified.
quasi-linear. In panicular, we have to determine& and Kj values so that
the solution of above equations(6)and (7)(cf 33) fits tests
measurements with !iquency sollicitation : one
possibility is to solve this equation analytically and
Kf values , and looping on those 2 values to miaimize the
I gap between numerical solution and test result :this leads
I to relatively long computations. - -
i 1st ZOlK : A better way consists of finding a straightforward .
I approximated solution of these equations, insofar as the
e x t e d loading is sinusoidal : the assumption is that - -
response will be h o m c too, with the same pulsation :
this is the "harmonic linearization method", or "Rib
averaging criterion" : cf references [3] and [8].
If Fd (t) = FuI sin ut,we have F and 8 in the form :
F = Fccosot+Fssinot
e = u c c o s u t + ussincut

m i j + c ( e @ + + e p + ~ - ~ , s i n ~= t o (7)
These expressions of F and 8 are only approximate

displacement J,' I ! _. solutions, so their substitution in equations (a) and (7)

i leads to non-zero values. However, according to R i b
/I I I
criterion, weighting factors U,, U,, F, and Fs can be
determined in such a manner that,over the period 2 IL / o,
the weighted averages of those non-zero values vanish,
Approprbfion test :tonion ampfitudes andficqucnciea
with the same weight functions cos ut and sin ut as
defined in F and 8 expressions;if y = Kj lfo2,and with the
Such a curve shows 3 zones, successively reached for above expression of F and 8, we have the conditions:
increasing excitation values, whose physical
interpretation is :
- the first zone is governed by friction : low excitation
levels do not allow the system to reach freeplay thresholds
: it behaves like a 1 d0.f. oscillator with friction, which ($
produces a frequency drop when the excitation level
increases (cf nonlinear fiction model : fi >h)
 - - c

9-7

Such an improvement will be tested io futon shimmy

analysis.

(% 42
.2
.4 Limitations of tests with appropriztion
Despite the relative ensiness of these test proadure. many
limitations exist :
-in 0lde.r to reach asymptotic Values forfrequencyand
These 4 integdons can be computed numerically by damping, &on level most be sufficiently
s u m F and 0 by theii appmrimw cxpnssions : high, all the more as the landing gear has a more
they cancel out for @mi& v a l w of weighting factom n 0 n - h behavioolr : as long as frequencies depend
U,,U ~ , Fand~.,wttichdcfine~andesol~tio~.
, on excitationkvel, hear behsviour is not reached
- if excitation level is not sufficimt.identification of
4.2.3 Erample of experimentalfiningof nonlinear frictionparameters becomes less auauate.
landing gear model -duringthistestcampaigu,somemodeswerediffidt
to appropriate : their meammd fquencies atd
Theexperimeatallandinggeartorsionikquencyrespoase damping wuc not PCCOVsteenough.
m e is shown as a solid line: this frequency varies
between 16Hz (in the linear mne) and 24 Hz Besides, the identi6cation of fdction parameters shown
above may not be accmk enough when coupling rppears
For ea& &on kvel, friction parameters weLe between modes if coarespondhg friaion is not
adjusted to fit the experimental m e : stars indicate the neglipabk,iasofarasitshypomesisisthebehavimlike
frequencylevel readwd from W e values. a single 1 &of. model,llILcollpledwith other d0.f.

28
I
4.3.1 Experimental characterization of non-lineanties
with banfer%tion measmments

To reduce the limitatiws mentioned above,

Mexsie4mty SA. p f a n n e d a secord modal teat
campaign on the same landing .gear with trpnsfer function

'b
I&m&n
.b M M . A &-a
P ~ ( W
3- In0

of dk. *qmimnki t o n i o n j y u r r y napow

mcasurrmenb, io order to obtain mom eigenfrsqueacies
(~1O)in~flee~canditiWgThclaDding
par was instnunented with 19 accelerometers, whose
localization allowed the measmemm of modes beywd
the first One.

Experimental hquency evolution is well mpcted,

especially amufd the asympotic frequency for higher
force levek the mow importaat gap betweem the
experimental m e and computed points is around the
m e d i a n m , seen above as the mownooliaearune of the
system, which makes it more difEcult to modcl precisely. qy t
It is possible to improve the non-bear behaviour of this
model,for example by dehiag forcefo as an expomntid
fuoctiwof angular velocity :its maximum absolute valw
is mater thao dipping value fo. , which represents
&Y +>
maximum stiddng before sliding.

e'
9-8

Insteadofusingafrequencyrrsponaefrmction,liLemany
freqnency drmain algorithms, this method ia formulated
fromthe impulse functioo (due to Dirac impulse), which
may tc obtained fmm the invemeFwriertnwform of the
Sepoency nsponse fuuuiow the tbeonticalideotification
, expesedin the
process is the same 88 ~ w n i n 2 . 2 . 2but
time domain. One of the advaatagea of this methodis the
possibility to take many excitation mums mon easily
into account.
Transfer function meanurmmts are today in progress, so
we have not yet 0-d all thart ChSnaLnsa' 'cs. We
sbow below an example of mch identiEcarioIb in
agmmeat with the first 3 modes of an experimentalFRP

Fnqwncy mxponurfor Bflsmnt8Xddrllon b 8 k

The following elements can be noticed fnnn this
experimentalnault : ,.
I e.

- the fust frequency peak is independent of excitation , I ..__

_.._
level, bot its amplitode valw, which charachxizm . *
. I

damping value (cf Hij (U,)expression in 2.1). depends *\ I

'.
on this excitation level, which shows a nonlinear
!I I
tchaViOUC

-thesecondfrequencypeakdependsontheexcitation
level, with a "jump" between two frequencies, whic4
also shows a noolioear behaviour, identical to the
freeplay 1 d0.f. model seen in 3.42.
-
Exprlnunt.IdidenU&dFRFVUL3&atndr.
As shown in 43.1. wch identificationmust bc made from
other tests on this system showednodhearbehaviour: with &&rmt &tation and diffaent
-wbenexcitingthesystemwithasinglefkxpency sweeping dinctions, asspr pry for &te&g
equal to this first peak, the response showed harmonic nO D - h W i t i G S .
cornponeMS of that frequency, which should mt
appiarina liaearsystan. 5. SAIMMY NUI)IERICALMODEL
- when exciting the system with a swept sine, some
peak frequencies depend on the direxion of sweeping 5.1 shtnmy lndd hypothesis
(haeasing or decreesing she),which shows -lay
,,* beha,+,u, with a ,*jump**pbenarmenOn 5.1.1 Degrees offncdom of shimmy mcdel
between the= frequencyvalues. A wmvlete shimmy numerical ' modelhasbeendeveloved
from +Vious &chnncteristics, including l&ar
43.2 Identification of non-lineanties fmm transfer
function measurements and non-linear behavioor of the main landing gear, in
a p m e n t with test measmments mentioned above.
The identitication of mn-lhdties consists of linding a This model takes into account 3 d0.f. of the landing gear,
model whose FRP (Fkqnency Response Fundom) agree located on the center of wheel axle :lateral displaQmeM
with all the experimental F W for each accelerometer : Uy,rotation Mx about ox longhdinal axis, whicb
this identitication is 0btaine.d with qe&c software, ~ple~en t laanl aexion, and rotation M. h u t m
both
integratedintothemeamwnentchaia vertical axis, which rrpresenrs tomion

Them are several ways to examine the measured This model must take into Dccount the following
frequency respome function to extract these data, which elemem:
are descrited in refezence 141. The numerical method used - s k l l d descriprion of landing gear
hen is called 'Tolyrefemce llme Domain Technique " -gymwpicefkcm
(cf nfezence [SI), and allows the extraction of modal - dylurmic tire
parameters, namely modal damping and Wuencies, -landing gear non-linearitiea
fromexperimentaldata. -appliedexciratioll
 9-9

5.1.2 Structural description of landing gear This nowlinear system is solved iteratively. with a
dimetisation of the derived temporal terms (Newmark
Landing gear s t r u a m is taken into account with : schema) and 6iction f o m (Crank-Nicholsonschma).
- 2 manices 3x3 [M]and [K]obtainedfmm a dynamic The solution depends on airnaft velocity, which appears
condeosation (Guyan method) on the center of wheel
axle of the landing gear finite element model, whose in the gyroscopic metrix G and in the fn?quential tire
model P(s).
calculated modes agne with modal tests with
appropriation when linear behaviour is reached (cf
53 Readtn of model compntation
43.1)
- a damping matrix 3x3 [C] , directly built fmm From the previous temporal equation, we can obtain
m u w n d damping coefficients with the same modal torsion respooses of landing gear, for different non-linear
tests (cf 4.2.1) parameter values and airplane velocities :
5.1.3 Description of gyroscopic effects -torsional freeplay valuea :correspond to a madation
freeplay measured of a staadard point of measure,
We must take into Bccount gyroscopic effects due to whose valws are 0 , 0 . 5 , 1 , 1 5 mm.
mlling wheels of the gear,inthe form of a3x3 matrix [C]
which contains a term P V I R :Jis the moment of inenia - torsional fiction values : maximum forcefo values
for the system wheel + tyre, R is the tire radius, Vis the ive 0,25 ,50 Nm, with a constant friction slope Kj in
longibldioal airplane velocity. agreement with modal tests (cf4.2.2).
- airplane velocity values :45,55,65,75,85 mls.
5.1.4 Description of dynamic tire
The torsional damping can be deduced from logarithmic
A dynamic tire frequencial model was conveyed by its deament of temporal torsionresponse.
manufacturer, in the form :
The results can be summarized in diagrams dlowing
Fpnru(s)= P(s) U 0 = P(s) /UyCs)MAS) Mz(sJ1 torsional damping v e m airplane velocity :each diagram
P(s) is a 3x3 transfer matrix (s is Laplace variable), made corresponds to one freeplay value, with the 3 fiaion
up with 9 elementary transfer functions, whose terms A
values : x : ONm ; : 25 Nm ; :50Nm+
agree with manufactum dynamic tire tests, and depend 8,
explicitly on airplane velocity.
5.1.5 Landing gear non-linearities
Experimental results show more frequency variations in
torsion tests than in flexion : comequently landing gear
non-linearities, which are principdly fiiaicm and
&play, are only locaked in torsional d.0.f. MI, with the
expression described in 3.3 :cfequatiom (6)and (7) . -2O B 3 3 3 3
5.1.6 Applied excitation 40 50 60 70 80 90
Fn.lthyOnun
An excitation in moment along the oz axis is introduced
in the model,to load the landing gear in torsion :this load I I I I I
will be applied in fume shimmy tests.

5 3 Shimmy model numerical m l n t i o n

This non-linear shimmy model must be solved with a
temporal analysis. Insofar as frequential analysis with a
Laplace variable is limited to linear models, we must
express the tire model in temporal variable, which -21 I I I I I
nguires the inverse Laplace transformation of each of the 40 50 60 70 80 90
niae aansfer functions of the above transfer matrix P(s): Fneplay 0 3 mm
this leads to a temporal expression Fpne.(t).
The general form of the 3 d.0.f. shimmy model is :
M(&J U + (C+G) U + &zJ U + F p d O +
[O 0 F j j ~ ] = F u r , with U = [UyM. MI]
2
and K&
M(&J dependon because of torsional freeplay.
0
The scalar form of Ffiz comes from equation (6) : -7I
40 50 60 70 80 90
FN&IUWI
9-10

10 friction behaves like UI dditional e$,so the

torsional kquency incrrpses c m p d v e l y to its
8 quasi-lioear d u e .
6 - whm heeplay incnases, the excitation level
4 becomes iosufficient to reach the fleeplay thrcsholda:
this system gets less wl& 80 thc tolsiwpl flquelxy
2 decnsses comparatively to its quui-linear ralue.
0 As shown in 4.2.1, the studied landing gear, wbcn
-2 behaving in a qud-lkar m m m ,han dose kqucncks
40 50 60 70 80 90 in torsion (16Hz) and in llanl flexiw (14Hz),wbicb
Fw+y 13 mm may ccau couplins brmem t h e 2 m o b W l m
wll-lkar friuion iwueea, the t o r s i d frqllelxy
Tbc results W b e also SummPriEedindiagnmsshowing iocresses too. so the 2 Wquenciw move aside : the
todmd damping vems abplam velocity :each diagmn coupling deuepres and this knding gear gets more stabk.
cOmspoDds to ollc friuion value 0 or23 Nm, w i U ~the 3 Inversely, wbcn wo-liocar fleeplay hcmtxs, the
fleeplay values : torsional fIequency decrrarws, and becanes doser to the
x :freeplay value 0 m m A :*lay value 0.5 mm. lateral flexion kquency, whicb lmdas thir larding gear
+ :bephyvdue1oun +:freeplayvalue 13mm. more unstabk :so thislading geargeta more wrbk w h
noIl-liIlear *lay defnnsts.

This qualitative influenceof oa~-baritieagets less dear

when both heeplay and hiction become imporun, U
shown in thc dingnm for 6nepl.y 15 mn :inpucicular,
for a friction value of 25 Nm. the system kcows
alternatively stable or UDStpbk when aircnA velocity
incnases. This d t is slso shown in the f O U 0 ~
temporal curas forvelocities values of 65.75 and 85 m/s
-21 I I 1 1 I
40 50 60 70 80 90
Fffctfon0

8, I I I I I

6
4
2
0
-2-
40 50 60 70 80 90
F&Um 25

5.4 Influence d non-hedtiea on stab*

These diagrams show mu the linearid landing gear om-
(fleeplay and friction equal to zero) appears to be stable,
but wn-limuities lead to ZCIO damping, wahicb
correspolds to limit cycled, and e v m to negative
damping. which comspada to divergena in toIsioQ
hetlce tk impapDce of taking wn-heaaes in0
account in shimmy models.
Ingened. this lauding geargetsmore stabk wbmfriaion
inueascs. and when heeplay decrenaes: this can be
e x p l a i d h the BmuJ d t s of t k 1 do.
nop-linear model shown in 3 .4.1 and 3.4.2 , md h m
flexion andmaion bquency v a l w :
-when friuion ilxWse8,thc excitationlevel beconlea
inslImCient to reach tk limit fowfo before sliding:
 9-11

o.O(r . Some mexk

J/!
1
,....... ' ........................... 1 embe m to obtain, W.
iastext.
testa resulta, nca-linearity
values to be intmduced into a shimmy model :however,
thsc values are not indmic to me landing gear, iosofsr
as thcy are highly dcpcDdaa on its wear.
Forthis reason. it is important to coltirme shimmy sludies
by exploring non-linear parameters b i d e mcir p i b l e
range of vadations, in older to de& thc most
unfavourable coafisurationsin terms of ddmmy stability.

:. . .. , ... ,
.....
Tbis is cllncntly in pm- at Mem'er4bvly SA., with
amoreacauatenodinearchnnaarzstton * . of the llllldied
main Larding gear hnn teats with traosser fuaaion
meammeas, and witb the development of a complete
Y- lllmmld' model,insgreementwithmodaltan
d t a .
such unpredi- caused by mn-lioearities call also
be fouad in thco~tical~ramples,such as thc "D@g Fbhm ddmmy tern on a flywheel will enable the
spring", w h behaviour is highly d t i v e to initial validation of thb non-linear numeric model, cmd thc teats
conditions, so far as to become chaotic, in spite of a of Consprations able to improve landing gear stability.
comparstivelysimple behaviouraleapressim :
Tbisimpmvedmodclizatiwofthclnndinggcarwillallow
x*+o.asx"+=7.z5c~t to aaalyse ita dynamic behaviour in its devdopment. At
4
I this stage, it wiU be therefom convenient to delhe and
qurlifypoasible~~mtofthede~ifnecessrryto
a c b i ~the rrquestcd dynamic stability of me landing
gear in its range of operaiim colrdittons.

7. REFERENCES
[l] Computer Simulation of stick-slip friction in
Mechanical Dynamic Systems- D.Kamopp - Joumal of
Dynamic Systems, Mwsurements and Conml - Marcb
1985. VoL 107
[2] On the modelling and simulation of HcIim - D.A.
Hnessing, Jr. B.Riedlrod- Joumal of Dynamic Systems,
i -4
a S l O l s 2 0 a o S O I O * m
Measmementa and coonol- TMMctl'om of the ASME
-September 1991,W.113

- -
D u m # xplinr reax&iv@ to Idd*leodiUowx
x & h :X(0) -3 X'(0) - 4
l :x(o) 3.01 xyo) 4.01
h r ~ .linr
[3] Vi- -J ~ c o ~ S ~AI m~ ,
[4] Modal testiDg :theory and praCtire -DJ.E h -Jdm
Mo~eover,such chaotic behaviour has been expressed in w*y, 1984
pmrious landing gear nonlinear studies, in particular in
refe- [61 pamgraph 4. [SI The polydea'em b e domain teclmique - F.
Deblauwe, D.L.Bmwn, RJ. Allemang - MAC 1987 p.
However.it is imponam to notice rhat this qualitative 832,845
mente of mn-linearities CMOot apply to all landing
gem, insofar as toIsiw and latael fluioll iiquencies [6l Modeling and Analysis of a Dual-Whel Nostgear :
canbe more or less dose to each other; foriastance, if Shimmy Iostrbility and hnpaa Motions - G. Li
Sreqoenciesin linear behaviour are 10 Hz in t d o n and 15
Hz io lateral flexion, increasing fidm haeases tasim m Fl5 Nose Landing Gear Shimmy Taxi Test and
flqnency, which gas closer to lateral fquency: the Conelptive Adysea - D.GnwMolprm
system may bewme more uaatable, in opposition to this
studiedlanding gear. [SI The behDvioor Of N o I I - ~ ~Wbrptioa S m W. -
SzcmpIiasLesntpnca ' - K l m A ~ & d ~ h b b &

6. CONCLUSIONS 191Bngincerios V & d W- D.Inmao- Red= Hall

calcularim d t a shown above prove the importance of [lo] L ' d de vibration des shllc~msimparfaiternea
non4nearities io shimmy pknomm, and their Lidaim - R. Dal - Revwe praaFpiSe de Mccsrdgue no
~uenceon~ggear.uabiity. 1987-3
 10-1

A Nonlinear Model for Landing Gear Shimmy with

Applications to the McDonnell Douglas F/A-lSA

Jeff Baumann
Senior Engineer, Structural Dynamics and Loads
McDonnell Douglas Aerospace
P.O. Box 516
St. Louis, MO 63166-0516
USA

-
SUMMARY Fo = force the strut must transmit to the wheel at
This paper discusses a nonlinear shimmy model that point 0 for dynamic equilibrium, lb
includes components and features typically found in
the nose landing gear of fighter aircraft. The gear
FA = force of the strut on the effective rigid body at
structural model includes inertia, coulomb and viscous pointA, 16
damping, stiffness, and torsional freeplay. Single and FL = resultant force in the lower piston bearing,
dual non-corotating wheels are considered. The tires lb
are represented with the Moreland point-contact tire FMS = foreign military sales
model. FP = freeplay, rad (deg)
Qualitative analytical results from modeling the nose Fu = resultant force in the upper piston bearing,
, landing gear of the McDonnell Douglas FIA-18A are lb
- presented. This landing gear, in the normal US Navy
H = overall length of the strut, in
carrier-based configuration, is stable. When the
- launch bar and its power unit are removed for land- E = angular momentum, in lb sec
based versions of the aircraft, the nose gear may, [Z]bod,, = effective inertia tensor for the rotating lower
- - under certain adverse conditions, experience shimmy.
This condition can be aggravated by rotating forces strut, inlbsec 2
due to tire unbalance or other irregularities. A number = diametral moment of inertia of the wheel,
of corrective modification have been attempted with ZD
I varying degrees of success. These range from in lb sec2
relatively simple measures, such as increasing =polar moment of inertia of the wheel,
attention to tire maintenance, to more complicated
Zp
schemes, such as adding a friction collar, adding mass in lb sec2
dampers or even adding a dummy launch bar. For the JCOLuR = colladlaunch bar moment of inertia,
most severe cases, the dummy launch bar seems to be
required, however, the friction collar with mass in 16 sec2
dampers does show a significant improvement.
KO = tire lateral stiffness, lb / in
KSD = steering damper torsional stiffness, in lb
NOMENCLATURE KTL =torque link torsional stiffness, in lb
- Kw = tire radial stiffness, lb / in
A = acceleration, in / sec2
L = trailing arm length, in
C = tire coefficient of yaw, 1/ lb
Lp = length of the piston, in
Ci = tire yaw time constant, sec
ii?, = moment the strut must transmit to the wheel at
CD = tire lateral damping, lb sec/ in
point 0 for dynamic equilibrium, in 16
CPC = Cleveland Pneumatic Company
= exponential steering damper coefficient,
C ~ D @,, = moment of the strut on the effective rigid
body point A, in 16
in lb secMSD
MDA = McDonnell Douglas Aerospace
CTL = torque link structural damping, in lb sec
me = effective mass of the strut, lb sec2 / in
D = distance between piston bearings, in
DEAD = deadband function MR =rotating unbalance, lb sec2

Paper presented at the 81st Meeting of the AGARD SMP Panel on The Design, Qualification and Maintenance
of Vibration-Free Landing Gear, held in Ban& Canada from 4-5 October 1995, and published in R-800.
10-2

MSD = steering damper exponent

-
o = angular velocity, rad I sec
2
mW =mass of the wheel, lb sec I in
R = wheel radius, in
Rsr = static wheel radius, in 1. INTRODUCTION
Z = position vector, in As efforts are made to increase the performance and to
= lower piston bearing radius, in decrease the weight of modern aircraft landing gears,
RL shimmy problems often result. Much effort has gone
RRoLL = tire rolling radius,
in into predicting the shimmy stability of landing gear
& = upper piston bearing radius, in designs and in studying ways to stabilize shimmy
S = position(s) of the wheel(s) on the axle, in prone gears.
sgn = signum function One of the most important factors in the shimmy
t =time, sec model is the modeling of the tires. Two main theories
T, = spring torque on steering damper, in lb have been used to describe the interaction between the
tire and the ground. The stretched string tire model,
T2 = required torque in exponential damper, in lb proposed by vonschlippe and Dietrich [ 13 analyzes
TA = torque of torque Magelsteering damper at the tire as if it were a taut string on an elastic
point A, inlb foundation. The point-contact tire model was
proposed by Moreland [2] In this model, the effects of
Tcp = total Coulomb friction torque at piston, in lb
the ground on the tire are assumed to act at a single
TpcoN = Coulomb friction torque at piston (constant), idealized contact point. The lateral displacement of the
in lb contact point and the tire drift angle are related by the
TSD = steering damper Coulomb friction torque,
model to the lateral restoring force and the vertical
restoring moment. Many researchers have since
in lb examined and compared these models. Collins [3], for
TsF = Coulomb seal friction, in lb example, finds that the point contact tire model is
1; = unit vector aligned with the axle adequate for qualitative and, in most cases,
quantitative studies. Also, Black [4] has shown that
U, = position of the CG, 5 component, in the point contact tire model can be derived from the
USN = United States Navy stretched-string tire model. Since the point-contact tire
V = velocity, in I sec model is much easier to implement and the required
parameters are better understood, it is used for this
V, = position of the CG, j3 component, in
work.
VTaI = taxi speed, in I sec (knots)
Since some of the original analysis and verification of
W = aircraft weight on the strut, lb
the analysis program was related to the nose landing
WG = position of the CG, component, in gear of the McDonnell Douglas F-15 it is worth
X = lateral strut tip deflection, in special mention at this time.
a = strut tip bending angle, rad (deg) The original nonlinear analysis of the F-15 nose gear
-
a = angular acceleration, rad I sec2 was performed by Cleveland Pneumatic Company
(CPC), the gears manufacturer. Southerland [51
P = position of rotating unbalance, rad (deg)
describes the original analysis and computer
Y = constant strut caster angle, rad (deg) simulation of the gear.
A = lateral tire deflection, in
Both this early analysis and subsequent dynamometer
e = steering angle, rad (deg)
testing suggested that no problems existed with the
ec = collar angle, rad (deg) design. As aircrafi were put into service and began to
OD = damper angle, rad (deg) wear, however, shimmy problems developed. These
h = tire camber angle, rad (deg) instabilities were traced to the presence of larger than
anticipated values for the torsional freeplay. When
P = coefficient of friction at piston bearings
shims were added to reduce this freeplay, the shimmy
PR = coefficient of rolling resistance disappeared.
PT = tire moment coefficient, in lb
Grossman [6] describes experimental and analytical
0 = angle between the coordinate system at the shimmy tests performed as part of the modification to
strut tip and the one at the tire, measured the design of the F-15. In the first phase of the study, a
about the axle, rad series of taxi tests were performed with a specially
@ = tire yaw angle, rad (deg) instrumented aircraft. Vibrations were measured as the
aircraft was taxied at a variety of speeds and
w = tire drift angle, rad (deg)
conditions.
R = wheel spin rate, rad I sec
 10-3

UPPER PISTON BEAR

STEERING MOTOR /

LOWER PISTON BEARIN

Fig. 1a. Model Geometry for Cantilevered Landing Gear Fig. lb. Cantilevered Landing Gear

The second phase of Grossmans study involved information as the simulation ran. This made the
repeating the CPC analysis with more realistic values simulation modular in that new tire or structural
for the parameters. An attempt was also made to models could be added later.
linearize the nonlinear terms in the equations.
Since the details of this analytical model have been
The analysis correlated well with the shimmy speed - presented elsewhere (Baumann, Barker and Koval[7]
freeplay relations found in the taxi tests. Absolute and Baumann 181, a more general description will
stability was obtained for torsional freeplay less than follow. Table 1 lists the degrees of freedodstate
0.75 degrees. Also, the linearized models were shown variables used for this model.
to agree with the nonlinear model for the range of
parameters tested. 2.1. Kinematics
2.1.1. Coordinate systems

2. NONLINEAR LANDING GEAR MODEL The inertial coordinate system XoYoZo is attached to
the runway (see Fig. la). The Zo axis point upwards,
The objective of this work was to develop software
that would aid landing gear engineers in assessing the the Yo axis is aligned with the runway axis and points
shimmy stability of a more general class of landing aft, the X o axis completes the right-handed triad and
gear designs. With this analysis capability, potential points to the port side of the aircraft.
problems could be identified early in the design
process. It was also desired to be able to predict how A moving X I Y,Zl is fixed to the top of the aircraft
changes to existing design would affect the shimmy strut. It is rotated away from the XoYoZo about the
stability.
negative X o axis by the constant caster angle y .
Special care was taken to realistically model the
torsional components that are typically found in The X z Y 2 Z 2 system is fixed to the bottom of the strut
fighter aircraft nose landing gears. Three different and accounts for bending of the strut. It is rotated
landing gear geometries were studied: cantilevered, away from the XIY,Zl about the Y, axis by the
fully-levered and semilevered. Only the cantilevered
variable strut bending anglea .
type, as pictured in Fig. 1, will be discussed here.
The strategy employed was to analyze the landing gear The X3Y3Z3 system is fmed to the bottom of the strut
structure and the tire separately from each other. The and accounts for rotations of the strut. It is rotated
structural model and tire model were programmed as away from the X z Y 2 Z , about the Zz axis by the
separate computer subroutines that would share variable steering angle0 .
 10-4

TABLE 1 - Degrees of Freedom The motion of this point will also be subject-to the
variable strut bending angle, a ,and steering angle,8 .
Strut Model
The center of mass of the effective rigid body, Point
x = lateral strut tip deflection, in
CG, of the effective rigid body which represents the
dx
x =- swiveling mass of the landing gear is located with
dt respect to point A by the arbitrary vector
a = strut tip bending angle, rad (deg)
da
a =-
dt Thus, motion of this point will also be subjected to the
8 = steering angle, rad (deg) three flexible degrees of freedom described above.
e =-
de
dt
8c = collar angle, rad (deg) 2.2. Tire Effects
Based on a complete kinematic description of the axle,
a separate tire model is used to determine the force
= damper angle, rad (deg) and moment that the strut must exert on the tire(s) to
OD
maintain dynamic equilibrium. This force may
include suspension, traction, rotating unbalance, and
Tire Model inertia effects. One such tire model is described in
p = position of rotating unbalance, rad (deg) detail in Section 3.
A = lateral tire deflection, in
For convenience, it is assumed that the force and
w = tire drift angle, rad (deg)
moment act at point 0, no matter where the tire is
mounted on the axle. Any transformations that are
away from the X 2 Y2Z, about the Z2 axis by the required are included in the tire model.
variable steering angle8 . The force and the moment will generally be of the
form
2.1.2. Rotational Kinematics
The angular velocity of the bottom of the strut is equal ig = FO(XO,E3)
to the angular velocity of the X3Y323 coordinate
system. This angular velocity accounts for the strut
Po = Po(Xo,,oL3)
bending rotation a plus changes in the steering angle which requires that previously defined expressions for
8 and is given by
To and E3 be substituted into the moment
z3= (asine >5+ (a case )j3+hi3. expressions.

The corresponding angular acceleration is

-
a 3 = ( g s i n +ci6cose)i3
~ +(ticose+ci6sine)j3 +hi3
2.3 Strut Forces
2.3.1. Elastic Restoring Forces and Moments.
The restoring force ( Fs ) and moment ( Ms ) due to
2.1.3. Translational Kinematics
the bending ( x , a ) of spring element AB is given by a
The position, velocity and acceleration of four key 2x2 stiffness matrix and a 2x2 structural damping
points on the strut are needed for this analysis.
Reference is made to Fig. 1 a. &
matrix. This force acts at point A in the ( ) direction.
The moment also acts at point A about the (- j 3 ) axis.
Point B is at the intersection of the strut axis and the
trunnion axis. This point is assumed to remain a
constant height above the runway and have a constant
taxispeed VTm. 2.3.2. Shock Strut Bearing Forces
Considering the forces and moments acting on the
Point A, the reference point at the bottom of the strut
pistonhpring at Point A (due to bending and the
is separated from point B by the strut of length
weight of the aircraft on the nose gear) statics can be
H which is held constant over the simulation. Also,
used to determine the forces in the upper and lower
the strut may bend laterally, with the tip (Point A)
piston bearings ( F, and FL , respectively). 'Ihe
deflecting a variable distance x .
resulting torque exerted on the piston (and hence on
Point 0, at the intersection of the strut and the axle, is
separated from Point A by the trailing arm length L .
 -
10-5

the lower strut) is

{
~ c =p ~pcolv+ P ~ + RLFL ]}sgn(8 )
[ R~u

where TpcoN is an additional friction torque arising

from Coulomb friction between the piston and
cylinder walls (independentof loading at point A).
'1
2.4 Dynamic Analysis

At point A, the force FA and the moment

ji?, represent the force and moment of the strut on the
rigid body. At point 0, the force Fo and the moment
Eorepresent the force and moment required by the I TA T
wheels for static equilibrium. I LOWER STRUT

TIRE MODEL
2.4.1. Translational Dynamics
From Newton's second law applied to the strut rigid I GROUND
body,
-
G = G + m , ACG Fig. 2. Model for Torques
where m, is the effective translational mass of the
strut. any torsional model. The modeling of the tire is
discussed in Section 3.
The lower shock strut can translate along the strut axis
2.4.2. Rotational Dynamics
with motion restricted by oleo-pneumatic forces
From Newton's second law for rigid body rotations within the strut. Coulomb friction will be present in
about the mass center of the strut rigid body, the upper and lower piston bearings. This friction will
oppose translation as well as rotation of the lower
shock strut. Gross rotations of the lower shock strut is
prevented with internal splines or with an external
torque linkage, as shown in Fig. lb.
The upper arm of the torque linkage is connected to
the steering collar. For carrier based aircraft, such as
Solving for ZA, the F/A-l8A described below, a launch bar would also
be attached to this steering collar and its effects would
have to be included in the torque model.
The steering collar is driven by a steering motor.
When not actively steering, this motor will serve as a
shimmy damper. This damper is modeled with a
spring, a Coulomb damper and an exponential
damper. The exponential damper will generate a
damping torque proportional to the relative angular
velocity across the damper raised to some exponent. If
2.4.3. Torque Models this exponent were 1.O, the damper would be a linear
Since shimmy involves rotations about the strut axis viscous damper. If this exponent were 2.0, the damper
and since most corrective measures involve torsional would be a "velocity-squared'' damper.
components (often torsional dampers), a good The connection between the steering collar and the
torsional model is very important. Figure 2 shows how damper will often have some freeplay. This freeplay,
the system of Fig. l b is modeled. caused by gear backlash and clearance between mating
parts, will allow some relative motion across the
Starting at the bottom of Fig. lb, the interaction
connection between the collar and the damper.
between the tire and the ground plays a large role in
10-6

A total of three degrees of freedom are required to set to zero, 8 is no longer considered to be a degree
describe the full torsional model described above. of freedom and one of the other torque models would
These degrees of freedom are the collar angle 8 , its be used.
derivative6 , and the damper angle 8 . The torque Applying Newtons second law of motion for rotations
model will contain the derivatives of these three to the collar yields
quantities as well as the torque TA which the above
components will apply to the lower strut rigid body.
Since some of these components may not be present or
may have only minimal effect, provisions are made to
remove one or more of the degrees of freedom from JCOLZAR
the model. Six separate cases involving various
Note that this equation requires that collar rotary
combinations of degrees of freedom were analyzed.
inertia be present ( j C O L L A#Ro ). If the rotary inertia
Each of these special cases are treated separately in
numerical computations. Only the first of these cases, is zero, 8 is not considered to be a degree of freedom
the full model as pictured in Figs. l b and 2, is and again one of the torque models would be used.
discussed here.
The final state variable derivative
It is first necessary to determine if the steering damper
is engaged to the collar. This depends on the absolute
difference between the collar angular position, 8 and
the damper angle, 8 D . is already defined since it is also a state variable.

a)Steering damper disengaged:lqc - qDl< FP. b) Steering damper engaged: le -8 Dl 2 FP.

The torque of the strut on the wheel, TA, is As before, the torque of the strut on the wheel, TA , is
transmitted through the torque links, so that transmitted through the torque links and is given by
. . TA =KTL(e-ec)+cTL(6-ec)
TA = KTL(e -8 c ) + C T L (-8
~ c)
If no structural damping is present in the torque links, If no structural damping is present in the torque links,
CTL can be set to zero. CTL can be set to zero.

Summing torques in the disengaged steering damper Because the steering damper is connected to the collar,
requires

Since 6 is a state variable, this is already defined.

This can be solved for 6 by fust defining
Applying Newtons second law of motion for rotations
= -KsD8 D to the collar,

and 0, = $ec)
T2 =ITrI-Tso.
If T2 is greater than zero, then the spring torque (q )
will overcome the Coulomb friction and need to be
reacted by the exponential damper. If T2 is less than
-cSDp C y sgn(6 C)] JCOLLAR
zero, the Coulomb friction alone will be sufficient to
react the spring torque. Summarizing:
Again notice that JCoLZAR must be nonzero for this
for T2 I O model to apply. This expression uses the deadband
function to account for freeplay. This function is given
by

Note that the latter expression requires a nonzero

steering damper coefficient, CsD . If this coefficient is
DUD(X, FP) = (a Fp
X-FP
for X < (-FP)
for 1x1< FP
forX>FP
 10-7

3. MORELAND POINT-CONTACT TIRE

DIRECTION OF
MODEL WHEEL HEADING
This model computes the effects of a single non-
corotating wheel on the strut. Morelands point- DIRECTION OF
contact theory [2] is used to describe the cornering
and drift of the tire.
It is assumed that the torque about the axis of rotation
is negligible. This means that no bearing friction or
braking of the wheel is permitted. A rotating
unbalance in the center plane of the tire is considered.

3.1 Kinematic Analysis

The following information regarding the strut axle is
required by the tire model: Eo, vo, osTRm,
h, S.
The acceleration of point 0 and the angular velocity of
the lower strut will generally not be known to the tire
model. They are assumed to be of the form:
Yo I
& =A:& + A i 3 0 + Agio,
-
asTRm = a x & +a,jo +a,ko. Fig. 3. Top of Wheel

3.1.1. Coordinate Systems

Three coordinate systems are employed in this tire
model. The first XoYoZo is an inertial system
aligned with the ground. The Yo axis points to the left
side of the runway, the X , axis is aligned with the
runway axis and points to the rear of the aircraft and
the Zo axis points up. This corresponds to the
XoYoZo axis of the landing gear strut model.
The second coordinate system XI & Zl is rotated from
the XoYoZo about the Zo axis by the yaw angle$
(Fig. 3). Fig. 4. End View of Wheel
The third coordinate system X2 Y2Z2 is rotated from
the XI & Z, about the & axis by the camber angle h
(Fig. 4).

Note that i. = li is the unit vector aligned with the The angular velocity of the axle is specified by the
strut model as
axle. Since liis specified in the strut model, it is
possible to solve for the angles g and h :
-S
O R =~ o x i o + o y j O +o,kO.

tang = uy l u x Generally, it will not be possible to find values for d,

tank = -Uz sing I u y .
and ithat will satisfy G2 =EflRuT for all three
components (there are three equations but only two
unknowns). The difference between the two angular
3.1.2. Rotational Kinematics velocities is a relative rotation about the C axis. This
The angular velocity of the axle used in this tire model rate 6 , and acceleration 6 can be solved and used to
accounts for changes in the yaw angle d, and changes account for the discrepancies between the strut and
wheel system.
in the camber angle iand is
10-8

3.1.3. Translational Kinematics vp.2, = 0

Point CG, the wheel center of mass is separated from
where 2, is the unit vector indicated in Fig. 3, and P
point 0 by a distance S along the axis. Thus,
is the contact point. The unit normal vector, 2, can be
written as
The general position, velocity and acceleration in= cos(+ +w)i0 +sin(+ +v)jo
.
equations can be used to determine the motion of CG.
The position of point P, the contact point, is given by

Rp =Eo +Si2 -k2+ A i 1 .

3.2. Kinematic Constraints The velocity of P is found by differentiating. This
Three kinematic constraints must be satisfied. They constraint is used to solve for the unknown A .
are the following:

3.3. Forces and Moments Acting on the Wheel

3.2.1. Tire Rolls Without Slipping
3.3.1. Ground Forces and Moments
This can be used to define R , the spin rate of the The moment exerted by the tire on the ground, by
wheel. This condition is virtue of side slip and the twist is given by Moreland .
[21 as
- A

-VCG ' RROLL~

where R R o u is the rolling radius of the tire. The left li?T = PTViO.
side of the equation gives the component of the wheel There are also forces applied to the tire:
CG velocity in the direction of the wheel heading. -
Differentiating with respect to time, FGROUND = ~ i . d +o F N ~ ;+ ~ $ 1

where
FmD = W - K w ( R - R s ) = tire radial force,
where h is the time rate of change of the wheel spin
rate. FN = K D A + CDA = Moreland tire side force
and
3.2.2. Vertical Constraint
FD = p (FmD cosh) = tire drag force.
This constraint sets the geometric radius, R ,of the
tire by requiring that the tire always maintains contact The final element of Moreland's model relates yaw to
with the ground at the contact point. The appropriate side force:
equation is
\ir = - -1( c F ~+w).
Cl
Differentiating,
3.3.2. Tire Unbalance
Tire unbalance forces are due to centrifugal effects of
These two equations can be solved for R and R a rotating unbalance. If the product of the unbalance
mass and radius is MR and the wheel spins at a
rate , then the unbalance force is
3.2.3. No Side Slip -
F~~~~~= M R R (-sin
~ pj2 + cos pi2)
A tire under the action of a side force will drift or yaw
(as it rolls) from the direction in which it is pointed where p is the rotation of the unbalance mass and is
(see Fig. 3). This angle is called the side-slip angle or
equal to ( a t + p ; p is the initial wheel position.
the drift angle and is denoted by y . A standard
condition is to require that as the wheel rolls and side
slips, the component of the velocity normal to the path
of travel becomes zero, i.e.,
 10-9

UTAW

Fig. 6. F/A-18 NLG (FMS)

3.4. Tire Oynamies 4.2 Configuration Differences (USNvs. FMS)

Summing forces by Newton's second law, the force The launch bar is used to connect the aircraft to the
exerted by the shut on the wheel is shuttle of a catapult and launch the airaaft off the
- deck of an airaaft Carrier. When the airnaft is
Fo =mw&G-FGRouND-Gm.u intended to be solely land-based, weight savings and
reduced maintenanceexpense could be realized by
Summing Moments about the wheel's mass center and deleting the launch bar (about 35 pounds) and
applying Newton's second law for lotations yields the associated systems (Fig. 6). The launch bar power onit
moment of the shut on the wheel manifold, Fig. 7, is a collar thatrotates with the torque
links and wheels. This manifold provides the
hydraulic power necessary to raise and lower the
launch bar. Seals which contain this hydraulic
pressure provide a stabilizing friction torque.
The condition of no wheel beadng friction is imposed
by requiring that FMS versions may also include taxi lights and other
minor differences not relevant to the w e n t WO&
Mo-2=0
Nnmaical analysis of the FMS landing gear by the
landing gear vendor prior to fabrication suggested no
problems would result from removing the launch bar
4. THE F/A-l8A alone or in combination with the manifold collar.
4.1. Description of the F/A-l8A
The F/A-l8A is a single place canier-based fighter
4 3 General Observations
attack aircraft built by McDonnell Douglas, originally
for the US Navy. First fight occurred in November Most reports of shimmy in the HA-18 have been
1978. Since then, the F/A-18A and its two-seat described as pilot annoyance and decreased comfort.
variant the F/A-18B have also been sold to Canada In the most severe conditions,however, safety-of-
(with the landing gear in the original USN flight concerns have been raised, primarily due to
configmtion) and to Australia and Spain (in aland- vibration levels snfiicient to make. cockpit
based foreign military sales, or FMS, configuration). instrumentation (excluding the head-updisplay)
difficult to read.
The landing gear is in the tricycle configurationwith a
dual wheel cantilevered nose landing gear (Fig. 5 ) and Wide variations are found in the ages of the aircraft
fully levered single wheel main landing gears. and the tires where shimmy is reported.
A nose wheel steering ( N W S )unit has three modes of
o p t i o n : bigh gain steaing (for low speed taxiing),
low gain steerinp, and off/passive damping. Thelow 4 A. Tires
gain mode is automatically active'for landing and The condition of the tires appears to play a large role
rollout but can be deselected by the pilot using a in shimmy. One common. apparently effective. short
switch on the control stick. tenn maintenance correction is to change the tires
10-10

FIXED ADAPTER
Fig. 7.Steering Manifold

Fig. 8. Mass Damper

when shimmy is reprted. As this reduces the tire Life, While the wheels m.not balanced, the tires are
this makes the aircraft more costly to operate. marked by the manufactnm with a balance mark that
mbe aligned with the valve stem on the wheel to
Two specific fmrshave been observed to be minimize static unbalance. Additionally, the tires
important. These factors lead to cyclic forcing of the should be inflated with dry air or nitrogen (to
nose gear and. while this M not shimmy, it can drive minimizeinternal moisbm) and be allowed to grow
similar vibrations. Ibe first effect is rotating for atleast 12 homs before loading. This ensures an
unbalance. T%e wheels and tires are checked to insure even stretch of the rubber. An effort should also be
specification compliance but the combined assembly made to match tire diameters and pressures for the two -
is not. Apmduction tire-wheel assembly was tested by nose gear tires. The tire predsure should be regularly
MDA on a commercial automotive wheel balancing checked.
machine and the unbalance was measured at 12 inch-
ounces.While this is a severe unbalance, it is roughly To prevent formation of flat spots, the aiman should
the sum of the two unbalances allowed by the be moved evay 48 hours 01 jacked if unused for long
specifications.Since the tires wear qnickly with use, it periods of time.
would seem impractical to balance the assembly on a
routine hapis.
4.5.2. Addition of a Friction Collar to Replace the
The second factor is cold set of the rubbar redulting in
Launch Bar Power Unit Manfold
flat spots. This can reanlt from loading the tire before
the rubber is fully stretched by the inflation pres sur^ The seals of the launch bar power unit manifold
or by leaving the aircraft parked on the tires for represent about 20%of the breakout torque and about
s e v d days. This was observed during a winter flight 10% of the running torquerequired to swivel the
test in St. Louis.The tires were inadvertently allowed wheels. This damping (along with some stabilizing
to cold set and shimmy resulted. mass)can be regained by adding a collar where the
manifold is usually situated.

45. Corrections
4.5.3. Addition of Simple Masses in Conjmnion with
Several modifications to the FMS landing gear
the CollarAssnnbly
contlluration have been pmposed and/or attempted to
correct shimmy. Ibeee will be discussed in d e r of Additional weights can be added to the above collar,
increasing cost and complexity. as shown in Pig. 8, to replace some or all of the mass
removed with the launch bar. Since this mass has 110
4.5.1. Increased Attention to Tire Maintenance mechanical function it can be placed where it will not
As earliu discussed. the condition of the tires has a interfexe with gear stowage. Due to the malluradiw
big effect on shimmy and increased e m p h i s on of gyration, this mass adds up to about 30% of the
maintenance can do a lot to mitigate shimmy. original rotational inertia The weights can be
 10-11

modular in design and be added incrementally to 4.6. Qualitative Analytical Comparison

correct the problem with a minimum of increased
weight. To demonstrate the analytical model presented in
sections 2 and 3, simulations of the USN and FMS
F/A-18 landing gear configurations were run under
identical conditions. A limited amount of correlated
4.5.4. Reduce Freepluy
input data was available (much of it came from the
As discussed above, freeplay has previously been gear vendors original analysis).
shown to be a significant factor in causing shimmy.
For this comparison, the aircraft is taxiing at 110 knoh
Closing the tolerances in the torque links by adding
and the gear is disturbed with an initial 40 in/sec
special bushings has been suggested. Another possible
lateral deflection rate. Figure 9 shows the lateral
source of freeplay is gear backlash between the
deflection of the axle as a function of time. As shown
steering motor/damping unit and the ring gear on the
in Fig. 9a, the USN configuration with a launch bar is
strut.
stable and the oscillations decay. When the launch bar
inertia and collar damping friction are removed, as for
the Spanish FMS gear (Fig. 9b), the model predicts a
4.5.5. Addition of u Dummy h u n c h Bur limit cycle oscillation will result for this condition.
This solution is more complicated than the simple
masses since it must be moved for the gear to retract.
As the USN configuration is quite stable, this will 5. CONCLUSIONS
solve shimmy problems.
Based on experiences described here with the FMS
F/A-l8A, the current FMS version of the F/A-lSC/D
retains the launch bar to ensure dynamic stability of
The Spanish F/A-18 problem was corrected by the nose gear.
incorporating a collar with a simple mass damper as
described in item three, above. A flight test program The analytical model described was developed
was undertaken to evaluate this correction. Qualitative partially in response to the problems encountered with
pilot comments were used as a basis for comparing the F/A-lSA. The model has demonstrated good
- different arrangements and to select a final design. qualitative agreement with the F-15 and F/A-l8A nose
.- landing gears and stands ready to aid designers and
. The Australians have had a more chronic problem.
This is believed to be due, at least in part, to operating
dynamicists in assessing the shimmy properties of
future MDA landing gear designs.
the aircraft from rougher runways. A flight test
program was flown from May through October 1991 Future analytical work may center on correlating the
and examined the five corrective actions described model with quantitative test data and in developing
above (Ranson, et al., [91). The flight test guidelines for assigning values to the parameters. The
instrumentation included triaxial accelerometers large amount of data required by this model is
attached near the bottom of the nose gear strut and on regarded as a drawback. Sensitivity analysis could be
the fuselage near the nose gear trunnion. Also, high- used to determine the relative importance of each
speed motion picture cameras were mounted on the parameter to the final result.
aircraft keel and at the wing tips. Nearly 300 touch-
and-go landings were performed over almost 60
flights as part of this test. ACKNOWLEDGMENTS
In the flight test program flown by the Royal The author wishes to thank Prof. C. R. Barker and
Australian Air Force, they identified a shimmy mode Prof. L. R. Koval of the University of Missouri-Rolla,
as well as a speed-dependent vibration mode of the who were the principal investigators in developing the
gear, with a frequency equal to the wheel rotation rate. analytical model described here; and Jeff Pew and
At typical landing speeds, this speed-dependent Suresh Pate1 who were involved with the design and
vibration mode is very close to the second harmonic of analysis of shimmy corrections for the FMS F/A-18
the shimmy mode, explaining why the wheel gear.
unbalance and flat spots seem to aggravate shimmy.
Thanks also to Craig Beebe, Pauline Bowermaster,
! The NWS was deselected for about half of the Tom Burkhart, Matt Jansen, and Bill Macy of MDA
landings performed. This decreased the severity of who have been involved in the analysis and correction
shimmy only slightly. Of the above solutions, only effort and who offered comments and suggestions for
the addition of the dummy launch bar was found to be this paper. The encouragement of Terry Rodewald
acceptable. The Australians are currently considering and John Coyle, both also of MDA, is acknowledged.
this retrofit.
 10-12

[5] Southerland, K. F., Shimmy Analysis for the F-15 I

Nose Landing Gear Report, McDonnell Aircraft
0 - o.5oot+oo Company Report MDC A0932, March 1971.
[6] Grossman, D. T., F-15 Nose Landing Gear
Shimmy, Taxi Test and Correlative Analysis,
SAE Paper 801239, October 1980.
[7] Baumann, J., Barker, C. R., Koval, L. R., A
0.000E+00
Nonlinear Model for Landing Gear Shimmy,
presented at the ASME Winter Annual Meeting in
Atlanta, GA, December 1-6, 1991.
ASME 91-WA-DSC-14
[8] Baumann, J., Aircraft Landing Gear Shimmy,
Ph.D. Dissertation, University of Missouri-Rolla,
. 0 - -o.5ooE+oo lltlc- 0.00~+00 TIM - 0.50M+00
1992.
[9] Ranson, S . M., Jurd, K. W., and Ford, N. J.,
Fig. 9a. Simulation of F/A-l8A (USN) Hornet Nose Wheel Shimmy Evaluation, July
0 - O.SOOE+OO
1992, Royal Australian Air Force Report
AR -006-562.

O.W)OE+OO

. *
- -o.5ooE+oo
-
0

TI%- O.W#+W TIHE 0.500E+OO

Fig. 9b. Simulation of F/A-l8A (FMS)

REFERENCES
[l] vonschlippe, B., and Dietrich, R., Shimmying of
a Pneumatic Wheel, Papers on Shimmy and
Rolling Behavior of Landing Gears presented at
the Stuttgart Conference, October 16-17, 1941.
NACA TM 1365, August 1954.
[2] Moreland, W. J., The Story of Shimmy, Journal
of the Aeronautical Sciences, December 1954,
pp. 793-808.
[3] Collins, R. L., Theories on the Mechanics of Tires
and their Applications to Shimmy Analysis,
AIAA Journal of Aircraft, Vol. 8, No. 4,
April 1971, pp. 271-277.
[4] Black, R. J., Application of Tire Dynamics to
Aircraft Landing Gear Design Analysis, NASA
CP-2264, Workshop on Tire Modeling, at
Langley Research Center, September 7 - 9, 1982,
pp. 71-93.
 REPORT DOCUMENTATION PAGE
1. Recipients Reference 2. Originators Reference 3. Further Reference 4. Security Classification
of Document
AGARD-R-800 ISBN 92-836-1032-6 UNCLASSIFIED/
UNLIMITED
I I ____ ____ I~_______

5. Originator Advisory Group for Aerospace Research and Development

North Atlantic Treaty Organization
7 rue Ancelle, 92200 Neuilly-sur-Seine, France
6. Title
The Design, Qualification and Maintenance of Vibration-Free Landing Gear
7. Presented atkponsored by
The 81st Meeting of the AGARD Structures and Materials Panel,
held in Banff, Canada 4-5 October 1995.

8. Author(s)/Editor(s) 9. Date
Multiple March 1996

10. AuthorsEditors Address 11. Pages

Multiple 156

12. Distribution Statement There are no restrictions on the distribution of this document.
Information about the availability of this and other AGARD
unclassified publications is given on the back cover.
13. KeywordsKlescriptors
Landing gear Stability
Aircraft landing Landing simulation
Aircraft Design
Vibration Maintenance
Damage Vibration analysis
14. Abstract
The Structures and Materials Panel of AGARD sponsored a Workshop focusing on the various
vibrational and stability problems (e.g. shimmy, antiskid induced vibrations) that must be
considered in the early design phase of landing gear systems, especially problems which are
related to vibrations of the combined structural system formed by the landing gear, its tyres and
the flexible aircraft structure. The intention was to indicate the impact of (combined) landing
gear/aircraft vibration problems on aircraft design and to discuss the state-of-the-art technology
in this area and to define possible future steps of development.
.
 a@-
NATO @ OTAN
7 RUE ANCELLE 92200 NEUILLY-SUR-SEINE DIFFUSION DES PUBLICATIONS
FRANCE AGARD NON CLASSIFIEEX
TBlkopie (1)47.38.57.99 TBlex 810 176
Aucun staclr de publications n'a existh AGARD. A partir de 1993. AGARD dtiendraun stack limit6 des publications a w o c i h aux cycles
de conf&ences et corn sgciaux ainsi que les AGARDographies et les rapports drs groupes de travad. o r d s et publies & panir de 1993
inclus. Les demandes de mseignements doivent etre adresseeS i AGARD par lettre ou par fax i I'adresse indiqu& ci-dessus. VeuiUez ne
pes tdiphonor. La diffusion initiale de toutes les publications de I'AGARD est effectuk aup& des pays membrcs de I'OTAN par
I'intWnediaire des centres de distribution nationaux indiqnh cidessous. Des exemplaires suppl~menmirespeuvent pmfois @e o b h u s
aup& de ces centra (& I'exception des Etats-UNs). Si vous souhaitez recevoir toutes les publications de I'AGARD. ou simplement c e l b
qui Concement cenaiaS Panels, vous pouvez demander & &e inch sur Ln liste d'envoi de I'un de ces centres. Les publications de I'AGARD
son1 en vente au@s des agenw indiqUees ci-dessous. sow forme de pholocopie ou de microfiche.
CENTRES DE DIFFUSION NATIONAUX
ALLEMAGNE ISLANDE
Fachinformationszentrum Karlsruhe Director of Aviation
D-76344 Eeeenstein-Leowldshafen 2 cl0 Ruerad
BELoIQUE Reykjacik
Coordonnateur AGARD-VSL ITALIE
Etat-major de la Force atnenne Aeronautica Militare
@artier Reine Elisabeth Uffcio del Delegato Nazionale all'AGARD
CANADA
Rue d'Evere, 1140 Bmelles

Directem, Services d'informauon scientifique

Cmr
Aero rto Ratica di Mare

LUXEMBOURG
Pomezia (Roma)

Ministbre de la Defense nahonale Voir Belgique

Ottawa, Ontano KIA OK2 NORVEGE
DANEMARK Norwegian Defence Research Establishment
Danish Defence Research Establishment Attn: Bihlioteket
Ryvangs AIK 1 P.O. Box 25
P.O. Box 2715 N-2007 Kjeller
DK-2100 Copenhagen 0 PAYS-BAS
ESPAGNJ? Netherlands Delegation to AGARD
INTA (AGARD Puhlieations) National Aeros ace Laboratory NLR
Pintor Rosales 34 P.O. Box W50$
28008 Madrid 10% BM Amsterdam
r ETATS-UNIS PORTUGAL
.. NASA Headquarters
Code JOB-I
- D.C. 20546
Washinzton,
Estldo Maior da FMqa A h
SDFA - Centra de DocumenwZo
Alfragide
2700 Amadora
FRANCE
O.N.E.R.A. (Dwection) ROYAUME-UNI
r -
29, Avenue de la Division Leclerc Defence Rescarch Information Centre
92322 Chgtillon Cedex Kentigern House
65 Brown Sheet
GRECE Glasgow 0 2 SEX
Hellenic Air Force TURQUIE
Air War College
Scientific and Technical Library M U Savunma B-lig (MSBI
Dekelia Air Force Base ARGE Dairesi Ba$kanlig (MSB)
Dekelia, Athens TGA IO10 06650 BakanliLLar-Ankara
Le centre de dismbntlan national des Etats-Unis ne detient PAS de stocks des pubtlcpaoas de I'AGARD.
D6ventuelles demandes de photocopies doiveni etre formulks direaemenl auprZs du NASA Center for Aerospace Information (CASI)
& I'adreme cidessous. Toute notificatian de changcment d'ladresse doit &a fail egalement suprZs de CASI.

AGENCES DE VENTE
NASA Center for ESMnformation Retrieval Service The British Library
Aerospace Information (CASI) European Space Agency Document Supply Division
800 Elkridge Landing Road 10. rue Mario Nibs Boston Spa. Wetherhy
Lintbicum Heights. MD 21090-2934 75015 Paris West Yorkshin U 2 3 7BQ
Etats-Unis France Royaume-Uni
Les demandes de microfiches ou de photocopies de documents AGARD (y compris les demandes faites aup& du C A S ) doivent
comporter la denomination AGARD, ainsi que le numero de s&ie d'AGARD (par exemple AGARD-AG-315). Des informations
analogues, telles que le titre et la date de publication son1 souhaitables. Veuiller noter qu'il y a lieu de spkiiier AGARD-R-nnn et
AGARD-AR-nnn Ion de la commande des rapports AGARD et des rapports consultatifs AGARD respectivement. Des dftnnces
bihliographiques complhtes ainsi que des r&sumh des publications AGARD figurent dans les journaux suivants:
Scientific and Technical Aerospace Reports (STAR) Government Reports Announcements and Index (GRA&I)
publid par la NASA ScienWic and Technical publi&par le National Technical Information Service
Information Division Springfield
NASA Headquarters (J'IT) Virginia 22161
Washington D.C. 20546 Etats-Unis
Etats-Unis (accessible kgalement en mode interactif dam la base de
donnks bihliographiques en ligne du NTIS, et sur CD-ROM)

ImprirnP p a r le Groupe Communicarion C a d

45, boul. SacrP-Crpur. Hull (Qu4bec). C& KIA OS7
 NATO -@- OTAN
7 WE AMCELLE S2ZW NEUILLY-SUR-SEINE DISTWIBUTII
FRhNcE AGAlUl
Tdefax (1)47.38.57.98 TSlex610 176

WCRWAY
NCWCplkPaWR
Apn: Biblioteket
P.O.Box 25
N-ZOOI Kjeller
PORTUGAL
EstndoMniordaFoq
-
SDFA Ccntro de Do

WAIN
INTA (AGARD Rtbli lions)
Pintor Roslles 34
28008 Madrid
TURKEY
MillS Ssvunma Ea@
ARGE Dairesi Bqkat
06650 Bakanhkh-An
UNlTED KINGDOM
Defence RLKaxh Inf<
Kcntigm House
..Bmvm
6s ~
Smct
. ~.~
Glasgow 0 2 8Ex
UNlTED STATES
NASA Hedquatbm
Coee JOB-1
Wadngton, D.C. 205

The UaHsd States Natlo0.l DLstribntiol Cmtre dom NOT hold rtodrs Ot ,GARI) pObtif.nolla
.
Applications for topics should kc made d i m t to the NASA Coara for AemSpnee lnfm tim (CAS0at the address below.
Change of address q u e s t s should also go to CASI.
Y'
SALES AGENCIES
NASA Canter for -
s
B . RetrievalseNice ThC British Libmy
Documcatsuppl Centre ,
Bmton Spa,We&y
West Yorlcrhin LS23 784
unimi states ~ FmEe United Kingdom
should indude the word 'ACARD'
rb a8 title and publication d.te in
ARD-R-nnn and AGARD-AR-mn,