M. Rades - Dynamics of Machinery 3
M. Rades - Dynamics of Machinery 3
M. Rades - Dynamics of Machinery 3
Dynamics of Machinery
III
2008
Preface
This textbook is based on the third part of the Dynamics of Machinery lecture course given since 1993 to students of the English Stream in the Department of Engineering Sciences (D.E.S.), now F.I.L.S., at the University Politehnica of Bucharest. It grew in time from a postgraduate course taught in Romanian between 1985 and 1990 at the Strength of Materials Chair and continued within the master course Safety and Integrity of Machinery until 2007. Dynamics of Machinery, as a stand alone subject, was first introduced in the curricula of mechanical engineering at D.E.S. in 1993. To sustain it, we published Dynamics of Machinery in 1995, followed by Dinamica sistemelor rotor-lagre in 1996 and Rotating Machinery in 2003. The course aims to: a) increase the knowledge of machinery vibrations; b) further the understanding of dynamic phenomena in machines; c) provide the necessary physical basis for the development of engineering solutions to machinery problems; and d) make the students familiar with machine condition monitoring techniques and fault diagnosis. As a course taught for non-native speakers, it has been considered useful to reproduce, as language patterns, full portions from English texts. For the students of F.I.L.S., the specific English terminology is defined and illustrated in detail. Basic rotor dynamics phenomena, simple rotors in rigid and flexible bearings as well as the rotor dynamic analysis tools are presented in the first part. Finite element modeling of rotor-bearing systems, hydrodynamic bearings, seals and floating ring bearings are treated in the second part. This third part is devoted to the analysis of rolling element bearings, gears, vibration measurement for machine condition monitoring and fault diagnosis, standards and recommendations for vibration limits, balancing of rotors as well as elements of the dynamic analysis of reciprocating machines and piping systems. No reference is made to the vibration of discs, impellers and blades.
May 2008
Mircea Rade
Prefa
Lucrarea se bazeaz pe partea a treia a cursului de Dinamica mainilor predat din 1993 studenilor Filierei Engleze a Facultii de Inginerie n Limbi Strine (F.I.L.S.) la Universitatea Politehnica Bucureti. Coninutul cursului s-a lrgit n timp, pornind de la un curs postuniversitar organizat ntre 1985 i 1990 n cadrul Catedrei de Rezistena materialelor i continuat pn n 2007 la cursurile de masterat n specialitatea Sigurana i Integritatea Mainilor. Capitole din curs au fost predate din 1995 la cursurile de studii aprofundate i masterat organizate la Facultatea de Inginerie Mecanic i Mecatronic. Dinamica mainilor a fost introdus n planul de nvmnt al F.I.L.S. n 1993. Pentru a susine cursul, am publicat Dynamics of Machinery la U. P. B. n 1995, urmat de Dinamica sistemelor rotor-lagre n 1996 i Rotating Machinery n 2005, ultima coninnd materialul ilustrativ utilizat n cadrul cursului. Cursul are un loc bine definit n planul de nvmnt, urmrind: a) descrierea fenomenelor dinamice specifice mainilor; b) modelarea sistemelor rotor-lagre i analiza acestora cu metoda elementelor finite; c) narmarea studenilor cu baza fizic necesar n rezolvarea problemelor de vibraii ale mainilor; i d) familiarizarea cu metodele de supraveghere a strii mainilor i diagnosticare a defectelor. Fiind un curs predat unor studeni a cror limb matern nu este limba englez, au fost reproduse expresii i fraze din lucrri scrise de vorbitori nativi ai acestei limbi. Pentru studenii F.I.L.S. s-a definit i ilustrat n detaliu terminologia specific limbii engleze. n prima parte se descriu fenomenele de baz din dinamica rotorilor, rspunsul dinamic al rotorilor simpli n lagre rigide i lagre elastice, precum i principalele etape ale unei analize de dinamica rotorilor. n partea a doua se prezint modelarea cu elemente finite a sistemelor rotor-lagre, lagrele hidrodinamice, etanrile i lagrele cu inel flotant. n aceast a treia parte se trateaz lagrele cu rulmeni, echilibrarea rotoarelor, msurarea vibraiilor pentru supravegherea funcionrii mainilor i diagnosticarea defectelor, standarde i recomandri privind limitele admisibile ale vibraiilor mainilor, precum i elemente de dinamica mainilor cu mecanism biel-manivel i vibraiile conductelor aferente. Nu se trateaz vibraiile paletelor, discurilor paletate i ale roilor centrifugale. Mai 2008 Mircea Rade
Contents
Preface Contents 8. Rolling element bearings
8.1 Rolling-element radial bearings 8.2 Kinematics of rolling bearings
8.2.1 Basic assumptions 8.2.2 Simple kinematic relations for angular contact ball bearings 8.2.3 Primary rolling element bearing frequencies 8.2.4 Kinematic relations for tapered roller bearings 8.2.5 General kinematic relations
i iii 1
1 3
3 4 6 7 8
8.3 Structural frequencies 8.4 Bearing mechanical signature 8.5 Rolling element bearing damage
8.5.1 Primary damage 8.5.2 Secondary damage 8.5.3 Other damages
9 10 13
14 14 15
16
16 17 18 21 22
23
24 25 28 30
35
iv
References
9. Gears
9.1 Gear types 9.2 Gear tooth action 9.3 Gear vibrations
9.3.1 Tooth engagement 9.3.2 Effect of tooth deflection 9.3.3 Effect of tooth wear 9.3.4 Ghost components 9.3.5 Modulation effects 9.3.6 Resonance effects
39
39 40 45
45 46 47 48 48 53
54 55
55 56 58
58
59 61 67
69 72 72
75
75 76
76 77 78
79
80 81 81
CONTENTS
10.3.4 Peak-to-peak vs. r.m.s. 82
85
85 88 91 94 96 98 100
101
101 108
References
112
115
115 116
116 117 119
120 121
121 123 127 130 135 138
141
141 145 151 152 155 159
vi 12 Vibration limits
12.1 Broadband vibration standards and guidelines 12.2 Vibration severity charts 12.3 Vibration limits for nonrotating parts
12.3.1 General guidelines 12.3.2 Steam turbine sets 12.3.3 Coupled industrial machines 12.3.4 Gas turbine sets 12.3.5 Hydraulic machines 12.3.6 Reciprocating machines
163
163 164 168
168 169 170 172 172 174
176
176 177 178 180 181 183
Annexes References
192 199
13 Balancing of rotors
13.1 The mass unbalance
13.1.1 Definitions 13.1.2 Static unbalance 13.1.3 Couple unbalance 13.1.4 Quasi-static unbalance 13.1.5 Dynamic unbalance 13.1.6 Static vs dynamic unbalance
203
204
204 205 205 206 207 207
208
208 209
CONTENTS
13.2.3 Three-trial-mass method 215
vii
217
217 223
225
225 225 226
229
229 232 234
References
235
14 Reciprocating machines
14.1 Single cylinder engines
14.1.1 Gas pressure excitation 14.1.2 Inertia effects 14.1.3 Kinematics of crank mechanism 14.1.4 Connecting rod and equivalent two-mass system 14.1.5 Unbalance of a single cylinder engine
237
237
237 239 241 242 243
246
246 250 251
256
256 258 261 274
References
284
Index
287
8.
ROLLING ELEMENT BEARINGS
This chapter presents the vibration characteristics of rolling element bearings, and techniques for detecting bearing damage.
The inner ring is mounted on the shaft and rotating with it. There is a track for the rolling elements incorporated in this ring. For most applications, the outer ring is mounted in a housing and usually fixed. It also contains a track for the
DYNAMICS OF MACHINERY
rolling elements. In some instances, both races rotate. The cage connects the rolling elements and keeps an equal spacing between them. It rotates about the shaft. The rolling elements are moving with the cage between the races. Generally, rolling elements rotate around their axes and simultaneously they orbit round the bearing axis. If pure rolling motion is considered, the absolute motion can be seen as the sum of a transport motion with the cage and a relative spinning motion with respect to the cage. In addition, a certain degree of sliding occurs on the raceways, called skidding. In ball bearings with zero contact angle, a ball may have a rotational sliding normal to the contact surface. At the same time, the ball can have another kind of motion due to gyroscopic moments. If the roller axis does not coincide with the rolling axis, a slight skew of the roller in roller bearings may exist. Other motions may occur due to the misalignment of the two raceways. The kinematics of rolling bearings is influenced by structural parameters, operating conditions, lubrication and manufacturing accuracy. Higher clearances and lighter loading can cause internal sliding. Roller bearings used in aircraft engines are sometimes assembled with out of round outer raceways to yield a certain amount of preload in the radial direction in order to reduce skidding.
a
Fig. 8.2 (from [8.2])
According to the shape of the rolling element, there are ball bearings and roller bearings. Figure 8.2,a shows an angular-contact ball bearing while Fig. 8.2,b illustrates a tapered roller bearing. For the latter, the inner ring is called the cone, and the outer ring is called the cup.
Fig. 8.3
Figure 8.3 shows an angular contact ball bearing. The index i is for the inner ring, o for the outer ring, B for the ball, and m for the cage. Dm is the pitch diameter, DB is the ball diameter, Di is the diameter of the inner contact circle, and Do is the diameter of the outer contact circle, is the contact angle
DYNAMICS OF MACHINERY
( = 0 for radially loaded deep groove ball bearings). Symbols n i , no and n B represent the rotational speeds of the inner ring, outer ring and ball. Clockwise rotations are considered positive.
(8.1)
n i Dm
D D B cos = 1 B cos n i Dm . 30 2 2 Dm 60
(8.2)
The linear velocity at the center of rolling elements is equal to the mean of the outer and inner raceway velocities at contact points (Fig. 8.3)
vm = vo + vi D D = + 1 + B cos 1 B cos no Dm n i Dm . 2 120 Dm Dm 120
(8.3)
60
nm Dm .
(8.4)
Equating the two equations, the rotational speed of the cage is derived as
nm =
1 2
DB no 1 + D cos + ni m
DB 1 cos . Dm
(8.5)
The rotational speed of the cage relative to the inner ring is equal to the difference between the absolute rotational speed of the cage and that of the inner ring n D (8.6) 1 + B cos nm i = nm n i = r , 2 Dm where nr is the relative rotational speed between the outer and the inner races
nr = no n i .
(8.7)
(8.8)
Fig. 8.4
The rotational speed of a rolling element around its own axis can be obtained blocking the cage (nm = 0 ) . If v m = 0 , then
nm i = n i ,
no m = no .
(8.9)
60
n i Di = vo =
60
n B DB ,
so that
n m i Di = n B DB
and
nB=
Similarly
Di n mi . DB
(8.10)
nB=
Do n om . DB
(8.11)
DYNAMICS OF MACHINERY
2 nr Dm DB 1 nB = cos . 2 DB Dm
(8.12)
(8.13)
The impact rate for an outer race defect is equal to Z no m , the number of rolling elements passing a given point on the outer ring per minute
Z no m = Z D 1 B cos nr . 2 Dm
(8.14)
The impact rate (per minute) for a ball defect is 2 nB , because the ball defect strikes two surfaces (inner and outer races) in one revolution. For a stationary outer ring, the impact rate for a cage defect is no m . Expressing impact rates per second as frequencies f = n 60 [Hz], one obtains outer race ball pass frequency
fo = fi = Z D 1 B cos fr ; 2 Dm Z D 1 + B cos fr ; 2 Dm
(8.15)
(8.16)
f B = fr
2 Dm DB 1 cos ; DB Dm
(8.17)
fc =
1 2
no ni DB DB 1 cos 1 cos + + 60 . D D 60 m m
(8.18)
Note that the above relations are approximate, assuming pure rolling motion and neglecting sliding motions. For normal speeds, these defect frequencies are usually less than 500 Hz. Amplitude modulations especially at the shaft rotational frequency can produce sum and difference sidebands.
Example 8.1
A radial-thrust ball bearing type 46305, GOST 831-54 mounted on a shaft with the rotational speed n i = 1000 rpm , has the following geometry:
ball diameter DB = 14.3 mm , pitch diameter Dm = 77.5 mm , contact angle = 26o , number of balls Z = 10 [8.4]. From the formulas for bearing frequencies (8.15)-(8.18) we obtain:
f c = 6.99 Hz , f B = 30.72 Hz , f o = 69.9 Hz , f i = 97 .1 Hz .
Example 8.2
A radial ball bearing type SKF6211, mounted on a shaft with the rotational speed n i = 3000 rpm , has the following geometry: ball diameter DB = 25 mm , pitch diameter Dm = 62 mm , contact angle = 0 , number of balls Z = 10 . The bearing frequencies (8.15)-(8.18) are:
f c = 20 Hz , f B = 260 Hz , f o = 205 Hz , f i = 295 Hz .
(8.19)
When the two rings rotate in the same direction, we obtain the following speeds
DYNAMICS OF MACHINERY
the cage speed the cage speed relative to the inner ring the outer ring speed relative to the cage the roller spin speed
nm = n i K1 + no K 2 ,
( ) no m = ( no n i ) K1 ,
nmi = no n i K 2 ,
nR = no n i
Dm ) 2D K1K 2 .
R
Note that the relations for angular contact ball bearings can be obtained from equations (8.20)-(8.23) by substituting DR = DB and K1 = 1 2 D 1 B cos , Dm
K2 = 1 2 DB 1 + D cos . m
(8.24)
fn =
k k 2 1
1 2 k +1 a
2
EI m
[Hz] ,
(8.25)
where k is the number of waves around circumference ( k = 2 , 3, 4 ) , a is the radius to neutral axis, I is the moment of inertia of cross-section, E is Youngs modulus, and m is the mass of race per linear length; ball natural frequency
f Bn = 0.848 DB E 2 [Hz] ,
(8.26)
where DB is the ball diameter and is the density of the ball material. These are the free natural frequencies of individual elements. It is difficult to estimate how these frequencies are affected by assembly into a full bearing and mounted in a housing. However it is indicated that resonances are not altered significantly. Resonance of the ball is usually far above the range of vibration analysis and can be ignored. The outer ring resonance can be excited by the rotating balls (rollers). They deform the race into a flexural pattern (with a number of wavelengths equal to the number of rolling elements) which rotates with the ball passing frequency. It can also be produced by the waving motion of the balls around their theoretical circumferential path. In rolling bearings the external load is carried by a finite number of rolling elements. Their number under load varies with the angular position of the cage. The elastic deflection produced by the Hertzian contact under load varies with the position of the rolling element relative to the line of load. This gives rise to a periodical variation of the total stiffness of the bearing assembly and generates the so-called varying compliance vibrations of the rotor [8.13]. Their fundamental frequency is equal to the ball (or roller) passage frequency over the outer ring. Higher harmonics are also excited, to a degree decreasing with their order, mainly due to deviations of the bearing parts from the perfect geometric shape. The magnitude of shaft movements is a function of the external load, number of rolling elements, radial clearance and the local stiffness
10
DYNAMICS OF MACHINERY
between rolling element and tracks, as given by the Hertzian theory for elastic contacts (H. Hertz, 1881). The parametrically excited vibrations of the rotor-bearing system, with strongly coupled vertical and horizontal movements, are described by nonlinear equations of motion with time varying coefficients. Variable contact compliance vibrations are of importance only at frequencies in the neighborhood of the rotational frequency of the bearing, and are generally of appreciable magnitude only for rather high radial loads. Structural resonances can also be excited by other distributed defects such as race misalignment or eccentricity, lack of roundness, waviness of the rolling surfaces and unequal ball diameters produced during the manufacturing process. These distributed defects often give rise to excessive contact forcers which in turn result in premature surface fatigue and ultimate failure. Note that waviness defines relatively widely-spaced surface irregularities. In principle, surface roughness is the same type of geometrical imperfection as waviness. Their distinguishing characteristic is the spacing of irregularities, which is finer for surface roughness. Waviness is used to imply irregularities up to an order of 200 waves per circumference, while surface roughness contains waves of a much higher order. Typical examples are the following: at a frequency of 300 Hz, the inner ring has 16 to 17 waves per circumference, and the outer ring has 24 to 27. At a frequency of 1800 Hz, the inner ring has 94 to 101 waves per circumference, and the outer ring has 147 to 166 [8.14]. Geometrical irregularities in the form of a waviness with a few cycles around the circumference give rise to low frequency vibrations. The vibrations of radially loaded bearings with stationary outer rings and positive radial clearances are primarily related to the inner race waviness and varying roller diameter, rather to other geometrical errors. The vibrations due to non-uniform roller diameters occur at cage speed harmonics, while vibrations due to inner race waviness occur at shaft speed harmonics with a side band spaced with the roller passage frequency occurring at the high harmonics [8.15].
11
mechanical signature can be related to the specific mechanical defects within the bearing. The amplitudes of these peaks are a measure of the energy transmitted by impacts and, therefore, of the smoothness of the bearing operation. Peaks generated by unbalance, misalignment and other sources have to be distinguished from bearing generated peaks.
12
DYNAMICS OF MACHINERY
A comparison of the mechanical signatures of two ball bearings of the same type would require data obtained at the same speed, since most of the vibration frequencies are proportional to speed. Rather than trying to hold speed constant, it was found better to have mechanical signatures independent of speed. This is accomplished by normalizing all frequencies relative to the fundamental rotational speed. The procedure is called order normalization. For stationary outer ring, the fundamental frequency of rotation is that of the inner ring. The spectra in Figs. 8.5 and 8.6 are plotted versus frequency orders.
The mechanical signature of a good bearing is shown in Fig. 8.7. The amplitude is calibrated for 90dB equal to 0.26 g. The noise floor is approximately 50dB or 0.0026 g. The first order is the only frequency evident in this spectrum. The amplitude of the spectrum is plotted in log scale to provide the greatest vertical magnification. This allows the detection of small defect frequencies in a measurement containing a large frequency component. Otherwise the random noise due to friction may dominate the spectrum making it difficult to locate frequencies that can be correlated with bearing defects. A spectrum averaging technique can be applied to enhance the signal-to-noise ratio of the periodic discrete frequencies generated by the ball bearing. A mechanical signature showing a ball defect is illustrated in Fig. 8.8. The presence of two large orders (5.80 and 1.00) generates sum and difference frequencies that can be identified at 5.80 1.00 and 5.80 2.00 . This bearing also shows orders associated with inner race defects that can be explained by a nonlinear (N.L.) theory taking into account race waviness, eccentricity and large ball diameter variations.
13
Generally, the outer race geometrical imperfections produce a vibration spectrum having peaks at the harmonics of the outer race defect frequency, with side bands spaced with the cage frequency. The inner race surface irregularities produce a spectrum having peaks at the harmonics of the inner race defect frequency. The side bands are spaced with an interval related to the cage frequency and the shaft running frequency.
14
DYNAMICS OF MACHINERY
15
as flaking or spalling. The flaking progressively increases in extent and eventually makes the bearing unserviceable. The life of a rolling bearing is defined as the number of revolutions the bearing can perform before incipient flaking occurs. The causes of premature flaking may be heavier external loading than had been anticipated, preloading on account of incorrect fits or excessive drive-up a tapered seating, oval distortion owing to shaft or housing seating out-of-roundness, axial compression as a result of thermal expansion, misalignment, etc. Flaking may also be caused by other types of damage, such as indentations, deep seated rust, electric current damage or smearing. Cracks Cracks may form in bearing rings for various reasons. The most common cause is rough treatment when bearings are being mounted or dismounted (hammer blows, excessive drive-up on tapered seatings, heating and mounting on shafts with wrong tolerances). Flaking acts as a fracture notch and may lead to cracking of the bearing ring. Cage damage Cage failures are due to vibrations, excessive speeds, wear and blockage by flaked material wedged between the cage and a rolling element. Misaligned rings produce oval ball paths that distort the cage once per revolution leading to fatigue cracks. The cage is the first component to be affected when the lubrication becomes inadequate. It is always made of softer material than the other components of the bearing and consequently it wears comparatively quickly. Two approaches have been used to study the vibration and acoustic response of rolling element bearings due to defects in the bearings. One is to run the bearings until they fail and monitor the changes in their vibration and acoustic response. Usually the failure is accelerated by overloading, overspeeding, or starving the bearings of lubricant. The other approach is to intentionally introduce defects in the bearings by techniques such as acid etching, spark erosion, scratching, or mechanical indentation. The vibration response of the bearings is measured and compared with the responses of good bearings.
16
DYNAMICS OF MACHINERY
Grooving shows as continuous circumferential indentation on balls produced by the balls running on the retaining diameter of the counterbored raceway. Brinelling. The term applies to a bearing which has been statically loaded to an extent such that the raceways and rolling elements are permanently deformed. A brinelled bearing has indentations in the raceways and often has corresponding flats on the rolling elements. Fretting is a corrosive form of wear caused by very slight movement between two metal surfaces under very high contact pressure. The formation of an iron-oxide paste between two fretting steel members is not uncommon. It is often seen between the inner ring and the shaft. Creeping is a relative movement between the bearing inner ring and the shaft, caused by inadequate interference fit for the applied load. Creeping is evidenced by circumferential scoring on the bearing bore and shaft. It may be an advanced stage of fretting. Spinning is an advanced stage of creeping. The relative movement between inner ring and shaft is much greater than in creeping and the sliding surfaces may become polished. The iron-oxide from the fretting phase may still be present and assist in further wear. Discoloration due to temperature indicates operation of the bearing elements with marginal lubrication or under excessive power conditions.
17
1 x= T
mean value
x (t )d t ;
0
(8.27)
xr .m.s . =
1 T
x
0
(t ) d t .
(8.28)
Usual practice is to measure the r.m.s. velocity of the overall vibration level at the bearing housing. Measured levels are compared with general standards or with established reference values for each bearing. By plotting the measurement results over time the trend in vibration can be followed and extrapolated to give a prediction of when the bearing needs replacement. However, because the overall vibration level often increases only in the final stages of failure, this method gives late warnings of failure. Two time-waveform indices used to get early warnings of the bearing failure the Crest Factor and the Kurtosis are presented in the following.
(8.29)
The curve in Fig. 8.9 shows a typical trend for the Crest Factor as the bearing condition deteriorates. Initially, for a bearing with no faults there is a relatively constant ratio of about 3.0. As localized faults develop, the resulting impacts increase the peak level substantially, but have little influence on the r.m.s. level. The peak level will typically grow to a certain limit. As the bearing condition deteriorates, more spikes will be generated per ball-pass, finally influencing the r.m.s. level, even though the individual peak levels are not greater. Towards the end of the bearing life, the crest factor may have fallen to its original value, even though both peak and RMS levels have increased considerably. The best way to trend the data is as illustrated in Fig. 8.9: peak and r.m.s. levels on the same graph, with Crest Factor inferred as the difference between the two curves (log scale).
18
DYNAMICS OF MACHINERY
Measuring the overall vibration level over a wide frequency range (10 Hz to 10000 Hz), the method is prone to interference from other vibration sources.
P ( x, x + x ) =
i =1
ti . T
(8.30)
When x 0 and T , one obtains the amplitude probability density p (x ) , giving the probability to have an amplitude x , plotted on the left of Fig.
19
8.10. The bell-shaped curve corresponds to the Gaussian (normal) distribution, which describes signals occurring in practice with sufficient precision.
Fig. 8.10
p (x ) dx = 1
(8.31)
as a function of the dimensionless variable x , where is the r.m.s. value for zero mean.
Fig. 8.11
20
DYNAMICS OF MACHINERY
It is found that 99.8% of all events occur in the range 3 . From that follows approximately that the peak value is 3 , which, divided by the r.m.s. value , gives for the Crest Factor (8.29) a value of 3.0 . An obvious measure of bearing condition is obtained by observing changes in the probability at particular amplitude levels, those above 3 providing most significant information.
A typical result for a bearing is shown in Fig. 8.12, where the vertical logarithmic scale was chosen to enhance the changes at low probability which have been found important in detection of bearing damage. Endurance tests have been carried out at constant speed and twice the recommended load, to accelerate fatigue failure. The overall acceleration level was measured in the frequency range 3Hz 5 kHz . The three curves correspond to increased test durations, expressed in terms of the bearing life L10 = 50 h . Note that L10 is defined as the rating life of a group of apparently identical rolling element bearings, operating under identical loads and speeds, with a 90% reliability before the first evidence of fatigue develops [8.21]. A fatigue spall of specific size ( 6 mm 2 ) is usually considered (ISO 281, 2006). In the early stages of the test, i.e. 0.067 L 10 ( 3.35 h ), when the bearing is undamaged, the distribution curve is an inverted parabola which indicates a normal (Gaussian) distribution. With incipient damage at 1.4 L10 ( 70 h ), pronounced changes occur in the tail of the distribution curves. This is consistent with the observation made on Fig. 8.9 that the measured peak acceleration level increases
21
but the r.m.s. level remains relatively unchanged. With increasing time, i.e. 1.6 L10 and advancing damage, the tail of the distribution curve initially broadens.
Mn =
x p (x ) dx ,
n
(n = 1, 2 , 3, ...) .
(8.32)
x=
__ 2
x p (x ) dx ,
(8.33)
x =
x p (x ) dx .
2
(8.34)
(x x ) p (x ) dx ,
2
(8.35)
where is the standard deviation (mean square error). Odd moments, i.e. n = 1, 3, 5,, etc., relate information about the position of the peak density relative to the median value. Even moments, n = 2, 4, 6 ... , etc., indicate the spread in distribution. Higher moments (n > 2 ) generally have the mean value removed and are normalized by the standard deviation. The third moment yields
skew (x ) =
x p (x ) dx
3
(8.36)
22
DYNAMICS OF MACHINERY
Kurtosis
kurt ( x ) =
x p (x ) dx
4
(8.37)
Skewness is a measure of symmetry, or more precisely, the lack of symmetry. The skewness for a normal distribution is zero.
8.6.5 Kurtosis
The Kurtosis factor is the ratio of the fourth central moment of the amplitude distribution to the second power of the second central moment. Kurtosis characterizes the relative peakedness or flatness of a distribution compared to the normal distribution (Karl Pearson in Biometrika, 1905). A normal distribution has a Kurtosis of 3 and is called mesokurtic. Indeed, for a Gaussian distribution
p (x ) =
the fourth statistical moment is
(x x ) 2 exp , 2 2 2
(8.38)
M4 =
( x x ) p (x ) dx =
4
( x x ) 4 exp (x x2)
2
dx .
Denoting
y=
we obtain
xx , 2
4 4
d x = 2 dy ,
M4 =
y exp ( y ) dy = 3
4 2
M2 =
(xx )
p ( x ) dx =
2 2
y exp ( y )dy =
2 2
23
kurt (x ) =
(M 2 ) 2
M4
=3.
A flat distribution with short tails has a Kurtosis value less than 3 and is called platykurtic. A peaked distribution with longer tails has a Kurtosis value greater than 3 and is said to be leptokurtic. Higher Kurtosis means that more of the variance is due to infrequent extreme deviations, as opposed to frequent modestlysized deviations. Kurtosis provides an early warning of surface damage (Dyer and Stewart, 1978). For a good bearing it equals 3. Bearing damage causes an increase in the impulsive components of the vibration signal, due to impacting. The signals become more spiky. A damaged bearing exhibits a non-Gaussian probability distribution with dominant tails which increase the Kurtosis value. The advantage of Kurtosis, as a parameter for detecting the condition of rolling element bearings, lies in the finding that it remains close to 3 ( 8% ) for an undamaged bearing and is insensitive to the load or speed of bearing. One disadvantage is that the Kurtosis value comes down to the level of an undamaged bearing (i.e. 3) when the damage is well advanced. Therefore, it has been suggested to measure Kurtosis in selected frequency bands [8.23]. Experiments have shown that initial damage increases Kurtosis in the lower frequency bands. As damage spreads, the Kurtosis value begins to decrease in the first band (2.5 5 kHz ) , while increasing in the other bands. At the end of the useful life of the bearing, the highest Kurtosis numbers are in the highest frequency band (40 80 kHz ) [8.24].
24
DYNAMICS OF MACHINERY
Particular emphasis is placed on changes in the frequency spectrum of the vibration signals. Additional processing techniques are used as an aid to interpretation of the spectrum, like synchronous time signal averaging and cepstrum analysis (see Sections 9.6.1 and 9.7).
25
Figure 8.13 shows how a developing fault changes the spectrum as a function of time. The frequency spectrum gives earlier warnings than monitoring of overall vibration. The level of overall vibration only increases after an increasing component has become the highest peak in the spectrum. Whenever an increase of the baseline (reference) level is detected, a further analysis is carried out for fault diagnosis. The frequency range in which the levels are exceeded gives an indication of what type of faults to expect. Band-pass analysis involves filtering the vibration signal above and/or below specific frequencies in order to reduce the amount of information presented in the spectrum to a set band of frequencies. These frequencies are typically where fault characteristic responses are anticipated. Changes in the vibration signal outside the frequency band of interest are not displayed. Vibrations produced by machines with rolling element bearings occur in three frequency regions (Fig. 8.14): a) the rotor related region, in the range of 1 4 to 3 times the shaft rotational speed. In the low frequency region, unbalance, misalignment, bent shaft, mechanical looseness, oil whirl, hysteretic whirl, etc. will be found; b) the prime spike (element passage) region, normally 1 to 7 times the element passage rate. In the medium frequency region, indication on wear and incipient faults in a gearbox will be found, as well as eccentricity, uneven gearwheels, misaligned gearwheels, etc.; c) the high frequency (spike energy) region, from 5 kHz to approximately 25 kHz . At very high frequencies, to the megahertz region, measured data may contain information related to incipient faults in rolling-element bearings, rubs, cavitation, valve noise, etc. Guideline alarm threshold values are [8.1]: 7.7 mm s peak for region a, 2.5 to 3.8 mm s peak values for region b, 3 to 4 g peak values for region c.
26
DYNAMICS OF MACHINERY
For a bearing with fixed outer ring, rotating inner ring and fixed load, Fig. 8.15 shows the signal produced by a defect in the fixed race
For a defect in the rotating inner race, it is important to consider the load distribution around the bearing circumference. This results in a modulation effect illustrated in Fig. 8.16.
When the load is not fixed in space, but rotating as for centrifugal forces, modulations are also generated for a fixed outer race defect. The intensity of impact energy is a function of pulse amplitude, pulse rate and pulse duration. This signal is processed by a Spike Energy detector (IRD Mechanalysis). A simplified flow chart of the Spike Energy signal processing is shown in Fig. 8.17. The vibration signal from an accelerometer is passed through a high frequency bandpass filter. The purpose of filtering is to reject the normal rotational
27
vibration components, such as unbalance and misalignment, while allowing the vibration generated by impacts to remain. The lower corner frequency, f c , can be selected between 100 and 5000 Hz, and the upper corner frequency, f d , is 65 kHz . The filtered vibration signal passes through a peak-to-peak detector with a properly selected output time constant, which detects and holds the peak-to-peak values. Then, it decays at the rate of the time constant until the next pulse occurs. The instrument repeats this process.
It is customary to measure accelerations in g units 1 g = 9.81 m s 2 . The acceleration measured to describe the energy produced by early bearing defects is measured in gSE units (acceleration units of Spike Energy). These faults produce a high frequency carrier and modulating sidebands. The carrier is the natural frequency of the excited bearing component. The modulating sidebands are caused by load and speed changes. The gSE reading is determined by the intensity of the high-frequency peaks in the vibration signal. Pulses with large amplitude and high repetition rate produce high overall gSE readings. In addition to overall Spike Energy measurements, a Spike Energy Spectrum can be obtained by fast Fourier transform (FFT) analysis of the signal from the Spike Energy detector. It is different from the acceleration frequency spectrum. The components in the gSE spectrum are modulation frequencies that are related to the high frequency carrier, such as the resonance frequency of the machine element.
28
DYNAMICS OF MACHINERY
Spike Energy readings can be affected by accelerometers and mounting methods. The gSE readings can be different if different transducers are used unless transducers have exactly the same frequency response characteristics. To ensure the consistency of gSE data, it is necessary to always use the same accelerometer and the same mounting conditions. Stud mounting is the best. Spike Energy readings are highly dependent upon the machine size, configuration and bearing details. Users must go through a learning phase, taking periodic readings, observing trends, noting failed bearings and building up historical background before accurate condition assessments are made. As an order of magnitude, energy alarm levels of 0.5 gSE have been used in an application with the dryer rolls on a paper machine.
The traditional method uses an analogue bandpass filter plus a rectifier and a smoothing circuit (Fig. 8.18). The filter extracts the resonance excited by the bearing fault from the frequency spectrum and the detector detects the envelope. In modern signal analyzers, zooming around a resonance excited by the bearing defect extracts the useful part of the frequency spectrum, and then the Hilbert transform generates the envelope of the time signal. The spectrum of the envelope is calculated to show the repetition frequency of the fault generated pulses. Envelope Detection or Amplitude Demodulation is the technique of extracting the modulating signal from an amplitude-modulated signal. The result is the time history of the modulating signal. This signal may be studied/interpreted as it is in the time domain or it may be subjected to a subsequent frequency analysis.
29
Envelope Analysis is done on the FFT frequency spectrum of the modulating signal.
Envelope detection is detailed in Fig. 8.19. The time signal is filtered around the frequency region where the increase is detected (in the kHz range). This leaves the high frequency signal which contains the pulse-excited vibration of the bearing housing without most of the contaminating signals. This signal is then rectified and low-pass filtered, at a frequency approximately one half the bandwidth of the bandpass filter. The signal now looks somewhat like the original pulses from the bearing, but of most significance, we have thus recreated the pulse frequency.
30
DYNAMICS OF MACHINERY
By analyzing this signal with an FFT analyzer, the pulse frequency can be determined exactly. Since the impulse rate can be calculated, see equations (8.15)(8.18), the source of the fault can be pinpointed. Note that the real frequency will be slightly lower than the calculated one due to sliding.
If the fault is on the rotating race, then it is sometimes possible to see the amplitude modulation from the varying load on the crack illustrated in Fig. 8.16. This modulation effect will turn up as sidebands around the lines corresponding to the pulse rate, spaced at the rotational speed (Fig. 8.20).
31
Empirical relationships were developed that provided both a measure of the theoretical lubricant film thickness between the bearing surfaces, as well as the overall condition of the bearing surface. The shock pulse analyzer detects impacts of very short duration arising from the presence of pits and spalls. Unlike conventional vibration analysis, that monitors a broad vibration band with the objective of detecting discrete frequencies, the shock pulse method (SPM) measures and evaluates the ultrasonic frequency band centered around 36 kHz.
32
DYNAMICS OF MACHINERY
Shock (or stress) waves that result from metal-to-metal contact are short duration bursts of energy that travel at the speed of sound through the material. As the wave travels, it dissipates energy through the structure, thereby reducing the wave pulse. The SPM is designed to detect the weak shock pulse signals using an accelerometer with a natural frequency of about 36 kHz, ideally placed very closely to the subject bearing. In fact, a patented design called Tandem-Piezo is used, which enables the accelerometer to accurately measure both shock pulse and vibration. To distinguish the shock pulses from vibration, a band pass filter around de 36 kHz shock pulse signal is used. This helps isolate the shock pulse from other interference created by machinery vibrations. The last stage of signal processing is the conversion from a waveform to analog pulses. This process provides a signal that then can be processed to determine bearing condition.
b
Fig. 8.22 (from [8.36])
Figure 8.21 shows the block diagram of an early shock pulse meter [8.14]. The accelerometer output (Fig. 8.22, a) is passed through a high-gain amplifier tuned at the resonant frequency of the accelerometer, the amplifier acting as a very sharp band filter. The filtered and amplified shock pulse is shown in Fig. 8.22, b.
The signal is rectified, averaged and then passed through a peak-sampleand-held circuit. This measures the information and displays it on a counter which records the number of peaks occurring above a defined peak amplitude;
33
alternatively, it presents the signal r.m.s. value. The amplitudes of analog shock pulses are displayed as function of time in Fig. 8.22, c. The bearing condition is defined by a string of pulses with varying magnitudes (Fig. 8.23). A shock pulse analyzer measures the shock pulse magnitude on a decibel scale, in dBsv (decibel shock value). It takes a sample count of the shock pulses occurring over a period of time and displays: LR (Low Rate of occurrence), the value for the relatively small number of strong shock pulses, and HR (High Rate of occurrence), the value for the large number of weak shock pulses in the pattern. The difference between LR and HR is called the delta value, .
The strength of the individual pulses, and the ratio between stronger and weaker pulses in the overall pattern, provide the row data for bearing condition analysis. The magnitude of these pulses is dependent on the bearing surface condition and the peripheral velocity of the bearing. In undamaged bearings, the shock level varies with the thickness of the lubricant film between the rolling elements and raceway. The relationship between stronger and weaker pulses, however, is only slightly affected (Fig. 8.24, a).
34
DYNAMICS OF MACHINERY
Surface damage causes an increase of up to 1000 times (60 dB) in shock pulse strength, combined with a distinct change in the ratio between stronger and weaker pulses (Fig. 8.24, b).
The shock pulse readings are evaluated and a code is displayed describing the general bearing condition. Code A is for a bearing in good condition. There is no detectable damage to the surfaces of the load carrying parts, and no extreme lack of lubricant in the rolling interface. Figure 8.25, a shows a typical shock pulse pattern from a good bearing: a low shock level and a normal delta value. Code B indicates a dry running condition, causing a high HR value and a low delta value (Fig. 8.25, b). Code C is for reduced condition defined by an increased shock pulse level with a large delta value (Fig. 8.25, c). This denotes incipient surface damage. Code D is for bearing damage characterized by a high shock level with a large delta value (Fig. 8.25, d). A contamination of the lubricant by hard particles causes a similar pattern.
35
Output data are displayed as in Fig. 8.26. The delta value = LR HR is plotted as a function of HR. The fields marked A, B, C, D correspond to the condition codes. The black point marks a shock pulse reading. For a bearing in good condition it is in the field A. Developing surface damage causes a marked increase of the delta value, HR remains low while LR increases. The marker point moves upwards, from A through field C towards D. For poor lubrication, the condition code changes from A to B then to D as damage develops and increases. The marker point moves to the right. Shock pulse is not limited to determining the condition of rolling element bearings. Any machine element with continuous metal-to-metal contact gives off shock pulse signals. Equipment such as gearboxes, lobe compressors, screw compressors and centrifuges can be monitored using SPM.
36
DYNAMICS OF MACHINERY
Figure 8.27 gives an example of the development of an outer race fault in a ball bearing, showing the spectra on the left and the cepstra on the right [8.39]. The initial cepstrum has a single peak at the quefrency equal to the rotational period. The second cepstrum (after 5 months) reveals the development of a fault by a series of new rahmonics. The quefrency of the first rahmonic is 4.1 times lower than the shaft speed quefrency (r.p.m.). This means that its corresponding frequency is 4.1 times the shaft speed. In this case it immediately identified the source as corresponding to the impact rate of the outer race for a particular bearing in the gearbox (which had 10 balls and a ball-diameter-to-pitchdiameter of 0.18). The cepstrum can only be used for bearing fault diagnosis when the fault generates discrete harmonics in the spectrum. This is usually the case for highspeed machines, where resonances excited by the fault represent a relatively low harmonic order of the ballpass frequencies involved, but is often not the case for slow-speed machines, where this order may be in the hundreds or even thousands, and these high harmonics are often smeared together. It should be noted that envelope analysis (see Section 8.7.3), where the envelope obtained by amplitude demodulation of the band-pass filtered signal is frequency analyzed, can be used in either case.
References
8.1. * Predictive maintenance through the monitoring and diagnostics of rolling element bearings, Bently Nevada Application Note AN044, June 1988. 8.2. Li, C. J. and McKee, K., Bearing diagnostics, Encyclopedia of Vibration, Braun, S., Ewins, D. and Rao, S.S., eds., Academic Press, London, 2002, p.143152. 8.3. Changsen, W., Analysis of rolling element bearings, Mechanical Engineering Publications, Ltd., London, 1991. 8.4. Scheithe, W., Vibration measurement a method for early detection of rolling element bearing failures, Practice of Vibration Analysis 13, Schenck C 1213e. 8.5. Hamrock, B. J. and Anderson, W. J., Rolling-Element Bearings, NASA Reference Publication 1105, June 1983. 8.6. Jones, A. B., The mathematical theory of rolling element bearings, Mechanical Design and Systems Handbook, H.A.Rothbart, ed., McGraw Hill, New York, 1964, p.13-1 to 13-76. 8.7. Jones, A. B., Ball motion and sliding friction in ball bearings, Journal of Basic Engineering, Trans. ASME, vol.81, March 1959, p.1-12.
37
8.8. Jones, A. B., A general theory for elastically constrained ball and radial roller bearings under arbitrary load and speed conditions, Journal of Basic Engineering, Trans. ASME, vol.82, June 1960, p.309-320. 8.9. Harris, T. A., An analytical method to predict skidding in high speed roller bearings, Trans. ASLE, vol.9, 1966, p.229-241. 8.10. Gupta, P. K., Dynamics of rolling element bearings, Journal of Lubrication Technology, Trans.ASME, vol.101, no.3, 1979, p.293-326. 8.11. Meyer, L. D., Ahlgren, F. F. and Weichbrodt, B., An analytic model for ball bearing vibrations to predict vibration response to distributed defects, Journal of Mechanical Design, Trans. ASME, vol.102, no.2, April 1980, p.205-210. 8.12. Tandon, N. and Nakra, B. C., Vibration and acoustic monitoring techniques for the detection of defects in rolling element bearings A review, Shock and Vibration Digest, vol.24, no.3, March 1992, p.3-11. 8.13. Sunnersj, C. S., Varying compliance vibrations of rolling bearings, Journal of Sound and Vibration, vol.58, no.3, 1978, p.363-373. 8.14. Collacott, R. A., Mechanical Fault Diagnosis, Chapmann and Hall, London, 1977. 8.15. Su, Y.-T., Lin, M.-H. and Lee, M.-S., The effects of surface irregularities on roller bearing vibrations, Journal of Sound and Vibration, vol.165, no.3, 1993, p.455-466. 8.16. Babkin, A. S. and Anderson, J. J., Mechanical signature analysis of ball bearings by real time spectrum analysis, Nicolet Instruments Application Note 3, May 1972. 8.17. * Bearing failures and their causes, SKF Repro 19208. 8.18. Roos, C. H., Vibration signature analysis of bearings and electronic packages, Paper SI-460, 41st Shock and Vibration Symposium, Colorado Springs, Oct 1970. 8.19. * Detecting faulty rolling-element bearings, Brel & Kjaer Application Note, BO 0210-11. 8.20. Dyer, D. and Stewart, R. M., Detection of rolling element bearing damage by statistical vibration analysis, Journal of Mechanical Design, Trans. ASME, vol.100, no.2, Apr 1978, p.229-235. 8.21. Lundberg, G. and Palmgren, A., Dynamic capacity of rolling bearings, Acta Polytechnica, Mechanical Engineering Series, vol.1, no.3, Stockholm, 1947. 8.22. Martin, H. R., Statistical moment analysis as a means of surface damage detection, Proc. 7th International Modal Analysis Conference, Schenectady, New York, 1989, p.1016-1021.
38
DYNAMICS OF MACHINERY
8.23. Stewart, R. M., Application of signal processing techniques to machinery health monitoring, Stewart Hughes Ltd., Southampton, U.K., 1981. 8.24. Volker, E. and Martin, H. R., Application of Kurtosis to damage mapping, Proc. 4th International Modal Analysis Conf., Los Angeles, 1986, p.629-633. 8.25. Daadbin, A., and Wong, J. C. H., Different vibration monitoring techniques and their application to rolling element bearings, International Journal of Mechanical Engineering Education, vol.19, no.4, 1991, p.295-304. 8.26. Mathew, J. and Alfredson, R. J., The condition monitoring of rolling element bearings using vibration analysis, Journal of Vibration, Acoustics, Stress and Reliability in Design, Trans. ASME, vol.106, July 1984, p.447-453. 8.27. Taylor, J. I., Identification of bearing defects by spectral analysis, Journal of Mechanical Design, Trans. ASME, vol.102, no.2, April 1980, p.199-204. 8.28. Angelo, M., Vibration monitoring of machines, Brel & Kjaer Technical Review, no.1, 1987. 8.29. Xu, M., Spike Energy and its applications, Shock and Vibration Digest, vol.27, no.3, May-June 1995, p.11-17. 8.30. Sidahmet, M. and Dalpiaz, G., Signal generation models for diagnostics, Encyclopedia of Vibration, Braun, S., Ewins, D. and Rao, S.S., eds., Academic Press, London, 2002, p.1184-1193. 8.31. Shea, J. M. and Taylor, J. K., Using Spike Energy for fault analysis and machine-condition monitoring, IRD Mechanalysis Technical Report 11, 1990. 8.32. Mignano, F., Envelope detection, Shock and Vibration Digest, vol.29, no.3, March 1997, p.18-23. 8.33. McFadden, P. D. and Smith, J. D., Vibration monitoring of rolling element bearings by the high frequency resonance technique. A review, Tribology International, vol.17, 1984, p.1-18. 8.34. Courrech, J. and Gaudet, M., Envelope analysis the key to rolling-element bearing diagnosis, Brel & Kjaer Application Note No. BO0187-11. 8.35. Butler, D. E., The shock-pulse method for the detection of damaged roller bearings, Non-Destructive Testing, April 1973, p.92-95. 8.36. Lee, G., What is shock pulse method?, www.reliabilityweb.com. 8.37. * Shock Pulse Analyzer A2011, Instruction Manual, SPM Instrument AB, no.71416.B, Nov.1992. 8.38. Lundy, J., Detecting lubrication problems using shock pulse, Lubrication and Fluid Power, Jan-Feb.2006, p.57-62. 8.39. Randall, R. B., Cepstrum analysis, Encyclopedia of Vibration, Braun, S., Ewins, D. and Rao, S.S., eds., Academic Press, London, 2002, p.216-227.
9.
GEARS
This chapter provides an overview of current vibration methods used for gearbox diagnostics. It presents gear defects, gear errors and condition indicators used for gear failure detection. Differences of various metrics are in the characteristic frequencies that are included, excluded, or used as a reference.
b
Fig. 9.1 (from [9.1])
Helical gears, used to transmit motion between parallel shafts, are shown in Fig. 9.1, b. The line of contact of helical-gear teeth is diagonal across the face of the tooth, so that there is a gradual engagement of the teeth and a smooth transfer of load from one tooth to another. Helical gears subject the shaft bearings to both
40
DYNAMICS OF MACHINERY
radial and thrust loads. Double helical gears (herringbone) are used for the transmission of high torques at high speeds, and to cancel out the thrust load. For power transfer between intersecting shafts, there are straight bevel gears (Fig. 9.1, c). Spiral bevel gears (Fig. 9.1, d) are the bevel counterpart of the helical gears. Their teeth are curved and oblique.
b
Fig. 9.2 (from [9.2])
Hypoid gears (Fig. 9.2, a) are like spiral bevel gears, but their pitch surfaces are hyperboloids rather than cones, and their pitch axes do not intersect. They operate more smoothly and quietly and are stronger for a given ratio. Crossed helical gears (Fig. 9.2, b), also known as spiral gears, are ordinary helical gears used in nonparallel shaft applications. The worm gearset (Fig. 9.2, c) consists of a worm, which resembles a screw, and a worm wheel, which is a helical gear, with the respective shafts 90 0 apart. They are quiet and vibration free, with lower Hertz contact stresses than the crossed-helical gears.
9. GEARS
41
The resultant force vector between a pair of meshing gears acts along the pressure line (also called line of action or generating line). The pressure line is tangent at points c and d to the base circles.
The angle between the pressure line and the common tangent to the pitch circles is the pressure angle, and it usually has values of 20 or 25 deg. The operating diameters of the pitch circles depend on the center distance used in mounting the gears, but the base circle diameters are constant and depend only on how the tooth forms are generated, because they form the base of the starting point on the involute profile.
42
DYNAMICS OF MACHINERY
Point a is the initial point of contact, where the flank of the pinion driving tooth just touches the tip of the driven tooth. This point is located at the intersection of the addendum circle of the gear with the pressure line. Should point a occur on the other side of point c on the pinion base circle, the pinion flank would be undercut during the generation of the profile. Point b is the final point of contact, when the tip of the driving tooth just leaves the flank of the driven tooth. This point is located at the intersection of the addendum circle of the pinion with the pressure line. For no undercutting of the gear teeth, point b must be located between the pitch point P and the tangent point d on the base circle of the gear. Line aP represents the approach phase of tooth contact, while line Pb is the recess phase. Tooth contact throughout the line of action ab is by both sliding and rolling, except for an instant at P when the contact is pure rolling. Sliding gives rise to friction forces that vary in magnitude and direction as the teeth go through the meshing cycle. During the approach action, the flank of the pinion tooth is sliding down the face of the gear tooth, producing a frictional force oriented upwards in Fig. 9.4. During the recess action, the face of the pinion tooth is sliding up the flank of the gear tooth, and the resulting friction force exerted by the pinion against the gear is oriented in opposite direction (downwards in Fig. 9.4). Friction forces produce a characteristic type of gear wear.
The zone of action of a pair of meshing gear teeth is shown in Fig. 9.5. The arc of action AB is the sum of the arc of approach AP and the arc of recess PB. In the unlikely situation in which the arc of action is exactly equal to the circular pitch, when one pair of teeth are just beginning contact at a, the preceding pair will be leaving contact at b. Thus, for this special condition, there is never more or less than one pair of teeth in contact. If the arc of action is greater than the circular pitch (their ratio is called the contact ratio) but less than twice as much, then when a pair of teeth come into contact at a, another pair of teeth will be still in contact somewhere along the line
9. GEARS
43
of action ab. Thus, for a short period of time, there will be two pairs of teeth in contact, one near the vicinity of A and another near B. As the meshing proceeds, the pair near B must cease contact, leaving only a single pair of contacting teeth, until the procedure repeats itself. Gears are not generally designed having contact ratios less than 1.20 because inaccuracies in mounting might reduce the contact ratio even more, increasing the possibility of impact between the teeth as well as an increase in the noise level. A contact ratio of 1.2 means 80 percent of the time single tooth contact, and 20 percent of the time double tooth contact.
The contact ratio is equal to the length of the line of action ab divided by the base pitch. The base pitch is the distance, measured on the line of action, from one involute to the next corresponding involute. In Fig. 9.6, a the mating teeth of the meshing spur gears are in contact at the pitch point. The number of tooth pairs in contact is shown in Fig. 9.6, b. The transition from single to double tooth contact produces variations in the mesh stiffness.
44
DYNAMICS OF MACHINERY
The tooth involute profiles are designed to produce a constant angular velocity ratio during meshing. Ideally, when two gears are in mesh, their pitch circles roll on one another without slipping. Denoting the pitch radii by r1 and r2 , and the angular velocities by 1 and 2 , the pitch line velocity is
v = r1 1 = r2 2 .
(9.1)
i=
2 r1 = . 1 r2
(9.2)
In order to transmit uniform rotary motion during meshing, a pair of gears must meet the following requirements (Fig. 9.7): a) the pitch point P must remain fixed on the line of centers O1O2 ; b) the lines of action for every instantaneous point of contact e must pass through the same point P; c) the generating (pressure) line must be always tangent to the base circles and normal to the involute profiles at the point of contact e. Deviations from the above requirements produce transmission errors giving rise to vibrations [9.6]. Changing the center distance, the above requirements are still satisfied, because it has no effect on the base circles used to generate the tooth profiles. Increasing the center distance increases the pressure angle and decreases the length of the line of action, but the teeth are still conjugate, and the angular velocity ratio
9. GEARS
45
is not changed. This increase creates two new operating pitch circles having larger pitch diameters but remaining tangent to each other at the pitch point. Interference might be produced by the contact of portions of tooth profiles which are not conjugate. It is eliminated by undercutting (which weakens the teeth), by using a larger pressure angle, or by increasing the number of teeth, hence increasing the pitch line velocity and making the gears noisier, which is an unacceptable solution.
fm = N f s .
(9.3)
For a pair of spur gears, if f s1 and f s 2 are the rotation frequencies of the two shafts, and N 1 and N2 are the corresponding number of teeth, the fundamental meshing frequency is the same for both gears in mesh
f m = N1 f s1 = N 2 f s 2 .
(9.4)
An epicyclic geartrain is shown in Fig. 9.8. It consists of three revolving planet pinions that engage the central sun gear and the coaxial ring gear with internal teeth, and a carrier in which the planet pinions are supported. For a planetary gear system, the following relationships can be used: the meshing frequency
f m = N s ( f s f c ) = Nr ( fc f r ) ,
(9.5)
DYNAMICS OF MACHINERY
fc =
Ns fs + Nr fr , Ns + Nr
(9.6)
where f r and N r are the speed (rps) and the number of teeth of the ring gear, and
f s and N s are the speed (rps) and the number of teeth of the sun gear.
In most planetary gear systems one of the elements is attached to the frame and has a zero input motion. Profile errors identical on each tooth, or deflection effects which are the same for each tooth mesh, produce vibrations with components at the tooth meshing frequency and its harmonics.
9. GEARS
47
tooth A is deflected back further and tooth D will momentarily lag. The final point of contact b is where the addendum circle of the driver gear crosses the pressure line.
This tooth deflection is very load dependent. For condition monitoring purposes it is necessary to make measurements always at the same loading, and this loading should be sufficient to ensure that the teeth are permanently in contact, and not able to move into the backlash.
When the point of contact of the engaging teeth reaches the pitch point, the direction of sliding reverses, causing a shock sometimes referred to as the pitch-circle impulse which is perpendicular to the axes of rotation of the two
48
DYNAMICS OF MACHINERY
gears. The two shafts are then subjected to bending stress reversals at the rate of the product of shaft speed and number of teeth. When a new pair of teeth takes part in the transmission of load, the driven gear retrieves its retardation by a renewed forward acceleration. It is subjected to an engagement shock, the impulse acting in a tangential direction at a rate of the product of rotational speed and number of teeth. These impulses cause the transmitted torque to fluctuate about a mean level, with variations of the angular velocity, producing a frequency modulation of the tooth-meshing frequency. The pitch circle and engagement vibrations are transmitted through the shaft and bearing housing causing casing vibrations. This vibration can be measured using an accelerometer mounted on the casing.
f (k , p , q ) = k f m p f s1 q f s 2 ,
k = 1, 2, .. ,
p , q = 0,1,2, .. . (9.7)
9. GEARS
49
Fig. 9.11
The existence of complex phase and amplitude modulation may also be interpreted as a nonlinear or cyclostationary phenomenon.
Amplitude modulation
When the excitation due to the tooth engagement occurs simultaneously with excitations having a frequency of once or twice per gearwheel rotation, amplitude modulation (multiplicative) effects are produced (Fig. 9.12).
Fig. 9.12
Typical once per revolution excitations are produced by: a) the accumulative effect of the pitch error; b) an isolated error of the tooth form; c) debris trapped in the teeth; d) eccentricity of mounting the gear wheel; e) load variation and f) unbalance. Typical twice per revolution excitations are produced by misalignment and wheel distortion (ovality).
50
DYNAMICS OF MACHINERY
Amplitude modulation of a carrier frequency by a lower frequency gives rise to a pair of sidebands in the frequency spectrum, spaced on either side of the carrier frequency by an amount equal to the modulating frequency.
by a lower-frequency cosine (plus d-c component), um (t ) . U c ( f ) and U m ( f ) are the corresponding Fourier spectra obtained by a forward Fourier transform. The final resulting spectrum consists of the carrier frequency plus two sidebands spaced at an amount equal to the modulating frequency. Indeed, transformation of the product of cosines in a sum yields
cos c t cos m t = 1 [ cos (c + m )t + cos (c m )t ] , 2 1 i c t e + e i c t , 2
(9.8)
where
cos c t =
and
c = 2 f c , m = 2 f m .
9. GEARS
51
Considering only amplitude modulation effects, Fig. 9.14 shows the effect of fault distribution on the sideband pattern. A very localized fault, e.g., on one tooth, would tend to give a modulation by a short pulse of length of the order of the tooth mesh period, but repeated once every revolution. Figure 9.14, a shows how this in the spectrum would result in the generation of a large number of sidebands of almost uniform level.
The effect of a more distributed fault is shown in Fig. 9.14, b. It is seen that as the envelope of the fault in the time signal becomes wider, it makes the corresponding envelope in the frequency domain narrower and higher. The resulting modulation products become more obviously sidebands grouped around the tooth meshing harmonics.
Frequency modulation
When the rotational speed of the gears is not constant, and the tooth spacing is not perfectly uniform, a frequency modulation of the tooth meshing frequency occurs. In fact, the same fluctuations in the tooth contact pressure which give rise to amplitude modulation apply a fluctuating torque to the gears, producing angular velocity fluctuations at the same frequency. Frequency modulation, even by a single frequency f1 , gives rise to a whole family of sidebands with a spacing equal to the modulating frequency, i.e., the same frequencies as produced by amplitude modulation by a distorted periodic signal (Fig. 9.15). Since in gears the two effects are virtually inseparable, the
52
DYNAMICS OF MACHINERY
resulting spectrum is a combination of the sidebands produced by both amplitude and frequency modulation.
Fig. 9.15
(9.9)
where = 1 is the modulation index. It represents the maximum deviation The decomposition into cosine components, and further decomposition into positive and negative frequency components, yields (giving details of the positive frequency components alone)
a=
A i t i ( + ) t i ( 0 1 ) t { C0 ( ) e 0 + C1 ( ) [ e 0 1 e ]+ 2
i ( 0 + 2 1) t
+ C2 ( ) [ e
i ( 0 2 1) t
(9.10)
The relative amplitude of the carrier frequency component is given by C0 ( ) and that of the nth order sidebands by Cn ( ) . Assuming in a gearbox that the carrier frequency f 0 is the tooth meshing frequency and the modulating frequency f1 is the rotating speed of the gear, then
f
f1
f f0
f 0 f1
=N ,
(9.11)
where is the relative speed fluctuation of the gear, and N is the number of teeth on the gear.
9. GEARS
53
It can be shown that, for << 1 only one pair of sidebands are important, while for < 1 most information is contained in the first two pairs of sidebands (Fig. 9.16). Frequency modulation tends to modify the relative amplitude of the sidebands produced by the amplitude modulation. Its additional effect is to increase the number of sidebands somewhat and to make the sideband patterns unsymmetrical by reinforcement/cancellation because of the different phase relationships of the sidebands.
54
DYNAMICS OF MACHINERY
For main transmission gears, if the gear assembly is located at a node of vibration and there are gear errors attempting to force a vibratory movement at that point, very large dynamic loads will be produced. By placing the gears away from a nodal point, the vibratory torque at the gears may be appreciably reduced. In the case of auxiliary drives, it is usually preferable to take off power at a node of free torsional vibration, since the vibration amplitude at a node is small and is therefore less likely to cause noise and wear or interfere with the functioning of a comparatively lightly loaded auxiliary system. In heavily loaded gears, the use of tip and/or root relief is usually desirable to accommodate tooth deflection under load. With helical gearing, tooth errors can excite torsional and axial vibrations or a combination of the two. In double-helical gearing, the type of vibration excited depends upon the phasing of errors in the right- and left-hand helices. If errors are in phase the tendency is to excite torsional vibrations. If the errors are 180 0 out of phase, the tendency is to excite axial vibrations. In difficult cases of resonance, it might be desirable to change the number of teeth in all gears whose meshing is responsible for the disturbance or, alternatively, modify the structural elements which are responding to the disturbance.
9. GEARS
55
Similar effects are produced by variations in mesh stiffness, further influenced by the bearing clearance, shaft and housing deflections and manufacturing errors. Teeth which have pitch or profile errors are prone to tooth separation, high impact loads and failure. This occurs especially for lightly loaded gears where the stress at the root of a tooth and the vibration may be excessive at critical speeds.
56
DYNAMICS OF MACHINERY
Frosting applies to fully hardened gear teeth profiles defined by the existence of small round or elliptical patches which (under high magnification) exhibit the general appearance of minute scorings. Destructive wear results in a corrosive change in the involute shape of the gear tooth. It would be accompanied by extremely rough operation, non-uniform motion and shock overloads, which would probably result in tooth breakage. Corrosive wear generally occurs in long lasting operations in the presence of chlorine or sodium based lubricants. This should not be confused with ordinary oxidation corrosion which may occur during no operation intervals due to improper preservation. It may be difficult to detect if the resultant pitting is fine. Interference wear is the effect of the tip of one gear tooth tip in contacting the fillet or root area of its mating tooth. When the initial contact occurs, the root of the pinion tooth engages with the tip of the wheel tooth. Deflection occurs at the point of maximum relative sliding speed, so that the tip may dig in and cause wear of the pinion root. Conversely, during recession, the tip of the pinion tooth deflects as it disengages from the wheel tooth causing wear of the wheel root. To overcome this, the tips of both pinion and wheel teeth should be carefully and accurately relieved. Burning indicates surface tempering or softening of the tooth, most probably accompanied by a total loss of lubricant. It is an advanced condition of discoloring. Discoloration is a term used to locate the existence of surface temper colouring of the active profile (the top band) of a gear tooth. The condition indicates a marginal lack of lubrication or an excessive power operation. Misalignment wear arises from operation at skewed axes.
9. GEARS
57
Micropitting occurs on surface-hardened gears and is characterized by extremely tiny pits, approximately 10 m deep. Spalling is a pitting condition whose origin can be physically detected at the apex of the fan-shaped portion of the damaged area. This is a surface-initiated type of fatigue, which has its origin in the surface tensile cracking which leads to the gradual erosion and exfoliation of increasingly larger pieces of gear material as the fan widens out in the direction of sliding action. The cracks will ultimately undermine the entire case of case-hardened gear teeth as the spalling approaches the extremities of the addendum.
Arrested pitting describes very small shallow pits that are not propagating into larger failure areas. It has been observed on spiral bevel gears and is frequently associated with the waviness condition referred to as barber pole. This pitting is often considered corrective in that it progresses immediately to the point of relieving local compressive stress of overload. Pitch line pitting belongs to the family of rolling contact fatigue and is truly subsurface in origin. It is not generally associated with a condition of lubrication distress but generally occurs at relatively high cycles of loading. In fully hardened, properly designed gears, it is seldom seen in less than 100,000 cycles of operation. Addendum pitting and dedendum pitting are terms which merely signify the site of origin of one of the foregoing types of pitting or spalling. Case crushing means shear failure of the core-case interface in casehardened gear teeth. It indicates insufficient case depth for the load magnitude.
58
DYNAMICS OF MACHINERY
Multiple cracking, often both transverse and longitudinal, is generally observed in the tooth flank. The Hertzian shear stress increases from zero at the surface to a maximum at a depth depending upon: a) surface load concentration; b) contact length at the surface, and c) relative curvature of the surfaces in contact. Case depth is the distance from the surface to the position where the hardness reaches that of the core material. For gear teeth, the maximum shear stress should be within the depth of hard casing.
High local temperatures arising from the grinding of hardened teeth without adequate coolant leaves the surface layers in a state of tension. In severe cases, grinding cracks may form at the root of tooth fillets. Such cracks act as nuclei for tooth fracture.
9. GEARS
59
Feature extraction is the process of extracting condition indicators (metrics) about the system input which are more informative than evaluating the raw input itself. It is a parametrization process, which often reduces the data volume. Feature extractors output only the information relevant for detecting the failure modes to which the associated components are susceptible. In vibration monitoring, damage detection techniques based on the Discrete Fourier Transform (DFT) have been traditionally developed. The absolute value of DFT contains an estimate of the signal power spectrum, which displays different behavior between health state and damaged state signals. However, the absolute DFT is insensitive to the shaft phase offset, which is random and thus a source of unwanted variation in signal characteristics. Given the geometry of rolling element bearings, it was possible to predict which frequencies, i.e. DFT coefficients, are affected by different failure modes. For gear systems, some fault-indicating signal characteristics are not well captured by the DFT, but are better enhanced by other transforms. Several condition indicators (figures of merit) have been introduced to detect localized damage in gears [9.11-9.13]. Ideally, these features are more stable and well behaved than the raw signature data itself.
60
DYNAMICS OF MACHINERY
Further preprocessing calculates the residual signal, which consists of the TSA signal with the primary meshing and driveshaft components along with their harmonics removed. Good results are obtained by high passing the data about some frequency and remove the meshing frequency and all harmonics. The cut-off frequency of the high pass filter is system dependent, and is selected between d.c. and the fundamental meshing frequency. Indicators NA4 (Zakrajsek, 1993) and NA4* (Decker, 1994) are determined based on the residual signal. The condition indicators FM4 (Stewart, 1977), M6A and M8A (Martin, 1989) are based on the difference signal, calculated by removing the regular meshing (RM) components from the TSA signal. The RM components consist of the driveshaft frequency and its harmonics, the primary meshing frequency and harmonics, along with the first order sidebands. It turns out that the difference signal can be obtained by removing the sidebands of the primary meshing frequencies from the residual signal. The NB4 condition indicator (Zakrajsek, 1994) is obtained from the bandpass mesh (BPM) signal. The TSA signal is band-pass filtered around the primary gear mesh frequency, including as many sidebands as possible. The Hilbert transform is then applied to the filtered signal to produce a complex time series.
9. GEARS
61
The real part is the band-passed signal and the imaginary part is the Hilbert transform of the signal. The envelope is the magnitude of this complex time signal and represents an estimate of the amplitude modulation present in the signal due to the sidebands. Kurtosis and r.m.s. may be performed at different preprocessing levels, while Demodulation and Enveloping may return multiple parameters.
The root mean square (r.m.s.) is a measure of the power content of the vibration signal. It is a general fault indicator, which provides no information on which component is failing, and shows no appreciable changes in the early stages of gear damage. Alone, it can be effective only in detecting major out-of-balance. The r.m.s. of a digital signal defined by a data series xn over length N is defined as
r .m.s . = 1 N
x
n =1
2 n
(9.12)
Delta r.m.s. is simply the difference between the current r.m.s. value and the previous one. This parameter focuses on the trend of vibration signal and is sensitive to changes in the vibration signal. Theoretically it allows selection of an alarm level which is not sensitive to load, however in practice it came out that it is sensitive to load change.
Kurtosis
Kurtosis is defined as the fourth moment of the distribution (about the mean), normalized by the square of the variance. It measures the relative peakedness or flatness of a distribution as compared to a normal distribution. Kurtosis provides a measure of the size of the tails of distribution and is used as an indicator of major peaks in a set of data. As a gear wears and eventually a tooth breaks, this feature should signal a defect due to the increased level of vibration. The equation for Kurtosis is given by
62
DYNAMICS OF MACHINERY
kurt =
1 N
n =1
( x n x )4
N
(9.13)
where x is the mean of the data and 2 is the variance. A more detailed presentation of the Kurtosis is given in Chapter 8 in connection to the condition monitoring of rolling element bearings.
Crest Factor
The Crest Factor is defined as the ratio of the peak level to the r.m.s. level of the signal [9.14]. In early stages of damage, there is no change in the r.m.s. value, while the peak value increases, therefore the Crest Factor increases. As the damage progresses, the r.m.s. value increases and the Crest Factor decreases. It is used to detect changes in the signal pattern due to tooth breakage, but is not considered a very sensitive indicator. A presentation of the Crest Factor is given in Chapter 8.
Energy Operator
The Energy Operator is defined as the normalized Kurtosis of a signal in which each point is computed as the difference of two squared neighborhood points of the original signal
EO =
1 N 1 N
( s n s )4
N n =1 N n =1
s n s 2
(9.14)
2 2 where s is the mean value of signal s , sn = xn +1 xn , and N is the number of points in the dataset x. In the case of endpoints, the data is looped around [9.15]. Specifically, when calculating the first point, the last point is used and vice versa.
Enveloping
Enveloping is used to monitor the high-frequency response of a gearbox to periodic impacts produced when a faulty tooth makes contact with the mating tooth. These impacts usually excite a resonance in the system at a much higher frequency than the vibrations generated by the other components. The corresponding high frequency energy is usually concentrated in a narrow frequency band. Tooth wear and breakage increase the amplitude of side bands near critical frequencies such as the output shaft frequency.
9. GEARS
63
The enveloping technique, already presented in Chapter 8 for rolling element bearings [9.11], consists of processing the structure resonance energy with an envelope detector (Fig. 8.17).
Demodulation
When the teeth wear, the relative sliding results in a change of amplitude or amplitude modulation of the vibrations at the gear meshing frequency f m and its harmonics. Demodulation identifies the periodicity in the modulation of the carrier. The carriers are basically f m and 2 f m . Demodulation techniques detect the amplitude modulation components induced by the gear wear at these frequencies. This differs from enveloping which detects the combined effects over a range of frequencies. The raw data is high-passed filtered at 0.85 f m and lowpassed filtered at 1.15 f m . The power spectral density of the filtered signal is searched to obtain the actual carrier frequency f m . The actual carrier is used to amplitude demodulate the filtered carrier signal. The power spectral density of the resulting signal is searched within 5% of the output shaft frequency. The condition indicators extracted from this technique are the frequency of the peak and the magnitude squared amplitude.
FM 0
Major tooth faults typically result in an increase of the peak-to-peak signal levels, but do not change the meshing frequency. The zero-order figure of merit FM 0 is defined as the peak-to-peak level of the TSA signal divided by the sum of the amplitudes at the gear-mesh frequency and its harmonics [9.16]. While the Crest Factor compares the peak value of the TSA signal to the energy of the TSA signal, the FM 0 compares the peak value of the TSA signal to the energy of the row signal. The equation for FM 0 is
FM 0 =
PPA
ak
k =1
(9.15)
where PPA is the peak-to-peak amplitude of the TSA waveform and a k is the amplitude of the kth mesh frequency harmonics.
64
DYNAMICS OF MACHINERY
FM 4 The indicator FM 4 was developed to detect changes in the vibration pattern resulting from damage on a limited number of gear teeth [9.16]. FM 4 is calculated as the Kurtosis of the difference signal 1 N 1 N
FM 4 =
n =1 N
(d n d )
N
4 2
n =1
(d n d )
(9.16)
where d is the difference signal, d is the mean value of the difference signal, and N is the total number of data points in the time record.
9. GEARS
65
The difference signal is obtained by removing from the original signal the gear meshing frequency, harmonics and first order sidebands. A flowchart for calculating FM 4 is shown in Fig. 9.20. It is assumed that a difference signal from a gear in good condition has a Gaussian amplitude distribution, therefore resulting in a normalized Kurtosis value of 3.0. As a defect develops in a tooth, such as a crack or pitting, peaks will grow in the difference signal resulting in a less peaked amplitude distribution with a Kurtosis value increasing beyond 3.0, typically larger than 7.0. If more than one tooth is defective, the data distribution becomes flat and the Kurtosis value decreases. NA4 The NA4 indicator was developed to improve the behavior of the FM 4 indicator when more than one tooth is damaged [9.17]. It is determined by dividing the fourth statistical moment of the residual signal by the current run time averaged variance of the residual signal, raised to the second power. The equation for NA4 is 1 N 1 M
NA4 =
n =1
( rn r )4
N n =1
1 N m =1
( r n ,m rm )
N
(9.17)
where r is the residual signal, r is the mean value of the residual signal, N is the total number of data points in the time record, and m is the current time record number in the run ensemble.
NA4 was developed to detect the onset of damage and to continue to react to this damage as it spreads and increases in magnitude [9.18]. If the gear damage spreads from one tooth to another tooth, NA4 grows a) because the first order sidebands increase, and b) the value of the average variance at the denominator increases slower than the numerator.
NA4 *
NA4 * (or ENA4 ) was developed as an enhanced version of NA4 , and was expected to be more robust when progressive damage occurs [9.19]. This added robustness is obtained by normalizing the fourth statistical moment with the residual signal variance for a gearbox in good condition, instead of the running variance, which is used for NA4 . This overcomes the rapid increase of the averaged variance at the denominator of equation (9.16) when the gear damage progresses.
DYNAMICS OF MACHINERY
NA4* =
n =1
( rn r )4
N
~ 2 (M 2 )
(9.18)
~ where M 2 is the variance of the residual signal for a gearbox in good condition. Energy Ratio Heavy uniform wear can be detected by the Energy Ratio [9.14]. It compares the energy contained in the difference signal, d, to the energy contained in the regular meshing (RM) signal
ER =
d , RM
(9.19)
where denotes the standard deviation. The basic idea is that the energy is transferred from the regular meshing component to the rest of the signal as wear progresses.
M 6 A and M 8 A
The theory behind M 6 A and M 8 A is the same as that for FM 4 , except M 6 A that and M 8 A are expected to be more sensitive to peaks in the difference signal. The M 6 A indicator is determined by dividing the sixth statistical moment about mean of the difference signal by the cube of variance of the difference signal. The M 8 A indicator is obtained by dividing the eighth statistical moment about mean of the difference signal by the fourth power of variance of the difference signal [9.20]. The equations for M 6 A and M 8 A are as follows: 1 N 1 N
6 (d n d )
N
M 6A =
n =1 N
n =1
2 dn d
M 8A =
1 N 1 N
n =1 N
(d n d )
N
8 4
n =1
2 dn d
. (9.20)
For a gear in good condition M 6 A = 15 and M 8 A = 105 . As a defect develops in a tooth, M 6 A increases beyond 45 and M 8 A increases beyond 300.
9. GEARS
67 NB 4
The NB 4 indicator is similar to NA4 except that, instead of using the residual signal, NB 4 uses the envelope of a band-passed segment of the TSA signal [9.21]. The idea behind this method is that a few damaged gear teeth will cause transient load fluctuations that are different from the normal tooth load fluctuations. The theory suggests that these fluctuations will be manifested in the envelope of a signal which is band-pass filtered about the dominant meshing frequency. The latter is either the primary meshing frequency or one of its harmonics, whichever appears to give the most robust group of sidebands. complex (i.e., analytic) signal, a (t ) + i H [ a (t )] , obtained by applying the Hilbert transform
H [ a ( t )] = d a ( ) t 1
(9.21)
(a (t ) ) 2 + H [a (t ) ] 2
(9.22)
NB 4 is then determined by dividing the fourth statistical moment about mean of this envelope signal by the square of average variance of the envelope of band-passed signals up to current time, with the following equation
1 N 1 M 1 N m =1
NB 4 =
n =1
( s n s )4
N n =1
( s n ,m sm )
N
(9.23)
where s is the envelope of the band-passed signal and s is its mean value.
68
DYNAMICS OF MACHINERY
lubricating oil. By detecting the number and size of particles in the oil one can identify gear pitting damage in an early stage, which is unidentifiable by vibration methods. Oil debris sensors are usually based on a magnetic or an optical principle. Magnetic sensors measure the change in magnetic field caused by metal particles in a monitored sample of oil. A disadvantage of oil debris analysis is that it does not localize the failure in complex gearboxes. The oil debris sensor records counts of particles in bins set at different particle size ranges. For each bin size range, an average particle size is first determined. Then statistical distribution methods are applied to particles collected from the lubrication system. The mean particle size is calculated as
E ( ) =
i =1
P [ i ] ,
(9.24)
where i is the average bin size, i the number of bins, and P [ i ] is the number of particles per average bin size per reading divided by the total number of particles per reading. The Variance is
Variance =
The Kurtosis is
[
i =1
E ( ) ] 2 P [ i ] .
(9.25)
Kurtosis =
The relative Kurtosis is
[
i =1
E ( ) ] 4 P [ i ] .
(9.26)
Re lative Kurtosis =
(Variance ) 2
Kurtosis
(9.27)
Laboratory experiments have shown that oil debris analysis is more reliable than vibration analysis for detecting pitting fatigue failure of spur gears. The increase in oil debris mass is related to damage progression, which is not detected by some vibration based condition indicators.
9. GEARS
69
In order to extract an intelligent feature from the accumulated mass measured by the oil debris sensor, a fuzzy logic analysis can be used. Integrating oil debris analysis and vibration measurement results in a monitoring system with improved damage detection and decision-making capabilities.
C ( ) = F 1 { log [ G ( f ) ] } ,
where G ( f ) is a frequency spectrum.
(9.28)
Thus, the cepstrum is a spectrum of a spectrum and for this reason, the name cepstrum was coined from spectrum by reversing the first syllabe. Other terms are coined in similar way, quefrency from frequency, rahmonic from harmonic, gamnitude from magnitude, saphe from phase, quefrency alanysis from frequency analysis, etc. [9.23].
70
DYNAMICS OF MACHINERY
signal g x (t ) , i.e. G xx ( f ) = F { g x (t ) } and F{ } represents the forward Fourier Transform of the bracketed quantity, the resulting cepstrum is termed a power cepstrum [9.24], defined by
2
C ( ) = F 1 { log [ Gxx ( f )] } .
(9.29)
Fourier Transform of a time signal g (t ) , the resulting cepstrum is termed a complex cepstrum, defined by equation (9.28) but where
G ( f ) = A ( f ) e i ( f ) ,
and
(9.30)
ln [ G ( f ) ] = ln [ A ( f ) ] + i ( f ) .
(9.31)
The independent variable, , of the cepstrum has the dimensions of time, but is known as quefrency. A high quefrency represents rapid fluctuations in the spectrum (small frequency spacings) and a low quefrency represents slow changes with frequency (large frequency spacings). When peaks in the cepstrum result from families of sidebands, the quefrency of the peak represents the time period of the modulation. Its reciprocal is the modulation frequency. Note that the quefrency says nothing about the absolute frequency, only about the frequency spacings.
Figure 9.22 shows the results of this type of analysis for a gearbox. The spectrum (Fig. 9.22, a) contains a large number of sidebands, but their spacing is difficult to determine. Within the display range of the cepstrum (0 30 ms) the first three rahmonics of the 8.28 ms (120.75 Hz) component and only the first rahmonic of the 20.1 ms (49.75 Hz) component are present (Fig. 9.22, b). The periodicity is not apparent in the frequency spectrum since the mixture of the two periodicities gives a quasi-periodic structure.
9. GEARS
71
Figure 9.23, a is a 400-line baseband spectrum from 0-20kHz of a gearbox vibration signal containing at least the first three harmonics of the toothmeshing frequency (4.3 kHz). Figure 9.23, b is a 2000-line composite spectrum extending in frequency from below the toothmesh frequency to above its third harmonic (3.5-13.5 kHz). It excludes the low harmonics of both shaft speeds. This degree of resolution is required to separate the individual sidebands with spacings equal to the shaft speeds, but the eye cannot see the spectrum details.
Figure 9.23, c shows the expanded 400-line section from 7500 to 9500 Hz. The eye still cannot readily see the sideband families because of the mixture of different spacings. The amplitude cepstrum (Fig. 9.23, d ) of the whole one-sided spectrum (Fig. 9.23, a), reveals that all rahmonics come from one of two families, corresponding to the speeds of the two gears in this particular gearbox (50 Hz and 85 Hz).
72
DYNAMICS OF MACHINERY
The cepstrum is considered to be an extremely useful tool for two tasks in vibration monitoring and analysis [9.25]: For fault detection: a) it is a sensitive measure of the growth of harmonic/sideband families; b) the data is reduced to a single line per family; c) it is insensitive to the location of the measurement point, to the phase combination of amplitude and frequency, and to gearbox loading. For fault diagnosis: a) it is an accurate measure of the spacing of frequency components; b) it can be calculated from any section of a spectrum; c) it can be used for the separation of different families of sidebands; and d) it is sensitive to tooth and blade differences but not to uniform wear.
References
9.1. Sidahmed, M. and Dalpiaz, G., Signal generation models for diagnostics, Encyclopedia of Vibration, Braun, S., Ewins, D. and Rao, S.S., eds., Academic Press, London, 2002, p.1184-1193. 9.2. Coy, J. J., Townsend, D. P. and Zaretsky, E. V., Gearing, NASA/RP-1152, 1985. 9.3. Shigley, J. E. and Mischke, C. R., Gearing. A Mechanical Designers Handbook, McGraw-Hill, New York, 1990. 9.4. Shigley, J. E., Mechanical Engineering Design, 2nd ed., McGraw-Hill Kogakusha Ltd., Tokyo, 1972.
9. GEARS
73
9.5. Collacott, R. A., Gear faults diagnostics, U.K. Mechanical Health Monitoring Group, Leicester Polytechnic, Nov. 1975. 9.6. Mark, W., Analysis of the vibratory excitation of gear systems: basic theory, Journal of the Acoustical Society of America, vol.65, 1978, p.1409-1430. 9.7. Randall, R. B., A new method of modeling gear faults, Journal of Mechanical Design, Trans. ASME, vol.104, April 1982, p.259-267. 9.8. Wilson, W. Ker, Practical Solution of Torsional Vibration Problems, Chapman & Hall, London, 1956. 9.9. Dempsey, P. J., Integrating oil debris and vibration measurements for intelligent machine health monitoring, NASA/TM-2003-211307. 9.10. Choi, S. and Li, C. J., Estimate gear tooth transverse crack size from vibration by fusing selected gear condition indices, Measurement Science and Technology, vol.17, 2006, p.1-6. 9.11. Lebold, M., McClintic, K., Campbell, R., Byington, C. and Maynard, K., Review of vibration analysis methods for gearbox diagnostics and prognostics, Proc. 54th Meeting of the Society for Machinery Failure Prevention Technology, Virginia Beach, VA, May 1-4, 2000, p.623-634. 9.12. Mosher, M. Pryor, A.H. and Huff, E.M., Evaluation of standard gear metrics in helicopter flight operation, 56th Mechanical Failure Prevention Technology Conference, Virginia Beach, VA, April 15-19, 2002. 9.13. Dempsey, P., Lewicki, D. G. and Le, Dy D., Investigation of current methods to identify helicopter gear health, NASA/TM-2007-214664. 9.14. Swansson, N. S., Applications of vibration signal analysis techniques to signal monitoring, Conf. on Friction and Wear in Engineering, Barton, Australia, 1980. 9.15. Ma, J., Energy operator and other demodulation approaches to gear defect detection, Proc. 49th Meeting of Soc. for Mechanical Failure Prevention Technology, Virginia Beach, VA, April 1995. 9.16. Stewart, R. M., Some useful data analysis techniques for gearbox diagnostics, Report MHM/R/10/77, Machine Health Monitoring Group, I.S.V.R., Univ. of Southampton, July 1977. 9.17. Zakrajsek, J. J., An investigation of gear mesh failure prediction techniques, NASA TM-102340, Nov.1989. 9.18. Zakrajsek, J. J., Townsend, D. P. and Decker, H. J., An analysis of gear fault detection methods as applied to pitting fatigue failure data, NASA TM-105950, April 1993. 9.19. Decker, H. J., Handschuh, R. F. and Zakrajsek, J. J., An enhancement to the NA4 gear vibration diagnostic parameter, NASA TM-106553, June 1994.
74
DYNAMICS OF MACHINERY
9.20. Martin, H. R., Statistical moment analysis as a means of surface damage detection, Proc. 7th International Modal Analysis Conference, Schenectady, New York, Jan 1989, p.1016-1021. 9.21. Zakrajsek, J. J., Handschuh, R. F. and Decker, H. J., Application of fault detection techniques to spiral bevel gear fatigue data, Proc. 48th Meeting of the Society for Machinery Failure Prevention Technology, Wakefield, MA, April 1994. 9.22. Randall, R. B., Cepstrum analysis and gearbox fault diagnosis, Brel&Kjaer Application Note No. 233-80. 9.23. Bogert, B. P., Healy, M. J. R. and Tukey, J. W., The quefrency alanysis of time series for echoes: cepstrum, pseudo-autocovariance, cross-cepstrum, and saphe cracking, Proc. Symp. Time Series Analysis, Rosenblatt, M., ed., Wiley, New York, 1963, p.209-243. 9.24. Randall, R. B., Advanced machine diagnostics, Shock and Vibration Digest, vol.29, no.6, 1997, p.6-26. 9.25. * Primer for Cepstrum analysis a powerful tool for simpler diagnosis of REB and gear vibrations, Brel & Kjaer Application Note No. BAN0026 -EN11. 9.26. Cohen, L., The Time-Frequency Analysis, Prentice-Hall, New Jersey, 1995. 9.27. Kaiser, G., A Friendly Guide to Wavelets, Birkhuser, Boston, 1994. 9.28. Wang, W. J. and McFadden, P. D., Early detection of gear failure by vibration analysis. Calculation of the time-frequency distribution, Mechanical Systems and Signal Processing, vol.17, 1993, p.193-203. 9.29. Dalpiaz, G., Rivola, A. and Rubini, R., Effectiveness and sensitivity of vibration processing techniques for local fault detection in gears, Mechanical Systems and Signal Processing, vol.14, no.3, 2000, p.387-412. 9.30. Gade, S. and Gram-Hansen, K., Non-stationary signal analysis using Wavelet Transform, Short-time Fourier Transform and Wigner-Ville distribution, Brel & Kjaer Technical Review, no.2, 1996.
10.
VIBRATION MEASUREMENT
This chapter describes instrumentation and procedures used for the measurement and evaluation of vibrations in machine condition monitoring and fault diagnostics.
76
DYNAMICS OF MACHINERY
When the machine has a relatively light rotor operating in a heavy stiff casing (Fig. 10.1, a), most of the energy generated by the rotor is dissipated in relative motion between the shaft and the bearing. On machines of this type (such as high pressure centrifugal compressors) with casing to rotor weight ratios of 30:1 or more, the relative displacement between shaft and bearing measured with a noncontacting probe is the best indicator of the machine condition.
77
With the opposite configuration, a relatively heavy rotor running in stiff bearings supported on flexible structure (Fig. 10.1, b), most of the energy developed by the rotor is dissipated in the structural motion. On this type of machine (fans, aircraft derivative gas turbines and machinery fitted with rolling element bearings), the velocity of the casing vibration is the best measure of condition.
Displacement probes should be mounted in the same plane and facing in the same direction at each bearing of a multi case machine string. Since it is often impossible to mount a displacement probe in the horizontal plane, due to interference from the horizontal splits of the bearing and casing, current practice is to mount both probes in the upper half of the bearing, 45 degrees either side of the vertical centerline. In this configuration, the probe on the right (looking from the driver end) is arbitrarily called the horizontal probe and applied to the horizontal axis of an oscilloscope to establish the correct orbital motion. The probe on the left is called the vertical probe. One must of course remember that the oscilloscope display is tilted 45 degrees from actual shaft motion. Apart from the probes measuring the shaft radial displacement, a phase reference probe is usually installed on each shaft. It is a standard displacement probe, located so that it can observe a once-per-revolution mark on the shaft, such as a keyway or a hole. The width of the mark should be at least twice the probe tip
78
DYNAMICS OF MACHINERY
diameter and a minimum of 3 mm deep. The probe should be gapped closer than a standard displacement probe, to generate a high spike at the output of the oscillator demodulator. The spike may then be applied to the Z axis of an oscilloscope to produce a blank spot for a phase reference in the waveform and orbital presentation. The phase mark may be fed to a tachometer for speed indication. It is used either as a reference for the horizontal axis of a spectrum plot to construct order plots, or for phase measurement in balancing, or to calculate a correction for shaft runout.
Figure 10.3 shows the recommended measuring points for pedestal bearings, while Fig. 10.4 shows the measuring points for housing-type bearings. Figure 10.5 shows the measuring points for small electrical machines. For vertical in-line reciprocating engines, the measuring points are shown in Fig. 10.6, where L and R define the left-hand and right-hand when facing the coupling flange, 1 the machine end of mounting, 2 the crankshaft level, 3 the top edge of frame, .1 the coupling end, .2 the mid machine, and .3 the free end of machine.
79
Preferred measurement positions are specified in Fig. 10.7 for multicylinder Vee engines and in Fig. 10.8 for a horizontal opposed cylinder machine.
The major precaution to be exercised in making measurements of casing vibrations is to ensure that the transducer mounting is solid and does not have a natural frequency within the frequency range to be examined. In general, cantilevered mounts should be avoided as well as mounting on inspection covers or unsupported areas of bearing caps.
80
DYNAMICS OF MACHINERY
velocity or acceleration) and value (zero-to-peak, peak-to-peak or r.m.s.), as well as the transducer which must be used.
If the motion consists of the fundamental frequency and the first harmonic, the displacement components x(t) and y(t) recorded along the two directions are periodic, and the precession orbit is as shown in Fig. 10.10. The maximum precession radius smax is a measure of the shaft vibration severity, as defined in the recommendations VDI 2059 [10.5]. When measurements are made at bearings, this value can be compared with the bearing clearance. In the standard ISO 7919 [10.6], which superseeded VDI 2059, the shaft vibration magnitude is defined as the higher value of the peak-to-peak displacement measured in two selected orthogonal measurement directions,
81
Fig. 10.11
When the rotating assembly is five or more times heavier than the case of the machine, the shaft absolute displacement is of interest. It can be measured in two ways: a) electronically summing the signals of both an eddy current probe measuring relative shaft motion with respect to the bearing, and an accelerometer measuring case absolute displacement (integrated twice) (Fig. 10.11), and b) using a shaft rider, which is a spring mounted device that physically rides on the surface of the shaft, normally a velocity sensor mounted on top of the shaft rider whose output is integrated electronically to displacement.
82
DYNAMICS OF MACHINERY
For most machine types, one value of vibration severity will characterize the vibratory state of that machine. However, for some machines this approach may be inadequate and the vibration severity should then be assessed independently for measurement positions at a number of locations.
83
acceleration would indicate the 1 3 X component to be only about 1 5 the amplitude of the 1X motion (Fig. 10.12).
Fig. 10.12
The shaft precession will more commonly generate these sub synchronous frequencies, and thus the amount of shaft deflection (measured in displacement), relative to the machine clearances, is the most important parameter to evaluate in terms of vibration severity and the significance of the various vibration frequency components.
84
DYNAMICS OF MACHINERY
The root mean square (r.m.s.) value of velocity may be calculated as follows
T
vr .m.s . =
1 T
v (t )dt ,
2 0
(10.1)
where v (t ) is the instantaneous value, and T is the sampling time, which is longer than the period of any of the major frequency components. For non-periodic steady-state vibrations, the r.m.s. value is defined as
1 t
v r .m .s . =
lim t
v (t )dt .
2 0
(10.2)
For harmonic vibrations of velocity amplitude v o , the following relations can be established
vr .m .s . =
1 v0 = 0.707 v0 , 2
v p = v0 ,
v pp = 2 v0 .
(10.3)
Figure 10.13, b illustrates the addition of a harmonic component having the same amplitude but a frequency 10 times higher than the initial harmonic vibration. The resulting periodic vibration would have a peak value almost twice that of the initial component, but an r.m.s. value only 1.4 times larger. Figure 10.14 shows the influence of the relative phasing on the compounding of two harmonic components. The higher harmonic has half the amplitude of the fundamental component and a frequency 3 times higher.
85
Though the peak values are different in the two cases, v2 p 1.4 v1 p , the r.m.s. values are the same, v2 r.m.s. = v1 r.m.s. . This means that the use of the r.m.s. vibration as a measure of the vibration severity gives better results in comparisons with allowable limit values than in detecting developing malfunctions by monitoring the change of broad-band vibration magnitude.
While the standard ISO 7919, based on VDI 2056, recommends the measurement of the r.m.s. velocity on the bearing cap, some API standards and [10.10] recommend the measurement of zero-to-peak velocity.
86
DYNAMICS OF MACHINERY
For a signal with a velocity level of 6 mm/s, the displacement amplitude is 1m at about 1000 Hz and disappears into the background noise of most commercially available measuring systems. In an extreme case, the 6 mm/s velocity at 10 kHz corresponds to an acceleration level of 400 m s 2 , i.e. approximately 40 g with a displacement of only 0.1m . Displacement is not an effective means of measuring high-frequency vibration because large forces are required at these frequencies to produce a measurable displacement and not because the instrumentation system is limited to a maximum frequency. Below
87
about 20 Hz, the displacement amplitude necessary to produce an easily identifiable acceleration signal is so large that endangers the transducer mechanical integrity. Thus, it is recommended to use displacement pickups from 0 to about 1000 Hz, velocity pickups from 10 to 3000 Hz and accelerometers from 20 Hz to well above 20 kHz, setting the lower limit for acceleration measurement at 0.4 m s 2 , and that for displacement at 2 m [10.1]. The trend is the extension of acceleration measurement at lower frequencies.
Once an accelerometer is selected for vibration measurement, the first critical consideration is to make sure it has a frequency range that includes the potential machine fault frequencies. The second key factor is to make sure the accelerometer will perform in the environment intended to be used. Some of the considered factors are: a) heat tolerance, b) moisture, c) chemical exposure, d) electrical interference, e) intrinsically safe requirements, and f) shock limit. Recommended ranges for displacement, velocity, and acceleration transducers are specified in [10.10] as well as frequency ranges of operation. Figure 10.16 illustrates the typical envelope ranges of velocity vs. frequency for several type transducers. Traditional vibration sensors fall into three main classes: a) noncontact displacement transducers, b) velocity pickups, and c) piezoelectric accelerometers [10.12].
88
DYNAMICS OF MACHINERY
The probe is a sensor that translates distance (gap) to voltage. It may be used to measure dynamic motion and static gap. The eddy probe is a flat pancake coil of wire molded into a universal threaded case. Usual probe tip diameter is 5 mm, with a body diameter of 8 mm and length of 25 mm.
Fig. 10.18
A typical eddy current transducer contains two coils: an active coil and a balance coil. The active coil senses the presence of a nearby conductive object,
89
while the balance coil is used for temperature compensation and to balance the output bridge circuit. The lead wire is a single conductor shielded cable [10.16]. The eddy probe driver generates a high frequency signal to the eddy probe and converts the return signal to a voltage which can then be displayed on a readout monitor and used for comparison of vibration levels with alarm set points. When the appropriate voltage is supplied to the eddy probe driver, it becomes an oscillator and generates a high frequency signal to the coil in the tip of the eddy probe. The coil creates a small magnetic field that induces eddy currents in metal targets (Fig. 10.18). These eddy currents absorb part of the energy and change the sensors oscillation amplitude. As the gap narrows, more and more energy is absorbed until finally the voltage output drops to zero at, or near, contact. A typical response graph of gap vs. voltage (Fig. 10.19) shows the sensitivity of 200 mV mil (8 mV m ) over the range of 100 mils (2.5 mm ) for a standard supply voltage of minus 24V of direct current. Note that even though the eddy probe has a response in excess of 2.5 mm , the linear response which ends the useful range stops at about 2.5 mm .
The direct current voltage output may be measured by a voltmeter, and by referring to the probe calibration curve the exact gap between the eddy probe and the observed surface can be determined. If the observed surface is moving, as with a precessing rotating shaft, the signal is not constant but varies in proportion to the amplitude of the movement. Therefore, both the negative d.c. voltage which gives average gap distance, and the a.c. component which gives dynamic motion are available.
90
DYNAMICS OF MACHINERY
This dynamic measurement provides not only the amplitude of the peak-topeak vibration, but also the frequency and waveform of the motion. This information is of utmost significance in both monitoring and machinery malfunction diagnostics. The eddy current measurement is not disturbed by non-conductive material in the gap between the probe and its observed surface, so that oil, steam and gases do not adversely affect the measurement. The main disadvantage of proximity probes is the sensitivity to shaft mechanical and electrical runout (glitch). Mechanical runout is shaft eccentricity and depends on the manufacturing tolerance. Electrical runout is a false indication of relative displacement due to shaft anomalies (magnetization or internal stresses) and is indistinguishable from actual displacement. A single transducer mounted radially at one bearing provides the vibration signal in only one plane. In order to obtain the shaft precession orbit, it is necessary to mount two probes at 90 0 to each other (Fig. 10.20), at the same radius.
In machine monitoring applications, an eddy current transducer is used as a phase reference probe observing an once-per-turn discontinuity (notch, hole, pin, keyway), to provide phase angle orientation information. This key-phase probe provides both speed and phase references, and a timer pulse for use with peak-topeak eccentricity measurements. An oscilloscope is used with X vs Y (orbit) display and Z-axis intensity. The instrument must display voltage waveforms from vibration transducers in time base and orbit format. The key-phase signal connected to the Z-axis intensity input
91
can be used to trigger the oscilloscope and to provide a reference point from which to make phase angle measurements. The horizontal probe is connected to the positive polarity input of the horizontal amplifier of an oscilloscope (Fig. 10.20). The vertical probe output is connected to the vertical amplifier jack. The key-phase probe output signal is connected to the trigger jack of the oscilloscope for synchronizing the scope, but more importantly, it is connected to the Z-axis input of the Cathode Ray Tube (CRT). This connection must be ac coupled to the Z-input, and the normal ground shorting bar must be removed. If the scope is ac coupled at this point, there is no problem. If it is dc coupled, a capacitor is needed in series with the signal. Shielded coaxial cables should be used, and only one earth ground is to be used for all equipment. The key phase signal is superimposed on the time base and orbit traces producing a bright/blank (or blank/bright) key phase mark. A notch-type keyphasor will produce a voltage pulse which goes negative and then positive. As the notch enters the probe face area (increase in gap) a boost in negative voltage (with negative slope) is produced. When the notch trailing wall passes the probe face (decrease in gap) a less negative burst (positive slope) of voltage occurs. As the gap changes through the notch start and notch end, the time base waveforms and the orbit trace are interrupted with a blank (break) mark, followed by a bright mark [10.18]. If the shaft rotates clockwise, then the blank/bright sequence on the orbit should also be clockwise if the shaft precession is forward. A bright/blank sequence would indicate backward precession. The usual old convention was to view, from outboard, the driving end facing the driven machine. Eddy current proximity sets can be used to monitor rotor-to-stator differential expansions and rotor positions relative to the thrust bearing.
92
DYNAMICS OF MACHINERY
operate, and it has relatively low electrical output impedance making it fairly insensitive to noise induction. A vibration pickup consists of a seismic mass supported by two membranes, so that part of the mass lies within the air gap of a magnetic circuit. The velocity pickup is a seismic instrument fastened to a vibrating structure. At frequencies above the resonance of the mass-spring system, the relative motion between the mass and casing sensed by the transducer is essentially the same as the motion of the structure under test. The seismic mass and the pickup casing vibrate 180 0 out of phase. Relative to a fixed (inertial) reference frame, the mass remains nearly stationary (becomes a fixed point) and the casing motion is measured with respect to it. The amplitude of the e.m.f. induced into the measuring coil is proportional to the velocity of the relative motion and hence to the vibration velocity of the structure under test.
The measuring coil, the damping cylinder and the additional damping coil are supported in the air gap. The damping cylinder reduces the influence of the transducers natural frequency on the measuring signal. The additional damping
93
coil can be energized to compensate for possible reductions in damping at high temperatures or to compensate for static sag if the transducer is used in a vertical attitude. A correction coil, wound round the magnetic flux source, i.e. permanent magnet, eliminates the influence of eddy-current damping on the flux. Limit stops are fitted to prevent excessive movement of the seismic mass. The electrodynamic velocity pickup PR 9266 made by Philips is shown in Figure 10.21 where: 1 permanent magnet, 2 correction coil, 3 measuring coil, 4 additional damping coil, 5 damping cylinder, 6 and 7 membranes, 8 casing, 9 output leads, 10 three-core screened cable, 11 and 12 limit stops. The frequency range is 10 to 1000 Hz, for displacement amplitudes up to 1 mm and accelerations up to 10 g. The undamped natural frequency is 12 Hz. The mass without cable is about 0.5 kg. The sensitivity is 30 mVpp mm s at 110 Hz. Another type of velocity transducer consists of an accelerometer with a built-in electronic integrator. This unit is called a "velometer", and is by all accounts superior to the classic seismic velocity probe
Fig. 10.22
In spite of these advantages, the velocity pickup has many disadvantages that make it nearly obsolete for new installations, although there are many thousands of them still in use today. It is relatively heavy and complex and thus expensive, and it has poor frequency response, extending from about 10 Hz to 1000 Hz. The spring and the magnet make up a low-frequency resonant system with a natural frequency of about several Hz (Fig. 10.22). This resonance needs to be highly damped to avoid a large peak in the response at this frequency. The problem is that the damping in any practical design is temperature sensitive, and this causes the frequency response and phase response to be temperature dependent.
94
DYNAMICS OF MACHINERY
10.4.4 Accelerometers
Due to their advantages light-weight, ruggedness, wide frequency response, good temperature resistance and moderate pricing - piezoelectric accelerometers are the most often used vibration sensing instruments. They are made in several different configurations, but the compression-type, illustrated in Fig. 10.23, serves to describe the principle of operation. This accelerometer is a seismic pickup in which the sensing piezoelectric ceramic discs form the elastic element of the spring-mass system.
The seismic mass is clamped to the base by an axial bolt bearing down on a circular spring. The piezoelectric element is squeezed between the mass and the base. When the accelerometer is subjected to vibrations, the mass will exert a variable force on the piezoelectric discs. The charge developed across the piezoelectric discs is proportional to the applied force, which in turn is proportional to the acceleration of the mass. For frequencies much lower than the resonance frequency of the accelerometer assembly, the acceleration of the seismic mass is equal to the acceleration of the whole pickup. Accelerometers have a very large dynamic range. The smallest acceleration levels they can sense are determined only by the electrical noise of the electronics, and the highest levels are limited only by the destruction of the piezo element itself. Acceleration levels can span an amplitude range of about 10 8 , which is 160 dB. The frequency range of the accelerometer is very wide, extending from very low frequencies in some units to several tens of kilohertz. The high-frequency
95
response is limited by the resonance of the seismic mass coupled to the springiness of the piezo element. This resonance produces a very high peak in the response at the natural frequency of the transducer, and this is usually somewhere near 30 kHz for commonly used accelerometers. A rule of thumb is that an accelerometer is usable up to about 1/3 of its natural frequency. Data above this frequency will be accentuated by the resonant response, but may be used if the effect is taken into consideration. The lower limit is determined by cable and preamplifier. The frequency response curve of an accelerometer is presented in Fig. 10.24.
Most accelerometers used in industry today are of the "ICP" type, meaning they have in internal integrated circuit preamplifier. This preamp is powered by a dc polarization of the signal lead itself, so no extra wiring is needed. The device the accelerometer is connected to needs to have this d.c. power available to this type of transducer. The ICP accelerometer will have a low-frequency roll-off due to the amplifier itself, and this is usually at 1 Hz for most generally available ICP units. There are some that are specially designed to go down to 0.1 Hz if very low frequency data is required. The resonant frequency of an accelerometer is strongly dependent on its mounting. The best type of mounting is always the stud mount - anything else will reduce the effective frequency range of the unit. When mounting an accelerometer, it is important that the vibration path from the source to the accelerometer is as short as possible, especially if rolling element bearing vibration is being measured. If an accelerometer is mounted on a surface that is being strained (bent), the output will be altered. This is known as base strain, and thick accelerometer bases are used to minimize this effect. Sheartype accelerometers are less sensitive because the piezoelectric crystals are mounted to a center post and not to the base.
96
DYNAMICS OF MACHINERY
Machine type Large steam turbine generator sets with fluid film bearings
Transducer type Noncontacting transducer Noncontacting and seismic transducer combination Velocity transducer or accelerometer Noncontacting transducer or axial probe Eddy current/inductive/optical transducer Noncontacting transducer
Measurement locations
Velocity or acceleration Power generation Shaft axial displacement Phase reference and rpm Medium and small industrial steam turbines with fluid film bearings Relative displacement
Shaft
Shaft, at each bearing Each bearing housing and turbine housing Thrust collar
Velocity transducer or accelerometer Noncontacting transducer or axial probe Eddy current/inductive/optical transducer
Axial Z
Mechanical drive
Shaft
Radial
97
As a summary of the characteristics of the different types of transducers, the following comparison is reproduced from [10.8].
Proximity probes
Advantages: a) measures directly the motion of the shaft (the origin of most large machine vibrations), b) measures in terms of displacement (the most meaningful engineering unit for fluid film bearing measurements), c) measurement is noncontact (will not influence the measured vibratory motion because of contact), d) solid-state with no moving parts, e) one sensor simultaneously measures both dynamic motion and (average) position, f) system is modular with the most inexpensive part, the probe, requiring only occasional replacement (because of abuse), g) one extra transducer can be used as a rotor speed sensor and a phase reference, h) excellent frequency response, i) small size, j) well-suited to most machinery environments, k) ease of calibration, l) accurate low frequency amplitude and phase angle information, and m) high level low impedance output. Disadvantages: a) control of observed shaft surfaces desirable to avoid excessive sensitivity to shaft mechanical and electrical runout, b) somewhat sensitive to various shaft materials, c) requires an external power source, and d) sometimes difficult to install.
Velocity pickups
Advantages: a) ease of installation due to external machine mounting, b) strong signal in the mid-frequency range, c) some are suitable for relatively high temperature environments, and d) no external power required. Disadvantages: a) relatively large and heavy, b) manufactured as a unit so that a transducer fault requires replacement of the entire pickup, c) sensitive to input frequency (tendency to emphasize higher frequencies), d) relatively narrow frequency response with amplitude and phase errors at low frequencies, e) has moving parts and is expected to degrade under extended normal use, f) difficult to calibrate, g) measures dynamic motion only (not static position), and h) can respond with excessive cross-axis sensitivity at high amplitude levels.
Accelerometers
Advantages: a) ease of installation due to external machine mounting, b) good frequency response (especially at high frequencies, although this could be a disadvantage by increasing the noise level from various external vibrations), c) small and light weight, d) some are suitable for relatively high temperatures, and e) strong signal in the higher frequency ranges. Disadvantages: a) most sensitive to input frequencies (although this can be an advantage when measuring very high frequencies), b) difficult to locate on the machine case for a meaningful measurement, c) very sensitive to the method of
98
DYNAMICS OF MACHINERY
attachment, d) output requires amplification, e) most sensitive to spurious vibrations (confusing the acquired data and making exact mounting location difficult), f) impedance matching (or charge amplifier) is needed, and g) normally requires some filtering for monitoring applications.
This full-time system, as outlined, provides monitoring of: a) speed, b) two-plane radial vibration of the shaft at each machine journal and each gear shaft, plus case radial vibration on the gear, and c) shaft axial position (dual or single) on all shafts (for protection against excessive thrust deflection). In addition, the following are available from the permanently mounted transducers for periodic monitoring and analysis information: a) shaft orbits, b) gear case orbits, c) phase of the turbine, gear, and compressor shafts, d) axial vibration, and e) eccentricity, or average position, of shafts. With the addition of a roving velocity transducer, case, foundation and piping vibrations can also be periodically monitored.
99
Figure 10.26 shows the layout of a comprehensive monitoring system for a large turbine-driven compressor.
100
DYNAMICS OF MACHINERY
Figures 10.28 indicate the location of seismic pickups used for periodic measurements on a motor driven fan using a) elastic coupling and low shaft (Fig. 10.28, a), and b) belt-driven high shaft (Fig. 10.28, b).
10.4.7 Instrumentation
Transducer signals are processed by a wide variety of electronic instruments. Conditioners, which include filters, analog integrators, and amplifiers, are used to enhance data. Digital recorders, tape recorders and digital computers are used for data recording, especially when transient vibration phenomena are measured. Electronic data collectors are used for storing r.m.s. or peak data that can be transferred to a digital computer; trends can then be established and reports generated. Data processing is carried out using tunable and swept-filter analyzers, tracking filters and FFT spectrum analyzers. FFT analyzers acquire a block of data over a designated frequency range during a period of time, digitize the data, and perform a frequency analysis using the FFT algorithm. They contain buffers capable of storing large quantities of data and can also produce spectrum cascade (waterfall) diagrams, i.e. amplitude versus frequency for various times or speeds. These analyzers can perform integration, r.m.s. band analysis, and compute power spectral density. The magnitudes of r.m.s.-based bands can be displayed in a linear or logarithmic format. Data display instruments include monitors, oscilloscopes, strip chart recorders, analog and digital plotters. Simpler instruments display r.m.s., peak, or average values of measured vibration. Apart from the mentioned transducers, optical (and magnetic) pickups are used in torsional vibration measurements, and general speed and phase measurements on rotating shafts. The optical pickup sends a voltage pulse to the oscilloscope or analyzer when energized by light pulses from a reflective tape (or
101
other marks) bonded on the shaft. The optical system includes a power supply and an amplifier.
Their interpretation makes it possible to determine the precession directivity (forward or backward), the bearing pre-loading (orbit distortion) and the existence of sub-harmonic and supra-harmonic components. Simple orbits are elliptical or Lissajous figures. The classical Lissajous figures may be obtained by compounding two perpendicular harmonic motions with two different frequencies. In Fig. 10.30 it is shown how the orbit is built up from the two components
3 A cos 2 t 450 . 4
x = A cos t ,
y=
102
DYNAMICS OF MACHINERY
The resulting orbit is known as a butterfly or rabbit ears. In the orbit construction, the time steps are labelled 1, 2, 3, etc. Thus the direction of orbital motion can easily be determined.
While Lissajous figures result from harmonic motions with different frequencies, actual rotor steady-state orbits result from two perpendicular components which both are periodic motions (sums of harmonic motions), and usually carry the same set of frequencies. Thus, if x contains 1X and 2X components, then most probably y will have 1X and 2X components, though with different amplitudes and phases.
103
Some machine faults generate periodic vibrations with sub- or superharmonic components. When the rotor vibration has a subharmonic component of order 1/N, the complex vector of the precession radius (for zero phase angles) is of the form
z = R1 ei t + R1 N e i ( N ) t ,
(10.4)
where R1 is the amplitude of the synchronous component (due to unbalance) and R1 N is the amplitude of the subsynchronous component. The plus sign is for forward precession while the minus sign is for backward precession. For N = 2 , and R1 R1 2 = 2 (dominant synchronous component), the orbit
is shown in Fig. 10.31, a for a forward subharmonic component, and in Fig. 10.31, b for a backward subharmonic component. When the two components have different phase angles, the orbits are no more symmetrical.
The corresponding orbits for N = 2 , and R1 R1 2 = 1 2 (dominant subsynchronous component) are shown in Fig. 10.32. Similar conclusions result from the analysis of periodic vibrations with superharmonic components, replacing N by 1/N in equation (10.4). Generally, orbits have external loops when the so-called 2X component is mainly due to shaft misalignment (including gear mesh and belt drives), coupling misalignment (Fig. 10.33) and resulting radial preload. Orbits with internal loops are mainly due to shaft asymmetry (such as with cracked shafts) together with radial preload (from misalignment, gravity or fluid flow). The effect of the radial preload on the shape of steady-state orbits is shown in Fig.11.9 (Chapter 11). With increasing force, the initial elliptical orbit may become banana shaped, then figure eight shaped.
DYNAMICS OF MACHINERY
The direct (unfiltered) orbit and timebase plots of the vibration signals measured at a particular location on a machine are quite complex. The orbit is far from elliptical and the timebase traces are combinations of several harmonic components (fig. 10.33, a). The half spectrum plot (Fig. 11.33, b) is a frequency domain version of the timebase plot for the Y probe. It makes the identification of frequencies and individual component amplitudes easier.
Most frequency analysis instruments display only the positive half of the frequency spectrum, because the spectrum of a real-world signal is symmetric around d.c. Thus, the negative frequency information is redundant. When only one vibration measurement (e.g., the vertical component) is made at a given point, the half spectrum plot is useful, e.g. for tracking changes in the spectral content over a period of time, provided timebase plots are available to check the vibration signal quality. Half spectrum plots reveal new frequency components and changes in the magnitude of previous data at a particular frequency. The phase information is lost. The half spectrum has been intensively used as a machine signature for assessing the machine condition, correlating the frequency and magnitude of the peaks with specific machine faults.
10.5.1.3 Full spectrum plots
In the general spectral analysis, the two-sided spectrum shows both the positive and negative frequency components of a signal. In machine diagnostics, full spectrum plots use data from two orthogonal transducers converted into information about the magnitude, frequency and phase of the directional (forward and backward) response components. Forward components are shown in the positive half of the full spectrum plot, and backward (reverse) components are shown in the negative half. The relative magnitude of the forward and backward components of equal frequency defines the direction of the precession.
105
The full spectrum plot for the example considered in the previous section is shown in Fig. 10.34. Though the right hand side of the full spectrum plot appears to be the same as a half spectrum plot for one probe, this is not true. Only forward components are shown on the positive side of the full spectrum plot.
106
DYNAMICS OF MACHINERY
In the full spectrum plot, a forward circular orbit is represented by a component present only on the right hand side. A backward circular orbit has a component only on the left hand side. When the components on the right and left hand sides are equal or of different amplitudes, the orbit at that frequency is elliptical. The component with the larger amplitude determines the direction of precession along the orbit. When the forward component is larger (Fig. 10.35), the precession is forward. When the backward component is larger (Fig. 10.36), the precession is backward. When the amplitudes are equal (Fig. 10.37), the orbit degenerates into a straight line.
The advantages of using a full spectrum plot are apparent in cases where two different machinery malfunctions produce the same half spectrum (Fig. 10.38, a). The full spectrum for oil whirl and whip is shown in Fig. 10.38, b, and that of a rub is shown in Fig. 10.38, c. In both cases there is an (1/2)X subharmonic component, but for whirl/whip it is forward, while for the rub is backward.
10.5.1.4 Mode shape plots
Mode shape plots (Fig. 10.39) display the precession orbits at selected sections along the rotor, and the dynamic deflected line at a given moment. They provide estimates of the nodal points along the rotor and for the internal clearances between the rotor and stator.
107
Trend plots are used for analyzing changes of observed data as a function of time (Fig. 10.40) determined by modifications in the operating parameters of the machine. They display both global level vibration data and other monitored parameters useful in the machine condition diagnostics.
108
DYNAMICS OF MACHINERY
Bod plots (Fig. 10.41) and polar plots (Fig. 10.42) reveal the critical speeds, the dynamic distorted shape and the mode shapes of the rotor, and the amplification factor at the synchronous frequency of the rotor-bearing system. The polar diagrams of the synchronous component 1X (filtered) are useful in the multiplane field balancing and the detection of cracked shafts.
109
Waterfall diagrams are used for examining the variation of the spectral components (synchronous, sub- and supersynchronous) with speed. This permits the detection of some instabilities (such as oil whirl/whip), of cracked shafts and rubs. Figure 10.43 illustrates a cascade spectrum plot and orbits of a cracked shaft.
Full spectrum cascade plots display the speed evolution of the directional (forward and backward) components of the instantaneous precession radius.
110
DYNAMICS OF MACHINERY
Figure 10.44 shows a normalized full spectrum cascade for a rotor in fluid film bearings. The resonance peaks can be seen on the 1 lines at the critical speed. The backward component is produced by the bearing anisotropy. For a rotor in rolling element bearings this component is missing. At high running speeds, above the onset speed of instability tresh , the unbalance response is dominated by the component with frequency equal to the lateral natural frequency of the rotor system.
Figure 10.45 is the full spectrum cascade plot of a preloaded shaft. It displays the known characteristics, i.e. forward precession, different amplitude forward and backward components, denoting an elliptical orbit, and primary 1X component for the entire machine startup. Other phenomena, such as preload induced rubs, cracked shafts or fluid produced instabilities are conveniently analyzed using such plots (see Chapter 11).
10.5.2.4 Journal center position diagrams
Diagrams of the shaft centerline position (Fig. 10.46) are useful for observing changes in the steady-state position of the rotor in the bearings. They give indications of bearing wear and major changes in the alignment state of the machine, indicating rotor preload due to misalignment or thermal effects. Shaft centerline plots warn about a journal operating near or above the center of the bearing, a usual cause of instability.
111
Graphs of the time variation of the broadband vibration amplitude (Fig. 10.47) permit to avoid the rotor thermal bow at start-up or when driven through the turning gear (due to nonuniform heating), detection of severe vibrations produced by rapid temporary variations of the steam temperature (due to boiler malfunction) or by partial admission, which affects the journal average position in the bearing, hence the stability of the precession motion.
Acceptance region plots are displays where filtered 1X (or 2X) vibration vectors are shown as a trend in polar format (Fig. 10.48). A user-defined normal operating range of the 1X vibration vector is determined within the polar plot to form what is called an acceptance region. Deviation of the 1X vibration vector tip from the acceptance region may be a vital warning of a shaft crack or other rotor disturbances.
112
DYNAMICS OF MACHINERY
Other display formats include a) d.c. gap voltage plots (for proximity probes), b) axial thrust position plots, c) rpm versus time plots and d) multiple orbit plots [10.32, 10.33].
References
10.1. Mitchell, J. S., An Introduction to Machinery Analysis and Monitoring, Penn Well Books, Tulsa, 1993. 10.2. * Vorbeugende Maschineninstandhaltung, Schenck Seminar C 50, Nov 1989, p.70. 10.3. ISO 10816-1, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 1: General guidelines, 1995. 10.4. ISO 10816-6, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 6: Reciprocating machines with power ratings above 100 kW, 1995. 10.5. VDI 2059 - Part 1, Shaft vibrations of turbosets. Principles for measurement and evaluation, Nov 1981. 10.6. ISO 7919-1, Mechanical vibration of non-reciprocating machines Measurement on rotating shafts and evaluation criteria, Part 1: General guidelines, 1996. 10.7. Bently, D. E., Crude vibration amplitude measurements: Peak to peak versus S max , Orbit, vol.15, no.3, Sept 1994, p.3.
113
10.8. * Machinery protection systems for various types of rotating equipment, Part 2, Bently Nevada Corporation, Application Note BNC-015, L0467-00, June 1980. 10.9. Federn, K., Erfahrungswerte, Richtlinien und Gtemastbe fr die Beurteilung von Maschinenschwingungen, Konstruktion, vol.10, no.8, 1958, p.289-298. 10.10. Jackson, Ch., The Practical Vibration Primer, Gulf Publishing Company, Houston, Texas, 1979. 10.11. ISO 13373-1, Condition monitoring and diagnostics of machines Vibration condition monitoring Part 1: General procedures, 2002. 10.12. Khazan, A. D., Transducers and Their Elements: Design and Application, Prentice Hall, 1994. 10.13. Bently, D. E., Proximity measurement for engine system protection and malfunction diagnosis, Bently Nevada Corp. Publication BNC-1, from Diesel and Gas Turbine Progress, March 1972. 10.14. Bently, D. E., Shaft motion and position Keys to planned machine maintenance, Annual Meeting of the Technical Association of the Pulp and Industry, Miami Beach, FL, 14-16 Jan 1974. 10.15. * Machine protection systems, Dymac Measurement and Control, Application Note Dymac MPS-1, Dec 1977. 10.16. Harker, R. G., A new turbine supervisory instrumentation package, Bently Nevada Corp. Publication BNC-3, Aug 1979. 10.17. * Bently Nevada Oscilloscope by Tektronix, Technical/Ordering Information L6026, Jan 1990. 10.18. Jackson, Ch., Balance rotors by orbit analysis, Hydrocarbon Processing, vol.50, no.1, Jan 1971, p.73-79. 10.19. * Machine Monitoring Systems, Equipment for electronic measurement of mechanical quantities, Philips Catalogue 79/80, p.49. 10.20. * Accelerometer calibration for accurate vibration measurements, Brel & Kjaer Application Note No. BR 0173. 10.21. Niemkiewicz, J., Standards for vibrations of machines and measurement procedures, Encyclopedia of Vibration, Braun, S., Ewins, D. and Rao, S.S., eds., Academic Press, London, 2002, p.1224-1238. 10.22. * Data presentation techniques for trend analysis and malfunction diagnosis, Bently Nevada Corporation, Application Note R7/79, July 1979. 10.23. Muszynska, A., Misalignment and shaft crack-related phase relationships for 1X and 2X vibration components of rotor responses, Orbit, vol.10, no.2, Sept.1989, p.4-8.
114
DYNAMICS OF MACHINERY
10.24. Tondl, A. ans Springer, H., Ein Beitrag zur Klassifizierung von Rotorschwingungen und deren Ursachen, Schwingungen in rotierenden Maschinen III, Irretier, H., Nordmann, R., Springer, H., eds., Vieweg, Braunschweig, 1995, p.257-267. 10.25. Laws, B., When you use spectrum, dont use it halfway, Orbit, vol.19, no.2, June 1998, p.23-26. 10.26. Southwick, D., Plus and minus spectrum, Orbit, vol.14, no.2, June 1993, p.16-20. 10.27. Laws, W. C. and Muszynska, A., Periodic and continuous vibration monitoring for preventive/predictive maintenance of rotating machinery, Journal of Engineering for Gas Turbines and Power, vol.109, April 1987, p.159167. 10.28. Gasch, R., Nordmann, R. and Pftzner, H., Rotordynamik, 2nd ed., Springer, Berlin, 2001. 10.29. Southwick, D., Using full spectrum plots, Orbit, vol.14, no.4, Dec 1993, p.19-21 and vol.15, no.2, June 1994, p.11-15. 10.30. VDI 2059 - Part 2, Shaft vibrations of steam turbosets for power stations, March 1983, p.6. 10.31. Bently, D. E. and Muszynska, A., Detection of rotor cracks, Proc. 15th Texas A&M Turbomachinery Symposium, Corpus Cristi, Texas, 10-13 Nov 1986, p.129-139. 10.32. * ADRE 3, Bently Nevada Corporation, Technical/Ordering Information L6024, Jan 1990. 10.33. Eshleman, R. L., Basic Machinery Vibration Analysis, Vibration Institute Press, Clarendon Hills, IL, 1999.
11.
CONDITION MONITORING AND FAULT DIAGNOSTICS
This chapter presents diagnostic techniques for determining machinery performance and predicting mechanical problems. Methods are based on condition monitoring, vibration measurement and analysis.
Three periods can be distinguished: 1) the running-in period, 2) the machine normal operation, and 3) the failure development period [11.2]. The beginning of a machine useful life is usually characterized by a relatively high rate of failure. These wear-in failures are typically due to design
116
DYNAMICS OF MACHINERY
errors, manufacturing defects, assembly mistakes, installation problems and commissioning errors. As the causes of these failures are found and corrected, the frequency of failure decreases. The machine then passes into a relatively long period of operation, during which the frequency of failures occurring is relatively low. This period of a machine life is called the normal wear period and usually makes up most of the life of a machine. There should be a relatively low failure rate during the normal wear period when operating within design specifications. As a machine gradually reaches the end of its designed life, the frequency of failures again increases. These failures are called wear-out failures. This gradually increasing failure rate is primarily due to metal fatigue, wear mechanisms between moving parts, corrosion, and obsolescence. The slope of the wear out part of the bathtub is machine-dependent.
117
Figure 11.2 shows an illustration of the relationship between the machine time in service, the load (or duty) placed on the machine, and the estimated remaining capacity of the machine. Whenever the estimated capacity curve intersects with (or drops below) the load curve, a failure will occur. At these times, repair work must be carried out. If the situation that exists fits within the seven rules outlined above, all related costs (repair work and downtime) will be minimized when using run-to-failure maintenance. 11.2.2.2 Preventive maintenance When specific maintenance tasks are performed at set time intervals (or duty cycles) in order to maintain a significant margin between machine capacity
118
DYNAMICS OF MACHINERY
and actual duty, the type of maintenance is called preventive (or scheduled) maintenance. Preventive maintenance is most effective under the following circumstances: a) data describing the statistical failure rate for the machinery is available, b) the failure distribution is narrow, meaning that the MTBF is accurately predictable, c) maintenance restores close to full integrity of the machine, d) a single, known failure mode dominates, e) there is low cost associated with regular overhaul/replacement of the equipment, f) unexpected interruptions to production are expensive, g) low cost spares are available, and h) costly secondary damage from failure is likely to occur [11.17].
Figure 11.3 shows an illustration of the relationship between the machine time in service, the load (or duty) placed on the machine, and the estimated remaining capacity of the machine when preventive maintenance is being practiced. Maintenance activities are scheduled at regular intervals in order to restore machine capacity before a failure occurs. In this way, there is always a margin between the estimated capacity and the actual load on the machine. If this margin is always present, there should theoretically never be an unexpected failure, which is the ultimate goal of the preventive maintenance. 11.2.2.3 Predictive maintenance Predictive (on-condition) maintenance requires that some means of assessing the actual condition of the machine is used in order to optimally schedule maintenance, in order to achieve maximum production, and still avoid unexpected catastrophic failures. Condition based maintenance should be employed when the following conditions apply: a) the machine is expensive or critical, b) a long lead-time is necessary for replacement parts (no spares are readily available), c) the process is uninterruptible, d) equipment overhaul is expensive and requires highly trained
119
personnel, e) failures may be dangerous, f) secondary damage may be costly, and g) failures are not indicated by degeneration of normal operating response [11.17]. Figure 11.4 shows an illustration of the relationship between the machine time in service, the load (or duty) placed on the machine, and the estimated remaining capacity of the machine when predictive maintenance is being practiced. Note that the margin between duty and capacity is allowed to become quite small, but the two lines never touch. This results in a longer time between maintenance activities than for preventive maintenance. Maintenance tasks are scheduled just before a failure is expected to occur. This requires the existence of a set of accurate measures that can be used to assess the machine integrity.
There are instances where a given machine will require different maintenance strategies during its operational life, e.g. scheduling the maximum time between overhauls during the early stages of the machine life, and increased frequency of monitoring as the age of the machine increases, looking only for unexpected failures.
120
DYNAMICS OF MACHINERY
MTBF, failure modes, g) redundancy, and h) safety (plant personnel, community, environment) [11.17].
121
physical measurements and the effects of the faults, i.e. the symptoms. It obviously requires the knowledge of faults that may occur in the machine, and their criticality, to be able to define the most suitable signature.
3. Detection
Detection involves data gathering, comparison to standards and recommendations (see Chapter 12), comparison to limits set in-plant for specific equipment, and trending over time. The signature characterizing the healthy state is compared to the one extracted from the real measurement. The fault is not defined.
4. Diagnosis
Diagnosis involves recognizing the types of fault developing and determining the gravity of given faults once detected and diagnosed. Sometimes it is referred to as fault isolation. When a signature is related to a specific fault, steps 3 and 4 may be imbedded in one-step detection/diagnosis, which happens often in vibration condition monitoring.
5. Decision
In this step, known as prognosis, the operator has to decide whether to stop the machine for maintenance and repair, or to continue operating. Prognosis involves estimating (forecasting) the expected time to failure, trending the condition of the equipment being monitored, and planning the appropriate maintenance timing. It may include recommendations for altering the operating conditions, altering the monitoring strategy (frequency, type), or redesigning the process or equipment. Sometimes it includes the root-cause failure analysis, and involves some research-type laboratory and/or in situ investigations.
11.4.1 Unbalance
Unbalance (also referred to as imbalance) exists when the center of mass of a rotating component is not coincident with the center of rotation. It is practically impossible to fabricate a component that is perfectly balanced, and even after balancing (see Chapter 13) some residual unbalance exists in rotors,
122
DYNAMICS OF MACHINERY
flywheels, fans, gears, etc. The causes of unbalance include excess mass on one side of a rotor (lost blade, eroded or damaged parts), low tolerances during fabrication (casting, machining, assembly), variation within materials (voids, porosity, inclusions), non-symmetry of design, aerodynamic forces, and temperature changes. Unbalance results in a periodic vibration signal with the same amplitude at each shaft rotation. The characteristic diagnostic symptom is a strong radial vibration at the fundamental frequency, 1X (1 x rotational speed). If the rotor is overhung, there will be also a strong axial vibration at 1X. The half spectrum has the higher peak at 1X (Fig. 11.6, a), the orbit is generally elliptical (Fig. 11.6, b) and the timebase waveforms have one key phase mark per shaft revolution (Fig. 11.6, c).
The high amplitude may be in both horizontal and vertical directions or it may be larger in one than the other. The latter is caused by asymmetrical radial forces such as might be produced by a pressure dam bearing, torque reaction in gears or orthotropic bearings (different horizontal and vertical support stiffnesses). In each case, the higher force tends to suppress the coincident vibration. A special form of unbalance is caused by thermal distortion of the rotor. It is produced by thermally unstable shaft forgings which take a bow when they are heated to an elevated temperature. This bow is a function of temperature and it will
123
not straighten out with time unless the rotor is allowed to cool off. Turbines are more affected than compressors and motors. To measure it, turbine rotor forgings are rotated slowly in an oven while temperature is increased and decreased several times, runout being recorded at several places along the rotor. Maximum allowable runout, at 500 C above operating temperature, is usually 8 m m of bearing span. Another form of unbalance is caused by bowed rotors, especially heavy rotors that have been allowed to sit idle for a long time. Such rotors are difficult to straighten, so they need to be balanced by adding counterweights. To avoid this condition the rotor should be turned when the machine is not in use.
Figure 11.7 shows the full spectrum cascade plot measured during the startup of a machine with an unbalanced rotor. The 1X and -1X frequency components have unequal peaks around 3000 rpm, denoting elliptical precession orbits. The two neighboring peaks denote a split critical excited by unbalance.
124
DYNAMICS OF MACHINERY
expansion preload, offset, or cocked bearing-related preload, and gear mesh forces also belong to this category. Misalignment between coupled machines can be caused by thermal expansion of the casing support structure, by settling or thermal distortion of foundation or baseplate, or by piping forces which deflect the casing and its support. It can be produced by a strong radial component of the fluid flow in fluidhandling machines, especially evident in single volute pumps, or in turbines during partial steam admission on the first stage nozzles. Due to the radial force, the rotor is displaced from the original position and moved to higher eccentricity ranges inside the bearings and seals. It may also become bowed, and rotate in a bow configuration. At these conditions the nonlinear effects of the system become active. Due to nonlinearity, the unbalance forced response of the rotor will contain not only the synchronous component 1X, but also its higher harmonics 2X, 3X, etc. There are two components of a coupling misalignment (Fig. 11.8, a): a) parallel (offset), and b) angular (face). Parallel misalignment occurs when shaft centerlines are parallel but offset from one another in a radial direction (Fig. 11.8, b). Angular misalignment occurs when the shaft centerlines meet at an angle (Fig. 11.8, c). The intersection may be at the driver or driven end, between the coupled units or behind one of the coupled units. A coupling misalignment is shown in Fig. 11.8, d.
The typical effect of misalignment is a vibration with a predominant 2X frequency component, as shown in Fig. 11.9. Shaft alignment techniques are presented in Annex 11.1.
125
Strong 1X vibrations with harmonics (usually up to the third, but sometimes up to the sixth) in the frequency spectrum are the usual diagnostic signatures. The harmonics allow misalignment to be distinguished from unbalance. High horizontal relative to radial vibration amplitude ratios (greater than 3:1) may also indicate misalignment.
The rotor precession orbits show some distortion from the effect of a misalignment load, as shown in Fig. 11.9, c. As the preload increases, the orbit will progressively shift from an ellipse to a banana and finally, in extreme cases, possibly to a figure eight (see Fig. 10.33). In severe cases of misalignment it is not uncommon to have the low or unloaded bearing become unstable due to the journal orbit location in the upper half of the fluid film bearing (Fig. 11.10). The analysis of the shaft centerline position can be used to diagnose excessive preloads. A combination of shaft position and orbit representation gives a clear indication of the position of the shaft in each bearing. In Fig. 11.10, preloads on the shaft are forcing the shaft down in one bearing and up in the other. Note the elliptical and banana nature of the orbits as a result of the preload and the bearing constraint. The key phase dots on the orbits indicate that, although the bearings are
126
DYNAMICS OF MACHINERY
preloaded in opposite directions, the ends of the shaft vibrate in phase with each other.
Misalignment is the deviation of relative shaft position from a collinear axis of rotation, measured at the points of power transmission when equipment is running at normal operating conditions. For instance, if the points of power transmission are 0.5 m apart and the maximum shaft centerline to projected shaft centerline offset is 0.5 mm , the deviation is one mm per m of power transmission distance. At 3000 rpm this deviation is acceptable. At 20,000 rpm the alignment deviation is unacceptable.
127
A general guideline for alignment tolerances is shown in Fig. 11.11. Acceptable amounts of misalignment must be tailored to suit each individual drive train application. Gear-type couplings and universal joint drives must have small amounts of misalignment for proper lubrication to occur. Staying within the acceptable misalignment band usually fills this requirement. Diaphragm couplings, on the other hand, should be aligned within the excellent range.
In a spectrum presentation (Fig. 11.12, a), instability shows up as a spectral component between approximately 40% and 60% running speed. In its early stages the subharmonic component generally fluctuates irregularly in amplitude. As instability progresses, the subharmonic spectral component will increase in amplitude and will remain at the higher amplitude for longer periods of time until it finally dominates the spectrum. When the latter occurs, the amplitude fluctuations generally stop and components at twice and higher multiples of the subharmonic may appear in the spectrum.
128
DYNAMICS OF MACHINERY
In a timebase presentation (Fig. 11.12, b), instability causes the running speed pattern to snake and bounce. The irregular bounce will likewise occur in an orbital presentation (Fig. 11.12, c) but the key indicator in the final stages of instability is two timing marks around the circumference of the orbit, indicating that the shaft rotates twice in the time necessary to complete the orbit. Instability is caused by a variety of factors. Oil whirl, where the shaft rides a pressure wave in the oil film circulating at approximately one-half shaft speed is perhaps the most common example. As a slight variation, the presence of oil whirl close to a critical speed may cause the combined effect to latch in at the critical. Other sources of instability can be hysteretic or frictional in nature, but all those mentioned share a common cause, a force component perpendicular to the rotor normal stabilizing force which, when resolved, produces a tangential whirling force in the same direction as the shaft rotation. Corrective measures for instability can range from minor changes in bearing design such as decreased clearances, decreased bearing area in the lower half to increase load or the addition of stiffness by substituting a pressure dam or lobed bearing design for a plain cylindrical bearing. Next, one generally goes to a much stiffer bearing design such as a tilting pad type with changes to the rotor itself, such as decreasing the bearing span and/or adding diameter resorted to as a final measure should all else fail. In general, operating changes such as varying oil temperature are not successful in eliminating instability although deliberate misalignment has been used in the past to temporarily stabilize a bearing and permit continued operation. A common type of malfunction in machines with fluid film bearings is whirl and whip. Whirl and whip are rotor instabilities (self-excited vibrations) generated by bearing, seal or main flow fluid dynamic forces. The malfunction is characterized by rotor forward subsynchronous precession, often at destructive levels, especially for whip conditions. Analysis of spectrum cascade plots and shaft orbital motion can be used to diagnose whirl and whip. Figure 11.13 shows a half spectrum cascade plot with orbits of an unbalanced rotor supported in oil bearings. At a threshold of stability, below the first resonance speed, the rotor undergoes whirl, as exhibited by the vibration component with frequency just below X (orbit a). When the machine speed increases and passes through the first resonance (orbit b), the amplitude of 1X vibration, caused by unbalance increases, resulting in higher oil bearing stiffness. It causes the suppression of the whirl. The bearing stiffness increases significantly for higher journal eccentricities. This characteristic is widely used for correction of oil whirl and whip malfunctions. A friendly radial preload (provided for instance by a misalignment), holding the journal in an eccentric position corrects effectively the oil whirl and whip in machines supported by hydrodynamic cylindrical bearings.
129
For higher running speeds, above the first resonance, when the shaft is lightly loaded and 1X amplitude is reduced, the whirl appears again (Fig. 11.13). It continues with the frequency just below X and then asymptotically approaches to the rotor first natural frequency (orbit c). Oil whirl is replaced by oil whip. The latter is much more violent and dangerous for the machine integrity because the shaft vibrates at its resonant conditions. Therefore, relatively large cyclical rotor bending stresses can be incurred, introducing a significant risk of high cycle fatigue failure if the steady state tensile stresses in the rotor are high enough.
Vibration data taken during machine startup are conveniently displayed as full spectrum cascade plots. Figure 11.14 shows such a plot of a machine with a threshold of stability of approximately 2300 rpm. For speeds less than approximately 4500 rpm, the vibration is composed almost entirely of high amplitude forward vibration components. The absence of backward components
130
DYNAMICS OF MACHINERY
indicates that the shape of the precession orbits should be circular and the precession is forward. The instability vibration frequency is approximately 0.45X for the whirl instability, and it begins to diverge when the system starts the transition into whip instability as the rotor speed approaches 5000 rpm. For machine speeds above 4500 rpm, small backward vibration components exist and the orbit for the whip instability is slightly elliptical (confirmed by measurements).
Full cascade plots should be used together with orbit/timebase plots [11.23]. The latter are not shown here for conciseness.
131
cycle ( 360 0 ) of its precession motion, and b) a partial rub, when the contact occurs occasionally during a fraction of the period of precession.
In the case of full annular rub, occurring mainly in seals, high friction forces cause the change of the precession direction from forward to continuous backward whirl (known as dry whirl). The waterfall plot of vertical vibrations of a rotor rubbing inside the seal (Fig. 11.15) shows that in the lower speed range, the rotor bounces inside the seal, producing multiple higher harmonics of 1X, while at higher speed a full annular rub occurs. In the case of short-lasting rotor/stator contact, the system becomes piecewise continuous with variable stiffness. The rub may be caused by a seal or other non-rotating part acting as a bearing during part of the shaft revolution. The periodic contact with an obstacle (Fig. 11.16, a), creating the effect of a third bearing, produces a periodic variation of the rotor stiffness which determines the self-excitation of the synchronous response and increases the average spring constant to a higher value (Fig. 11.16, b). This tends to raise the rotor critical speed (Fig. 11.16, c).
132
DYNAMICS OF MACHINERY
Systems with periodic stiffness variations exhibit parametric vibrations described by Mathieu-type equations of motion. The solutions correspond to submultiples of the running speed frequency. If the rotor resonance is less than 1/4, 1/3, 1/2 etc. of operating speed, has some unbalance and is lightly damped, the resonance of the rotor system will be increased by the rub to exactly coincide with the nearest higher fraction of running speed. The rotor will lock on this exact submultiple.
c
Fig. 11.16 (from [11.24])
The total shaft motion orbit (Fig. 11.16, d) has two fixed once-per-turn timer marks indicating that the rotor requires two full turns to complete one orbit. The orbit of the filtered 1X and (1/2)X components show inverse precession directions due to the shaft kicking back as it rubs. Generally, multiples of (1/2)X are also produced by the nonlinearity of the normal/tight rub [11.25]. The partial rotor/stator rub, or rub in oversized or poorly lubricated bearings, causes steady subharmonic vibrations of the frequency equal exactly to half of the rotational speed. The range of the possible subharmonic vibrations varies, however, with the rotational speed. When the rotor operational speed is higher that 3 times its first natural frequency, the resulting steady subharmonic vibrations can have the range 1/3 (light rub) or 1/2 (heavy rub). This condition can be generalized to any value of the rotational speed. If it exceeds the value i times the rotor first natural frequency, then the rotor response will consist of the synchronous component 1X and one subsynchronous component with the lowest frequency equal to (1/i) , or (1/(i-1)) , or (1/3) , or (1/2) with increasing rub force [11.26].
133
b
Fig. 11.17 (from [11.26])
134
DYNAMICS OF MACHINERY
The half spectrum cascade plot in Fig. 11.17, a shows the subharmonic vibrations in the case of light rub. An increase of the rotational speed causes the change of the subharmonic order from higher to lower range. Figure 11.17, b presents the case of a higher rub force. The subharmonic vibrations of the order 1/2 are steadily maintained while increasing the rub force and rotational speed. The samples of the rotor precession orbits were taken at rotational speeds = 227, 404, 595, and 790 rad/s. The steady rotor response consists of two main harmonics: a synchronous component 1X due to unbalance, and a subsynchronous component (1/2)X, (1/3)X, (1/4)X or (1/5)X, only one at a time. Minor higher harmonics are present in the frequency spectrum. The synchronous orbit is always reduced to a straight line inclined to the left, i.e., the vertical and horizontal subsynchronous components are 180 0 out of phase [11.26]. At certain rotational speeds, the thermal effect of rubbing causes an ever changing thermal bow-related unbalance of the shaft.
An interesting feature of the full annular synchronous rub is illustrated by Fig. 11.18. When the rotor speed is increased, the unbalance creates enough force to cause the rotor to contact the stator before the resonance peak. The rotor gets stuck on top of its critical. The phase lag is usually about 80 to 100 degrees. As speed increases, the rub increases, so the system dynamic stiffness increases. This raises the critical speed, thus establishing this lockup situation. If the rotational speed is well above the critical, an impact on the shaft can remove the continuous rub condition, and the machine operates at much higher efficiency.
135
The unbalance force carried by the rotor may occasionally exceed the gravity force and/or other lateral forces applied to the rotor and pedestal. This
136
DYNAMICS OF MACHINERY
causes a periodic lifting of the pedestal, resulting in system stiffness softening, its cyclic variability and impacting. As a result, the rotor may exhibit changes in the synchronous response, and an appearance of fractional subsynchronous vibrations ((1/2)X, (1/3)X,) in some speed ranges. Most common is the occurrence of the (1/2)X vibration component, often measured on rotating equipment. 11.4.5.2 Excessive rotor/bearing clearance Specific dynamic phenomena are caused by increased looseness in bearings (often referred to as dead band), usually due to poor lubrication. Excessive clearances between journals and plain bearing bushes, as well as between rolling element bearings and housing, produce periodic variations of the stiffness of rotor/bearing system (Fig. 11.20), thus providing conditions for parametric unbalance-related excitation which can lead to rotor instability.
These phenomena are similar to those occurring during the rotor-to-stator rubbing, namely variable stiffness, impacting and friction. The similarity is, however, of the mirror image type. The rubbing system is described as normaltight, while the system with increased clearances is described as normal-loose [11.25]. As shown in section 11.4.4, a rubbing rotor becomes periodically stiffer, which leads to an increase of the average stiffness. In the rotor/bearing system with excessive clearances, the average stiffness decreases. This tends to lower the rotor critical speed. If the normal rotor resonance is greater than 1/2 the normal operating speed and the system is lightly damped, the resonance of the rotor will be lowered by the effective decreased stiffness to coincide with the nearest lower fraction of running speed. The rotor will lock into this exact submultiple [11.25]. The diagnosis of excessive clearance, and distinguishing it from the rubbing, should be based on the rotor centerline position and 1X data, frequency spectrum and orbit analysis. While exhibiting similar spectra, the journal/bearing contact is usually maintained during a longer fraction of the vibration period than the rotor/stator rubbing contact, thus the orbits are substantially different from the rub case. While maintaining the contact, the journal slides on the bearing surface, and a part of the orbit follows the bearing clearance circle. The journal remains close to the bearing surface even when the contact is broken. This is different from rubbing when more impacting and unsteady transient motion occurs.
137
Figure 11.21 shows the half spectrum cascade recorded during start-up of a journal rotating in a bearing with relatively large radial clearance in a brass bushing. Subsynchronous vibrations of (1/2)X and (1/3)X, as well as self-excited vibrations are present. 11.4.5.3 Loose rotating parts Looseness may occur at discs or thrust collars mounted on rotating shafts or at bearings untightened in bearing pedestals. A loose disc will still rotate, but at a different speed than that of the rotating shaft. A loose bearing may start rotating, dragged into rotation by the shaft. Their response is a function of clearances, the friction conditions between the shaft and the loose part, as well as the tangential external force applied to the loose part. Depending on a particular machine, the drag force can drive the loose part at a higher frequency than the rotational frequency (e.g.: a loose turbine disc) or
138
DYNAMICS OF MACHINERY
constant. If it does not differ very much from the rotational speed, , the resulting vibrations exhibit the characteristic of beat (Fig. 11.22). Most often, however, the looseness of a rotating part leads to transient conditions. The loose part related vibrations have most often a subsynchronous frequency tending to the natural frequency of the rotor. These vibrations look somewhat similar to fluid whirl/whip vibrations, and may sometimes be confused with the latter.
slow down the loose part. At steady-state conditions, the friction and fluid drag may balance each other, and the loose part rotational frequency, l , becomes
The time signal from a bearing that is loose on a shaft will also be truncated (clipped). The extent and shape of the truncation depend on the physical characteristics (stiffness, mass and damping) of the transmission path between the rotor and stator. Spectral analysis of a truncated waveform yields a number of discrete sum and difference frequencies.
139
The changes in the synchronous 1X amplitude and phase, measured by proximity probes, can be monitored under normal operating conditions to provide alarming and early warning of a shaft crack. The polar plot with acceptance regions (Fig. 10.48) is an excellent format for documenting these shifts. Deviation of the 1X vibration vector tip from an acceptance region may be a vital warning of a shaft crack. More effective is the shaft crack detection using transient data. Figure 10.40 shows a half spectrum cascade plot that documents 1X, 2X and other vibration components from slow roll to maximum available speed. A very useful diagnostic tool is the full spectrum cascade plot (Fig. 11.23). It shows the resonance produced by a 2X excitation force when the operating speed is near half the first critical speed.
The usual format of a full spectrum cascade plot is shown in Fig. 11.24. It clearly shows a peak in the 2X amplitude at about 1390 rpm. The first bending resonance of the machine is approximately 2700 rpm. Note that there is a peak in the 3X amplitude at about 900 rpm, and a peak in the 4X amplitude at about 700 rpm.
140
DYNAMICS OF MACHINERY
When operating at 1390 rpm, i.e. near half the first critical speed, the journal orbit (Fig. 11.25) has an internal loop which is characteristic for signals containing two vibration components with the same direction of precession [11.29]. A more detailed study (not presented here) implies analysis of orbit/timebase plots of the filtered 1X and 2X components. For the examined case it was found [11.19] that the 1X component is forward and slightly elliptical. The 2X component is forward, more elliptical, and larger than the 1X component.
It is important to correlate the changes in the 1X vector with changes of the system and process parameters as well as to trend the 2X vector, to determine whether vector changes are caused by a rotor crack or other factors such as load, field current, steam conditions, or other operating parameters.
141
142
DYNAMICS OF MACHINERY
For optimum performance, a casing-monitoring system similar to that shown in Fig. 11.27 is recommended. The signal obtained from the casing sensor is divided, within the monitor, into bands related to specific mechanical components. The first band, encompassing the frequencies around the running speed frequency, is measured in terms of velocity and fitted with a low-pass filter to eliminate interference from the impeller-vane passing frequency. If cavitation is likely, a bandpass filter can be used to enclose the vane-passing frequency and one or two of its multiples into a second monitored band. In centrifugal pumps with rolling element bearings, high frequency resonant transducers are used to assess pulses of energy above some adjustable threshold to provide early warning of an impending failure.
In high-head applications, where an appreciable amount of thrust force may be developed if internal clearances are lost, it is advisable to include either an axial-thrust position monitor or a thrust-bearing temperature indicator, plus an alarm to warn of impending problems. 11.5.1.2 Centrifugal compressors Modern centrifugal compressors, operating on hydrodynamic bearings, generally have a relatively large casing-to-rotor weight ratio and a stiff support structure (Fig. 11.28). Most of the low-frequency energy developed by the rotor is dissipated by relative motion between journal and bearing, within the bearing clearance. A noncontact relative-motion displacement monitoring system, like that shown in Fig. 11.29 (installed on a compressor of older design), has the fastest and most easily recognizable response at small changes in mechanical condition.
143
At higher frequencies, vane-passing and above, there may be significant excitation generated by aerodynamic turbulence or an impeller resonance. This excitation is relatively easy to detect as acceleration, but its displacement most likely will be below the minimum detectable amplitude of a typical industrial
144
DYNAMICS OF MACHINERY
displacement-monitoring system. Fortunately, this type of problem occurs so infrequently that the additional instrumentation required for protection is not usually warranted on a permanent basis. Thrust position monitoring capability should be included as a part of any centrifugal compressor monitoring system. A typical position-monitoring system consists of an axial-displacement sensor and an appropriate monitor. Two sensors are recommended in high-head or critical applications. Thrust-temperature monitoring is mandatory on high-differential-head compressors, where the failure of a balance drum seal can overload the bearing to failure. 11.5.1.3 Centrifugal fans Centrifugal fans and blowers used for forced- or induced-draft and primary-air service generally have large diameter rotors operating from 500 to 900 rpm in pillow-block bearings, supported on structural steel or concrete foundations (Fig. 11.30).
145
As a rule, the major problem with fans is unbalance caused by: a) uneven build-up or loss of deposited material, and b) misalignment. Both are characterized by changes in vibration at or near the rotational frequency, which can be monitored effectively with either a shaft-displacement or a casing system. Selection of the monitoring system is dictated by the type of construction. If the bearings are supported on stiff, reinforced-concrete pedestals, most of the dynamic force developed by the rotor will be dissipated as relative motion within the bearing clearance. A shaft-monitoring system is best suited for this construction (Fig. 11.31). If bearings are supported on structural steel, the dynamic force probably is dissipated as structural vibration, and a casing seismic monitoring system, using sensors attached to the bearing housings, gives best results. For optimum results, characteristics not specifically related to mechanical condition should be eliminated by filtering the fan casing vibration signal to a bandpass extending from approximately 50% of the running speed to three or four times it.
146
DYNAMICS OF MACHINERY
A continuous thrust-monitoring system using a single axial-position sensor generally provides sufficient protection.
Usually axial compressors are driven by gas turbines so that the monitoring system must be designed having in view the characteristics of both machines. 11.5.2.2 Steam turbines Steam turbines designed for driving process equipment in petrochemical plants, or boiler-feed pumps vary significantly in dynamic response from turbines used for producing electric power. The former usually operate between 5000 and 12,000 rpm and deliver from 6000 to 30,000 hp, with inlet steam pressures up to about 120 bar. Utility turbines operate much slower, at synchronous and halfsynchronous speed, generally are much larger in size, and may use steam at pressures above 250 bar. A relative-motion shaft-displacement system, like that shown in Fig. 11.33, serves best on large process-drive and boiler-feed pump turbines, units with moderate to high casing-to-rotor weight ratios and relatively stiff support structures. It provides excellent data at the low frequencies near the rotational speed, and where shaft instability might present problems. Further, it is the only way to monitor shaft radial position. On critical high-speed turbines, the monitoring system includes backup accelerometers mounted at each bearing that monitor absolute shaft motion. They are useful to avoid problems caused by inadvertent location of shaft sensors at nodal points, in-phase motion of the bearing housing, forces caused by vibration at high frequencies, or a large amount of opposing runout, when relative-motion shaft-displacement systems may not exhibit abnormal changes in the machine condition. Casing accelerometers have a wide frequency range, allowing them to
147
observe both the low rotational frequencies (as a primary or backup means of monitoring) and the high blade- and flow-related frequencies.
Thermocouples imbedded in thrust pads are recommended for warning of a thrust-bearing overload. Journal-bearing temperature, obtained from sensors imbedded in these bearings, is a valuable indicator of bearing performance.
148
DYNAMICS OF MACHINERY
Two sensors are installed for axial position measurement and thrust monitoring, because some turbine conditions, such as blade fouling, can overload the thrust bearing. Anyhow, steam turbines generally are less susceptible to thrustbearing problems than centrifugal problems, which depend on pressure balancing to maintain thrust load within tolerable levels. In addition to the measurements discussed for process-drive and boilerfeed-pump turbines, the unique nature of large utility turbines dictates some modifications. Figure 11.34 shows that the monitoring system for a thermal power station turbine generally incorporates some means for obtaining absolute shaft motion at each radial bearing (either a shaft-riding seismic sensor or a relativemotion and a casing-absolute-motion sensor, electronically subtracted). Rotor position indication, accomplished with a noncontact displacement sensor located at the thrust bearing, should be provided on all turbines. For large turbines with long bearing spans, it is also necessary to measure rotor eccentricity while the unit is on turning gear, to warn of a thermal rotor bow, which could result in packing rubs. Other capabilities of the monitoring system should include phase reference and speed measurements, made with noncontact sensors, and valve position indication, accomplished with potentiometer or similar device.
149
Casing-expansion sensors are necessary to ensure that the sliding shoes are free and functioning properly, to accommodate the large axial growths of high temperature turbines. The differential rotor and casing expansion must also be monitored to avoid rubbing between wheels and diaphragms. As shown in Fig. 11.34, differential expansion can be measured by using a noncontact axial position sensor attached to the casing, and observing the rotor at the end opposite the thrust bearing.
150
DYNAMICS OF MACHINERY
Examples of turbine-generator monitoring systems for different machine train configurations are given in Figs. 11.35 to 11.37, where the usual conventions for transducer numbering are shown, as well as in Fig. 10.27. 11.5.2.3 Gas turbines Compared to other types of industrial rotating machinery, gas turbines have: a) relatively low casing-to-rotor weight ratios, b) light, flexible casings, and c) flexible support structures. Though speeds vary greatly, these machines generally operate at moderate to high speeds. Aircraft derivatives are not considered herein.
Gas turbine vibration signatures, particularly those from units with two or more independent rotors, contain a large number of spectral components, spanning a wide frequency range. Along with several running frequencies, the signatures also may contain components generated by power takeoffs, load and accessory gearing, turbine- and base-plate-mounted auxiliaries, compressor and turbine blades, as well as numerous harmonics and sum and difference combinations. A casing system using accelerometers (Fig. 11.38) is suitable for gas turbines because of its: a) ability to monitor the mechanical condition of several components simultaneously, b) quick response to a variety of problems, c) ability to withstand high temperatures, and d) ease of installation and replacement. Shaftvibration sensors are not favored because they cannot collect the data that define blade and gear condition, and are ineffective on machines using rolling element
151
bearings. However, they may be required in special applications involving rotor stability. Casing accelerometers should be mounted at each bearing and at midspan. For maximum protection, axial position and journal bearing temperature sensors should be included.
High speed industrial gears have moderate bearing preloads and relatively flexible casings, so that casing vibration monitoring systems are favored. One sensor at the coupling end of the high speed shaft provides adequate protection on small gears. On large gears, two accelerometers are usually attached to the gear
152
DYNAMICS OF MACHINERY
casing on, or adjacent to, the coupling end bearings of both the high- and lowspeed shafts. Generally, filters are used to divide a gear signature into manageable segments. The first segment, containing the rotational frequencies of both shafts, starts at about 50% of the lowest running speed and extends to the fourth or fifth multiple of the high speed shaft. The second segment should include frequencies around 1-2 kHz. The third band should enclose the gear mesh frequency and its sidebands. The fourth band (if provided) will cover the very high frequencies generated by pitting and spalling of gear teeth. The layout of the monitoring system of a motor driven large compressor is shown in Fig. 11.39. Similar information is given in Fig. 10.25 for a turbine driven compressor.
Since the main bearings of a reciprocating compressor are typically fluid film bearings, problems related to bearing wear, crankshaft unbalance, or misalignment can be detected with proximity probes in an X-Y configuration at each main crankshaft bearing. Continuously measuring temperature at the main
153
bearings and crosshead slipper provides an early indication of either overloading, bearing failure, or insufficient lubrication. A typical layout of the transducers used to monitor a horizontal reciprocating compressor is illustrated in Fig. 11.40. Proximity probes mounted at crankshaft are not shown.
Figure 11.41 shows the overall layout and the transducer locations and orientation of a large ammonia and CO2 compressor with a crankshaft with six throws and rotational speed 330 rpm. The velomitors (piezo-velocity sensors) are installed horizontally at each end of the crankcase centerline. Six accelerometers are installed vertically on the transition sections which connect the cylinders to the crosshead slipper guides. They are intended to measure the high frequency signals generated by impacts associated with piston rod looseness and knocking. Rod drop measurements are made on all stages that have rider rings. At each monitored cylinder, a proximity probe is mounted vertically on the crosshead oil wiper stuffing box, where it measures the relative position of piston rod. A thermocouple or RTD mounted near each valve measures the gas temperature. Figure 11.42 shows the transducer locations on a large polyethylene reciprocating compressor. The monitoring system collects and processes the following data: a) valve temperature on 48 valves using resistive temperature devices, b) rider band wear using proximity probes on 6 piston rods, c) crankcase velocity with 4 piezo-velocity sensors per compressor, and d) crosshead acceleration with 6 accelerometers.
154
DYNAMICS OF MACHINERY
Vibration analysis of reciprocating compressors is better carried out in time domain [11.43]. Velocity transducers are installed on the crankcase in the horizontal plane parallel to the pistons to detect changes in running speed vibration and in crankcase deflection. They can be mounted on the cylinder head, or on the crosshead at a 45 0 angle in the plane of the piston motion, to detect both vertical and cylinder vibration called stretch motion. Malfunctions can be detected looking at timebase data in the region of 1X and 2X running speed Proximity probes, which measure both the position and motion of the piston rod, detect rider band wear. They can be used to measure crankshaft displacement relative to the main bearings which, under normal conditions, follows an elliptical orbit. Accelerometers installed on the crosshead or distance piece of each cylinder detect impact events. Accelerometers can also be used to confirm valve problems when they are temporarily installed on the valve covers. Valve leaks can be detected evaluating the timebase waveform relative to a reference that defines top dead center and bottom dead center. The portion of the stroke in which the high frequency vibration (produced by gas escaping through the valve) occurs can isolate the problem to either a suction or discharge valve.
155
Annex 11.1
Shaft alignment
By alignment, machine shafts are positioned to have their axes collinear when the machine is running under normal load. Coupling alignment measurements provide data for calculating the offsets of each shaft centerline relative to the other, across the coupling distance, in the horizontal and vertical directions. The maximum alignment deviation is the largest from driver offset and driven offset values. For the machine operating speed, the diagram from Fig. 11.11 shows if realignment is necessary.
There are basically two cold alignment methods: a) measuring the axial and radial displacement of one machine with respect to the other (Fig. A11.2, a), and b) measuring the radial displacement of both machines (Fig. A11.2, b).
a
Fig. A11.2
156
DYNAMICS OF MACHINERY
This method is subject to several sources of error and is not as accurate as the reverse dial indicator and laser methods, though in some situations it is necessary to use this method [11.45].
Fig. A11.4
The reverse dial indicator method is preferably used with the coupling in place (Fig. A11.4). The indicators are mounted on brackets with extension arms, to read the outside diameter of the opposite coupling hub. They are mounted at topdead-center on the hub and calibrated to read zero. Readings are taken every 90 0 as the shafts are rotated. These readings are reduced and plotted on graph paper. Alignment corrections are measured directly from the graph. This is probably the most accurate of the dial indicator-based methods. The method is used when laser systems are unavailable or unsuitable for the machine.
157
The addition of special alignment bars has improved the reverse dial indicator method, as both indicators can be read with ease as the shafts and coupling rotate through 360 0 .
158
DYNAMICS OF MACHINERY
Hot alignment is achieved by replacing the dial indicators with alignment bars and transducers (Fig. A11.5). The Dodd bar method [11.47] uses proximity probes mounted on bars. The tubular bars are in a triangular arrangement with supporting stiffeners between the tubes. One bar contains mounted probes which are referenced to the other bar for the probe gap. Movement of the machines is measured by the change in probe gap, and the relative alignment of the train is calculated and plotted. The alignment bars, representing the projected center lines of the two machines, are attached to the inboard bearing housing of each unit. The noncontacting transducer probes measure the relative movement between the bars, indicating the thermally induced travel of the two shafts. Indicating blocks are mounted on the bar fastened to the driven machine. Four proximity probes and two probe brackets are mounted on the bar attached to the driver. The probes and indicating blocks are positioned to measure the horizontal and vertical movement at each coupling hub. The proximity probes measure the air gap between the probe and indicating block. A proximitor amplifying unit conditions the electrical energy supplied to the probe and linearizes the return signal. The proximitor output signal is routed to readout meters calibrated to display the different movements in displacement units.
159
3. Laser alignment
Laser alignment units consist of a laser fixed to the shaft on one side of the coupling, behind the hub, and a prism affixed to the shaft on the opposite side of the coupling. In Fig. A11.6, 1 is the laser support, 2 prism, a laser, b lens, c focusing device, d filter, e lens, and f detector, 4 adapter, 5 driver, 6 driven machine. The laser and prism are connected to a dedicated computer. As the shaft is rotated, the computer records the alignment readings at multiple positions, typically every 90 0 . The process is usually repeated two or three times to ensure accurate results. Given machine dimensions, the computer will calculate the amount of misalignment at the coupling and the corrections necessary at each machine foot to achieve a correct static, or cold alignment.
4. Optical alignment
Optical alignment equipment generally consists of a precision jig transit or sight level accurate to 1 arc-second ( 25.4 m over 5.2 meters), a portable instrument stand, measurement scales, and tooling for mounting the scales on machines. This method is very accurate and especially useful on long machine trains. It directly shows the alignment of each rotor in the machine train, and the catenary shape of the entire shaft system. This is done by placing scales directly on the shaft and obtaining readings from the scales. Optical alignment is used on machines that have rigid couplings or are not easily measured using the previously presented methods. Examples are large turbine-generator trains and hydroturbines.
References
11.1. * Notes on the use of vibration measurements for machinery condition monitoring, Brel & Kjaer Application Note, 14-227. 11.2. Sidahmed, M., Diagnostics and condition monitoring, basic concepts, Encyclopedia of Vibration, Braun, S., Ewins, D., and Rao, S.S., eds., Academic Press, London, 2002, p.376-380. 11.3. Collacott, R. A., Vibration Monitoring and Diagnostic, Wiley, New York, 1979. 11.4. Mitchell, J. S., Machinery Analysis and Monitoring, Penn Well Books, Tulsa, 1981. 11.5. Reeves, Ch., Vibration Monitoring Handbook, Coxmoor Publ. Comp., May 1998.
160
DYNAMICS OF MACHINERY
11.6. Beebe, R. S., Predictive Maintenance of Pumps Using Condition Monitoring, Elsevier, Oxford, 2004.
11.7. Bloch, H. P., Practical Machinery Management for Process Plants, vol.1: Improving Machinery Reliability, 3rd ed., Gulf Professional Publ., Oxford, 1998. 11.8. Eisenmann, R. C. Sr., and Eisenmann, R. C. Jr., Machinery Malfunction Diagnosis and Correction, Hewllet Packard Professional Books, Prentice Hall, Upper Saddle River, NJ, 1997. 11.9. Bloch, H. P. and Gleitner, F. K., Practical Machinery Management for Process Plants, vol.4: Major Process Equipment Maintenance and Repair, 2nd ed., Gulf Publishing Comp., Houston, 1997. 11.10. Gleitner, F. K. and Bloch, H. P., Practical Machinery Management for Process Plants, vol.5: Maximizing Machinery Uptime, Gulf Professional Publ., Oxford, 2006. 11.11. Forsthoffer, W. E., Forsthoffers Rotating Equipment Handbooks, vol.5 Reliability Optimization through Component Condition Monitoring and Root Cause Analysis, Elsevier, Amsterdam, 2005. 11.12. Macdonald, D., Practical Machinery Safety, Newness, Oxford, 2004. 11.13. Mobley, R. K., An Introduction to Preventive Maintenance, 2nd ed., Butterworth-Heinemann, Amsterdam, 2002. 11.14. Scheffer, C. and Girdhar, P., Practical Machinery Vibration Analysis and Predictive Maintenance, Newnes, Oxford, 2004. 11.15. Barron, R., Engineering Condition Monitoring: Practice, Methods and Applications, Addison Wesley Longman, London, 1996. 11.16. Adams, M. L., Rotating Machinery Vibration: From Analysis to Troubleshooting, Marcel Dekker, New York, 2001. 11.17. Mechefske, C. K., Machine condition monitoring and fault diagnostics, Chap.25 of C.R.C. Handbook, Taylor and Francis, 2005. 11.18. Mitchell, J. S., Bearing diagnostics: An overview, ASME Winter Ann. Mtg., 10-15 Dec 1978, San Francisco, p.15-24, 1978. 11.19. Southwick, D., Using full spectrum plots, Part 2, Orbit, vol.15, no.2, June 1994, p.11-15. 11.20. Piotrowski, J., Shaft Alignment Handbook, 2nd ed., Marcel Dekker Inc., New York, 1995. 11.21. Muszynska, A., Vibrational diagnostics of rotating machinery malfunctions, Course on Rotor Dynamics and Vibration in Turbomachinery, von Karman Institute for Fluid Dynamics, Belgium, 21-25 Sept 1992.
161
11.22. Laws, W. C. and Muszynska, A., Periodic and continuous vibration monitoring for preventive/predictive maintenance of rotating machinery, Journal of Engineering for Gas Turbines and Power, vol.109, April 1987, p.159-167. 11.23. Southwick, D., Using full spectrum plots, Orbit, vol.14, no.4, Dec 1993, p.19-21. 11.24. * Machinery protection and diagnostics topics, Bently Nevada Application Note 003, Feb 1977. 11.25. Bently, D. E., Forced subrotative speed dynamic action of rotating machinery, ASME Paper 74-Pet-16, Petroleum Mechanical Engineering Conference, Dallas, Texas, Sept 1974. 11.26. Muszynska, A., Partial lateral rotor to stator rubs, I. Mech. E. Conference Publication 1984-10, Proc. Third International Conference on Vibrations in Rotating Machinery, Heslington, England, 11-13 Sept 1984, p.327-335. 11.27. Bently, D. E., Basic rotor-to-stator thermal rubs which exhibit rotative speed (1X) symptom only, Orbit, vol.17, no.3, Sept 1996, p.4-6. 11.28. Gasch, R., Nordmann, R. and Pftzner, H., Rotordynamik, 2nd ed., Springer, Berlin, 2001. 11.29. Muszynska, A., Misalignment and shaft crack-related phase relationships for 1X and 2X vibration components of rotor responses, Orbit, vol.10, no.2, Sept.1989, p.4-8. 11.30. Pfleiderer, C., Petermann, H., Strmungsmaschinen, 6th ed., Springer, 1990. 11.31. Mitchell, J. S., Putting vibration and other operating variables to work in a monitoring system, Power, May 1977, p.87-89. 11.32. * Centrifugal Compressor for Ultra-High Pressures, Druckschrift MA 25.69 en/9.83, Mannesmann Demag, 1983. 11.33. Eck, B., Ventilatoren, Springer, Berlin, 1957. 11.34. Kostyuk, A. G. and Frolov, V. V., Steam and Gas Turbines (in Russian), Energoatomizdat, Moskow, 1985. 11.35. Mitchell, J. S., Monitoring the complex vibration characteristics of bladed machinery, Power, July 1977, p.38-42. 11.36. * Rotating Machinery Information Systems and Services, Bently Nevada, Publ. L1001-00, April 1993. 11.37. Hayashida, B., Advancement of Turbine Supervisory Instrumentation continues to help solve machinery problems, Orbit, vol.13, no.1, Feb 1992, p.6-11.
162
DYNAMICS OF MACHINERY
11.38. Murray, G., Mucci, J., and Brier, S., Analysis of generator rotor unbalance, Orbit, vol.14, no.1, March 1993, p.25-29. 11.39. Swan, P., Torsional vibration problems with asynchronous motor, Orbit, vol.18, no.1, March 1997, p.22-24. 11.40. * Monitoring reciprocating compressors, Orbit, vol.11, no.3, Dec 1990, p.20-23. 11.41. Silcock, D., Reciprocating compressor instrumented for machinery management, Orbit, vol.17, no.2, June 1996, p.10-12. 11.42. Smith, T., Quantum Chemical uses reciprocating compressor monitor to improve reliability, Orbit, vol.17, no.2, June 1996, p.14-16. 11.43. Schultheis, S. M., Vibration analysis of reciprocating compressors, Orbit, vol.17, no.2, June 1996, p.7-9. 10.44. * Vorbeugende Maschineninstandhaltung, Schenck Seminar C 50, Nov 1989, p.70. 11.45. Bognatz, S. R., Alignment of citical and noncritical machines, Orbit, vol.16, no.1, March 1995, p.23-25. 11.46. Campbell, A. J., Static and dynamic alignment of turbomachinery, Orbit, vol.14, no.2, June 1993, p.24-29. 11.47. Dodd, V. R., Shaft alignment monitoring cuts costs, Oil and Gas Journal, Sept 1971.
12.
VIBRATION LIMITS
This chapter presents guidelines and recommendations providing acceptable vibration limits for different types and sizes of machinery.
164
DYNAMICS OF MACHINERY
Vibration severity is defined as the maximum value of the broadband rootmean-square velocity in the specified frequency range (typically from 10 to 1,000 Hz), as evaluated on the structure at prescribed points. The Guideline VDI 2056-1964 [12.4] was the basis for the standards ISO 2372-1974 [12.5] and ISO 2373 [12.6] superseded now by ISO 10816 [12.7][12.12]. ISO standards are used as a basis for the corresponding national standards. The obsolete standards still contain useful information. 3. Rotating machinery having flexible rotors, such as large steam turbine generators, multistage pumps and compressors. The machine may be set into different modes of vibration as it accelerates through one or more critical speeds to reach its service speed. On such a machine, the vibration amplitude measured on a structure member may not be indicative of the vibration of the rotor. For example, a flexible rotor may experience very large amplitude displacements resulting in failure of the machine, even though the vibration amplitude measured on the bearing cap is very low. Therefore, it is essential to measure the vibration on the shaft directly. The Guideline VDI 2059-1981 (first draft in 1972) comprised 5 parts [12.13]-[12.17] devoted to general guidelines and four types of turbines. It was the basis for the standard ISO 7919-1996 which has also 5 parts [12.18]-[12.22]. 4. Rotating machinery having quasi-rigid rotors, such as low-pressure steam turbines, axial-flow compressors, and fans. Such machinery contains a special class of flexible rotor where vibration amplitudes measured on the bearing cap are indicative of the shaft vibration. In addition to the International Standards Organization (ISO), various trade organizations such as American Petroleum Institute (API), American Gear Manufacturers Association (AGMA) and the American National Electrical Manufacturers Association (NEMA) have developed and published vibration standards, which are widely accepted and applied. In most cases, these standards have been developed by consensus of consumers and manufacturers, and their use is considered voluntary.
165
Six curves limit the zones for different mechanical conditions, ranging from very smooth to very rough. These categories are for overall (broad-band) vibration measurements at the machine bearing housing. Above 20 Hz, the boundaries are defined by lines with slopes of ( 1) on the log-log diagram, which plots the peak-to-peak displacement amplitude, in mils, as a function of frequency (1 mil = 25.4 m ). They represent constant peak velocity lines. The line indicating the sensory perception level [12.24] is also included.
Subsequent development of vibration severity charts followed this format. The chart developed by Blake [12.25] is given in Annex A12.1. Horizontal displacements measured on bearings have to be multiplied by a service factor.
166
DYNAMICS OF MACHINERY
Figure 12.2 shows the general severity chart for bearing cap measurement (filtered readings) developed first by H. G. Yates (1949) and reworked in 1964 by IRD Mechanalysis [12.26]. It was used only as a guide in judging vibrations as a warning of impending trouble. The diagram plots the peak-to-peak displacement amplitude versus frequency as constant peak velocity lines, with a step of 2 between severity levels (1 in s = 25.4 mm s) . Measurement of the peak-to-peak vibration level was the current practice in the U.S.A. until 1974.
Meanwhile, the VDI Vibration Group developed the Guideline VDI 2056, first released in 1960, then revised and completed in 1964 [12.27]. The vibration intensity was defined by the root-mean-square of the vibration velocity. The guideline was limited to mechanical vibrations above 5 Hz measurable at the surface, at bearings or at fixing points. An assessment scale was built up, starting from the average limit of human perception, 0.112 mm s , and progressing in a ratio of 1.6 (4dB) for the limits of
167
vibration intensity levels. The reason was that experience has shown that a 1.6 times increase in the velocity is distinctly perceptible or detectable in its effects and of importance for the stressing of the machine. A second improvement was the differentiation of the four quality classes: A good, B allowable, C just tolerable and D not permissible, for six different groups of machines. The four groups of machines for which vibration intensity limits have been suggested are the following: Class I (Group K) - individual parts of engines and machines, integrally connected to the complete machine in its normal operating condition; Class II (Group M) - medium sized machines (typically electrical motors with 15 kW to 75 kW output) without special foundations, rigidly mounted engines or machines (up to 300 kW) on special foundations; Class III (Group G) large prime-movers and other large machines with rotating masses mounted on rigid and heavy foundations which are relatively stiff in the direction of vibration measurements; Class IV (Group T) - large prime-movers and other large machines with rotating masses mounted on foundations which are relatively soft in the direction of vibration measurements (for example, turbogenerator sets and gas turbines with outputs greater than 10 MW). The vibration severity ranges for the four groups of machines are listed in Table 12.1 [12.4]. The operating zones B and C cover double-step severity ranges.
Table 12.1
168
DYNAMICS OF MACHINERY
Increased recognition of VDI 2056 by both manufacturers and users of prime movers and driven machinery led to the standard ISO 2372-1974 [12.5], merely the English version of the VDI 2056 recommendations. The term vibration intensity was replaced by vibration severity, so that the r.m.s. value of vibration velocity (over the frequency range 10 to 1000 Hz) was recognized as the best figure of merit for vibration effects on non-rotating parts of machinery. Table 12.1 can also be found in the above standard. It was maintained in the first release of ISO 10816-1 [12.7] as a short term expedient only, until the relevant parts of the standard became available. Although the absolute values suggested by these criteria are not always relevant, due to the different mobilities of the machine structures at the measurement location, they are useful in that they indicate the significance of various degrees of vibration level increases. For example, a level increase by a factor of 2.5 (8dB) is a significant change as it is the span of one quality class. Likewise, an increase by a factor greater than 10 (20dB) is serious as it can take the classification from good to not permissible.
169
The measurement of vibration is broadband and the frequency band is sufficient to ensure that the particular machine is adequately covered, which depends on the type of machine under consideration. For example, the frequency range necessary to assess the integrity of a machine with rolling element bearings should include frequencies higher than those of machines with fluid film bearings. For long-term operation, the standard shows how to set operational vibration limits under the form of alarms and trips. Alarms provide a warning that a defined value of vibration has been reached or a significant change has occurred, at which remedial action may be necessary. If an alarm situation occurs, operation can continue for a period whilst investigations are carried out to identify the reason for the change in vibration and define any remedial action. Trips specify the magnitude of vibration beyond which further operation of the machine may cause damage. If the trip value is exceeded, immediate action should be taken to reduce the vibration or the machine should be shut down.
Shaft rotational speed, rpm Zone boundary A/B B/C C/D 1500 or 1800 2.8 5.3 8.5 3000 or 3600 3.8 7.5 11.8 Vibration velocity, mm s r.m.s.
The designation of zones is the following: Zone A - new machines that can be operated without restriction. Zone B - acceptable for unrestricted long term operation. Zone C - machines that may be operated for a limited time until a suitable opportunity arises for remedial action to be taken. Zone D - vibrations of sufficient severity to cause damage to the machine. The vibration measurement system should be capable of measuring broadband vibrations in mm s r.m.s. over a frequency range 10-500 Hz. If,
170
DYNAMICS OF MACHINERY
however, the instrumentation is also to be used for diagnostic purposes, or for monitoring during machine run-up or run-down, or overspeed, a wider frequency may be required. This standard includes the vibration criteria shown in Table 12.2. It is based on bearing housing/pedestal vibration velocity amplitude ( mm s r.m.s.) for turbine generator sets exceeding 50 MW, and with nominal speeds of 1500, 1800, 3000 and 3600 rpm. The values apply to in situ measurement under steady-state conditions. It is recommended that the alarm value should not normally exceed 1.25 times the upper limit of zone B. In general, the trip value will be within zone C or D, but it is recommended that the trip value should not exceed 1.25 times the upper limit of zone C.
(mm s, r.m.s.)
2.3 4.5 7.1 3.5 7.1 11.0
Velocity
Significant differences in design, types of bearings, and types of support structures require a division of this standard into two machinery groups, namely: 1) large machines with rated power above 300 kW, or electrical machines with shaft heights over 315 mm; and, 2) medium size machines with a rated power above 30 kW up to and including 300 kW, or electrical machines with shaft heights from 180 mm to 315 mm. The larger machines normally have sliding bearings and the range of operating or nominal speed is relatively broad with ranges from 120 rpm to 15000 rpm. The recommended criteria are shown in Fig. 12.3.
171
Classification of the vibration severity zones for large industrial machines with rated power from 300 kW to 50 MW (group 1) is shown in Table 12.3. The zone descriptions are the same as in ISO 10816-2.
Table 12.4
Flexible
Classification of the vibration severity zones for medium size industrial machines with rated power from 15 kW to 300 kW (group 2) is included in Table 12.4.
Fig. 12.3
Based on ISO 10816-3, specific vibration severity standards have been established for pumps [12.28]. They apply to pumps with multivane impeller (centrifugal, mixed or axial flow above 15 kW) and separate or integrated driver.
172
DYNAMICS OF MACHINERY
Zone boundary A/B B/C C/D Vibration velocity, mm s r.m.s. 4.5 9.3 14.7
This standard encompasses machines which may have gears or rolling element bearings, but does not address the evaluation of the condition of those gears or bearings. The zone descriptors are the same as in ISO 10816-2.
173
The recommended criteria values (in mm s r.m.s.) vs. shaft rotational speed (in rpm) for hydraulic machines with nominal power above 1 MW, and nominal speeds between 120 and 1800 rpm are shown in Fig. 12.4. The zone descriptions are the same as in ISO 10816-2.
174
DYNAMICS OF MACHINERY
Vibration limits set by the Hydraulics Institute for horizontal clear liquid pumps, measured on bearing housing are given in Fig. 12.5. Vibration tolerances set by the Hydraulic Institute Application Standards B-74-1: 1967 [12.26] for centrifugal pumps are shown for comparison in Annex A12.2.
Based on experience with similar machines, the damage that can occur when exceeding the guide values is sustained predominantly by the machinemounted components (e.g., turbochargers, heat exchangers, governors, pumps, filters, etc.), connecting elements of the machine with peripherals (e.g., pipelines), or monitoring instruments (e.g., pressure gauges, thermometers, etc.). For rigidly seated reciprocating piston engines, vibration levels are measured at the top edge of the frame or cylinder cover. This standard generally applies to reciprocating piston
175
machines mounted either rigidly or resiliently with power ratings above 100 kW. The vibration criteria for seven different classes of reciprocating machines are presented in Table 12.6. The class definitions are: 1) balanced opposed type rigidly mounted reciprocating gas compressors; 2) multi-throw type rigidly mounted reciprocating gas compressors; 3) single-throw type rigidly mounted reciprocating gas compressors; 4) no example; 5) and 6) industrial and marine diesel engines (<2000 rpm); and, 6) and 7) industrial and marine diesel engines (>200 kW). The zone descriptions are the same as in ISO 10816-2. The values in Table 12.6 were derived from constant displacement in the range 2 Hz to 10 Hz, constant velocity from 10 Hz to 250 Hz and constant acceleration from 250 Hz to 1000 Hz. Vibration values for reciprocating machines may tend to be more constant over the life of the machine than for rotating machines. Therefore zones A and B are combined in this table. In future, when more experience is accumulated, guide values to differentiate between zones A and B may be provided.
The guideline VDI 2063 [12.2] prescribed as limit values 1 mm peak displacement in the range 2 Hz to 10 Hz, 45 mm/s r.m.s. velocity (or 68 mm/s peak velocity) from 10 Hz to 100 Hz, and 4g peak acceleration from 100 Hz to 300 Hz (Fig. 12.6). The limiting curve is overlaid on the vibration severity grade nomograph given in the standard ISO 10816-6: 1995 [12.3] in Annex A12.5. Guidelines for the evaluation of vibrations of reciprocating internal combustion engine-driven alternating current generating sets are given in the standard ISO 8528-9: 1995.
176
DYNAMICS OF MACHINERY
177
as the basis for the evaluation of a new machine, in the absence of any other established knowledge of the satisfactory running characteristics for a machine of that type. 2) Changes in shaft vibration displacement, even though the limits in 1) are not exceeded, may point to incipient damage or some other irregularity. Consequently, such changes relative to a reference value should not be allowed to exceed certain limits. If this reference value changes by a significant amount, and certainly if it exceeds 25% of the reference level, steps should be taken to ascertain the reasons for the change and, if necessary, appropriate action should be taken. In this context, a decision on what action to take, if any, should be made after consideration of the maximum value of vibration, and whether the machine has stabilized at a new condition. The standard ISO 10817-1 [12.32] describes the sensing device (transducer), signal conditioning, attachment methods, and calibration procedures for instrumentation to measure shaft vibration.
Shaft rotational speed, rpm 1500 1800 3000 3600 Peak-to-peak relative displacement of shaft, m 100 200 320 90 185 290 80 165 260 75 150 240
The vibration levels specified here define four quality zones for both relative and absolute shaft vibration measurement at, or close to, the main loadcarrying bearings, at rated speed and under steady state conditions. Higher levels of vibration can be permitted at other measuring locations and under transient conditions, such as start-up and run-down (including acceleration through critical speed ranges).
178
DYNAMICS OF MACHINERY
The recommended shaft vibration amplitude values for large steam turbine-generator sets, in micrometers peak-to-peak, measured relative to the bearings, are shown in Table 12.7 for relative shaft to bearing vibrations, and in Table 12.8 for absolute vibrations.
Table 12.8
Shaft rotational speed, rpm 1500 1800 3000 3600 Peak-to-peak absolute displacement of shaft, m 120 240 385 110 220 350 100 200 320 90 180 290
In both tables, zone A represents new machines that can be operated without restriction; zone B is acceptable for long-term operation; zone C represents machines that may be operated for a limited time until a suitable opportunity arises for remedial action to be taken; and zone D is identified as a trip level as these values are considered to be of sufficient severity to cause damage.
Figure 12.7 shows for comparison the limit values for shut-down given in VDI 2059-2 [12.14]. They represent maximum values of the radius of shaft precession orbit, hence peak displacements. It can be noticed that the values in Table 12.7 at the zone boundary C/D are almost the same, though they represent peak-to-peak values, suggesting to multiply by 2 values in Tables 12.7 and 12.8.
179
presented in terms of both steady running conditions, and any changes that may occur in these steady values.
This standard applies to coupled industrial machines with fluid film bearings, comprising: turbocompressors, turbines, turbine-generators, and electric drives, all having maximum rated speeds in the range 1000 to 30,000 rpm, and powers between 30 kW and 50 MW. In Fig. 12.8, limits are shown for the evaluation of the peak-to-peak shaft displacement relative to the bearing, d, as a function of the rotational speed, n. The three lines at zone boundaries are defined by the following equations
180 Zone limit A/B Zone limit B/C Zone limit C/D
DYNAMICS OF MACHINERY
d A B (in m) = d B C (in m ) = dC
(in m) =
The numerical values specified in Fig. 12.8 are not intended to serve as the only basis for acceptance specifications. In general, the vibratory condition of these machines is usually assessed by consideration of both the shaft vibration and the associated structural vibration. As a result, this Standard should be used in conjunction with ISO 10816-3 [12.9]. The zone descriptions of Fig. 12.8 are the same as in ISO 7919-2. For comparison, the corresponding chart from the recommendations VDI 2059-3 [12.15] is reproduced in Annex A12.6.
181
182
DYNAMICS OF MACHINERY
The guidelines are given for the application of shaft vibration criteria as measured close to the bearings of coupled hydraulic sets, under normal operating and steady state conditions, and any changes that may occur in these steady values. The numerical values specified in Fig. 12.10 present rotor displacements relative to the bearings vs. shaft rotational speed. It is limited to the range of nominal rotational speeds from 60 to 2000 rpm. The zone descriptions are the same as in ISO 7919-2. For comparison, the corresponding chart from the recommendations VDI 2059-5 [12.17] is reproduced in Annex A12.8.
183
Fig. 12.11 (from [12.29]) Figure 12.11 shows the flow diagram for selection of measurements and evaluation of vibration severity.
184
DYNAMICS OF MACHINERY
Figure 12.12 gives suggestions when to make absolute vibration measurements and when to make shaft relative vibration measurements.
Fig. 12.12 (from [12.29]) In general, machines equipped with rolling element bearings will tend to have high bearing stiffness, a stiffness ratio less than 1, and are better suited to vibration measurements at the pedestal and/or casing. Conversely, machines using fluid film bearings and supported on relatively soft pedestals, will have a much higher stiffness ratio, and are better suited to shaft vibration measurements. Table 12.9 [12.29] shows example dynamic stiffness ratios, and the applicability of the reference standard.
185
Table 12.9
Machine High pressure turbine Low pressure turbine Large generator High pressure centrifugal compressor Large fan Small fan & pump Vertical pump Large steam turbine generator set
ISO 10816 (pedestal) Moderate Moderate Moderate Not Good Good Good Good Moderate
ISO 7019 (shaft) Good Good Good Good Moderate Moderate Not Good Good
186
DYNAMICS OF MACHINERY
Figure 12.13 shows guidelines developed by the high speed gear industry [12.35]. The following peak velocity limit values measured on the bearing caps of gear boxes were recommended by Jackson [12.36]: smooth - 5 mm/s and less, acceptable - 5 to 7 mm/s, marginal - 7 to 10 mm/s, planned shutdown repairs 10 to 15 mm/s, and immediate shutdown - 15 mm/s.
Standard API 610 [12.37] API 611 [12.38] API 612 [12.39] API 616 [12.41] API 617 [12.42] API 672 [12.44] API 613 [12.40]
(including runout)
1.25 12000 nmax
0.1 peak
187
188
DYNAMICS OF MACHINERY
The limit lines in the chart correspond to constant values of the quantity
Z=
[ mm s ]
2 3
(12.4)
where a0 is the amplitude of acceleration. Taking Z s = 10 mm 2 s 3 as a reference value, the dimensionless vibration intensity, measured in vibrar, is given by
189
S = 10 log
Z 2 3 = 22 log x0 f Zs
[vibrar ]
(12.5)
2 3 f and S are given for the three zone boundaries in Values of both x0 Fig. 12.14. For comparison, lines of constant peak velocity are indicated on the chart, as well as the danger limit, according to the 1939 release of DIN 4150 [12.47]. It was considered [12.46] that little risk of damage is probable for values 2 3 f less than 50 mm 2 f 3 , a limit that corresponds to S = 37.37 vibrar . of x0
The same criterion was taken into account in [12.48], where a classification of vibrations according to their effect on buildings is given. The limit values of ranges are presented in Fig. 12.15, plotted in coordinates peak displacement vs. frequency with solid lines.
190
DYNAMICS OF MACHINERY
For comparison, r.m.s. velocities of 2.5, 5 and 10 mm/s, as well as peak velocities of 2.5, 5, 12 and 50 mm/s are also plotted with broken lines. Based on our experience [12.49], the allowable limit of building vibrations is in the range of 30-40 vibrars and corresponds to 5 mm/s r.m.s. velocity at frequencies between 5 and 50 Hz.
Range I II III IV
Velocity, mm/s r.m.s. Below 2.5 2.5 5.0 5.0 10.0 Over 10.0
Effect damages not possible damages very improbable damages not probable damages possible, stress check necessary
The standard ISO 4866 [12.50] facilitates the rough evaluation of stationary floor vibrations by measurement of the peak displacement and frequency. However, vibration limits are expressed in terms of vibration severity (Table 12.11). This is the maximum r.m.s. velocity, measured as the largest orthogonal component of vibration, determined at prescribed measuring points on the structure.
191
The safe limit appears to be somewhere below 10 mm/s r.m.s. velocity. Constant r.m.s. velocity lines of 2.5, 5 and 10 mm/s are plotted in Fig. 12.15 with broken lines.
A simplified vibration severity chart (Fig. 12.16) was published in [12.49], based on our experience that the upper limit for the range in which damages from sustained steady-state vibrations are most probable can be chosen at 7 mm/s r.m.s. velocity. However, as in all vibration standards, this must be taken only as a guide in judging the vibration level and as a warning of impending damage.
192
DYNAMICS OF MACHINERY
193
Annex A12.3 In-service vibration severity criteria for centrifugal compressors as a function of shaft speed Compressed Air and Gas Institute
194
DYNAMICS OF MACHINERY
Annex A12.4
195
196
DYNAMICS OF MACHINERY
Annex A12.6 Shaft displacement severity chart for industrial turbosets VDI 2059 Part 3
197
Annex A12.7 Shaft displacement severity chart for gas turbosets VDI 2059 Part 4
198
DYNAMICS OF MACHINERY
Annex A12.8 Shaft displacement severity chart for hydraulic machine sets VDI 2059 Part 5
199
References
12.1. Maedel, P. H. Jr., Vibration standards and test codes, Shock and Vibration Handbook, 5th ed., Harris C. ed., McGraw-Hill, 2001, p.19.1-19.11. 12.2. VDI 2063, Measurement and evaluation of mechanical vibrations of reciprocating piston engines and piston compressors, Sept 1985. 12.3. ISO 10816-6, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 6: Reciprocating machines with power ratings above 100 kW, 1995. 12.4. VDI 2056, Beurteilungsmastbe fr mechanische Schwingungen von Maschinen, Okt 1964. 12.5. ISO 2372, Mechanical vibration of machines with operating speeds from 10 to 200 rev s - Basis for specifying evaluation standards, Nov 1974. 12.6. ISO 2373, Mechanical vibration of certain rotating electrical machines with shaft heights between 80 and 400 mm Measurement and evaluation of the vibration severity. 12.7. ISO 10816-1, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 1: General guidelines, 1995. 12.8. ISO 10816-2, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 2: Land-based steam turbines and generators in excess of 50 MW with normal operating speeds of 1500 r min , 1800 r min , 3000 r min and 3600 r min , 2001. 12.9. ISO 10816-3, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 3: Industrial machines with nominal power above 15 kW and nominal speeds between 120 r min and 15000 r min when measured in situ, 1998. 12.10. ISO 10816-4, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 4: Gas turbine driven sets excluding aircraft derivatives, 1998. 12.11. ISO 10816-5, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 5: Machine sets in hydraulic power generating and pumping plants, 2000. 12.12. ISO 10816-7, Mechanical vibration Evaluation of machine vibration by measurements on non-rotating parts, Part 7: Rotordynamic pumps for industrial application, 2004. 12.13. VDI 2059 - Part 1, Shaft vibrations of turbosets. Principles for measurement and evaluation, Nov 1981.
200
DYNAMICS OF MACHINERY
12.14. VDI 2059 - Part 2, Shaft vibrations of steam turbosets for power stations, March 1983. 12.15. VDI 2059 - Part 3, Shaft vibrations of industrial turbosets, Nov 1981. 12.16. VDI 2059 - Part 4, Shaft vibrations of gas turbosets, Nov 1981. 12.17. VDI 2059 - Part 5, Shaft vibrations of hydraulic machine sets, Oct 1982. 12.18. ISO 7919-1, Mechanical vibration of non-reciprocating machines Measurement on rotating shafts and evaluation criteria, Part 1: General guidelines, 1996. 12.19. ISO 7919-2, Mechanical vibration of non-reciprocating machines Measurement on rotating shafts and evaluation criteria, Part 2: Land-based steam turbines and generators in excess of 50 MW with normal operating speeds of 1500 r min , 1800 r min , 3000 r min and 3600 r min , 2001. 12.20. ISO 7919-3, Mechanical vibration of non-reciprocating machines Measurement on rotating shafts and evaluation criteria, Part 3: Coupled industrial machines, 1996. 12.21. ISO 7919-4, Mechanical vibration of non-reciprocating machines Measurement on rotating shafts and evaluation criteria, Part 4: Gas turbine sets, 1996. 12.22. ISO 7919-5, Mechanical vibration of non-reciprocating machines Measurement on rotating shafts and evaluation criteria, Part 5: Machine sets in hydraulic power generating and pumping plants, 1997. 12.23. Rathbone, T. C., Vibration tolerance, Power Plant Engineering, vol.43, Nov 1939, p.721-724. 12.24. Reiher, H. and Meister, F. J., Die Empfindlichkeit der Menschen gegen Erschtterungen, Forschung auf dem Gebiete des Ingenieurwesens, vol.2, no.11, 1931, p.381-386. 12.25. Blake, M. P., New vibration standards for maintenance, Hydrocarbon Processing and Petroleum Refinery, vol.43, no.1, Jan 1964, p.111-114. 12.26. * A practical guide to in-plane balancing, IRD Mechanalysis, Technical Paper No. 116, 1981. 12.27. Federn, K., Erfahrungswerte, Richtlinien und Gtemastbe fr die Beurteilung von Maschinenschwingungen, Konstruktion, vol.10, no.8, 1958, p.289-298. 12.28. Beebe, R. S., Predictive Maintenance of Pumps Using Condition Monitoring, Elsevier, Oxford, 2004, p.93. 12.29. Niemkiewicz, J., Standards for vibrations of machines and measurement procedures, Encyclopedia of Vibration, Braun, S., Ewins, D. and Rao, S.S., eds., Academic Press, London, 2002, p.1224-1238.
201
12.30. Mechefske, C. K., Vibration standards and acceptance limits, Course MECH 458, Part 5, Machine condition monitoring and fault diagnostics, Queens University, Kingston, Canada, 2007. 12.31. * General guide lines for vibration on Clark centrifugal compressors, Dresser Industries Inc., Clark Turbo Products Division, N.Y., 1971. 12.32. ISO 10817-1, Rotating shaft vibration measuring systems Part 1: Relative and absolute sensing of radial vibration, 1998. 12.33. ISO 10816-0 (draft by H. Kanki), Guidelines for selecting vibration evaluation methods, including shaft relative, shaft absolute, and pedestal vibration measurements, 2003. 12.34. ISO 8579-2, Acceptance code for gears, Part 2: Determination of mechanical vibration of gear units during acceptance testing, 1993. 12.35. AGMA 426.01, Specification for measurement of lateral vibration on high speed helical & herringbone gear units, The American Gear Manufacturers Association. 12.36. Jackson, Ch., Shop testing Is it worth it?, Orbit, vol.19, no.2, June 1998, p.10.. 12.37. ANSI/API Std 610, Centrifugal Pumps for Petroleum, Petrochemical and Natural Gas Industries, 10th ed., Oct 2004. 12.38. ANSI/API 611, General-Purpose Steam Turbines for Petroleum, Chemical and Gas Industry Services, Jan 1997. 12.39. ANSI/API 612, Petroleum, Petrochemical and Natural Gas Industries Steam turbines Special-Purpose Applications, 6th ed., Nov 2005. 12.40. ANSI/API 613, Special Purpose Gear Units for Refinery Service, Chemical and Gas Industry Services, Jun 1995. 12.41. API Std 616, Gas Turbines for Petroleum, Chemical and Gas Industry Services, Aug 1998. 12.42. API Std 617, Axial and Centrifugal Compressors and ExpanderCompressors for Petroleum, Chemical and Gas Industry Services, 7th ed., July 2002. 12.43. API Std 619, Rotary-Type Positive-Displacement Compressors for Petroleum, Petrochemical and Natural Gas Industries, 4th ed., Dec 2004. 12.44. API Std 672, Packaged, Integrally Geared Centrifugal Air Compressors for Petroleum, Chemical and Gas Industry Services, March 2004. 12.45. API Std 673, Centrifugal Fans for Petroleum, Chemical and Gas Industry Services, Jan 2002.
202
DYNAMICS OF MACHINERY
12.46. Steffens, R. J., Vibrations in buildings, Part I, Building Research Station Digest, no.117, May 1970. 12.47. DIN 4150, Erschtterungen im Bauwesen, Teil 1, 2, 3, May 1986. 12.48. Koch, H. W., Ermittlung der Wirkung von Bauwerksschwingungen, VDIZeitschrift, vol.95, 1953, p.733-737. 12.49. Rade, M., Vibration limits for industrial buildings, The Shock and Vibration Digest, vol.26, no.3, May/June 1994, p.11-14. 12.50. ISO 4866, Mechanical vibration and shock. Vibration of buildings. Guidelines for the measurement of vibrations and evaluation of their effects on buildings, 2000.
13.
BALANCING OF ROTORS
Unbalance is the most common malfunction of rotating machines. It produces specific once per rotation (synchronous) precession of rotors and lateral vibrations of their supporting structures. Corrective balancing aims to reduce machine unbalance-related vibrations. Properly balanced rotors are important for the smooth running and integrity of rotating machines. However during manufacturing irregularities are produced by machining errors, cumulative assembly tolerances, distortions due to heat treatment, flaws or inclusions in castings, and material non-homogeneity. Irregularities occurring during operation include uneven wearing and erosion, unsymmetrical buildup of deposits, missing or loose rotor parts, and load-related and thermal distortion of the rotor. Because of these, the actual axis of rotation does not coincide with one of the principal axes of inertia of the body, and variable disturbing forces are produced which result in vibrations. In order to avoid/remove these vibrations and ensure/establish proper operation, balancing becomes necessary. Balancing is the process of adding (or removing) mass on a rotor in order to shift a central principal inertia axis to coincide with the geometric axis of rotation. In a thin disc, the mass centre is moved towards the centre of rotation. In a longer rotor, two or more planes along the axis of the shaft may be selected to redistribute the rotor mass. An appropriate set of masses is applied (or removed) at proper angular orientation and as close as possible to the proper plane along the rotor. This gives a counter effect and balances the unit by making the free centrifugal forces acting on the rotating body as small as possible. Balancing is done by drilling, welding, sticking with adhesive, milling, grinding or attaching screws. Typical items requiring balancing are electrical armatures, spindles and tool-holders, crankshafts, ventilators, turbo-machinery, pump impellers, drive assembly components, turbo charger rotors and car wheels. The forces generated by an unbalance are proportional to the rotating speed squared. Therefore, the balancing of high-speed equipment is especially important.
204
DYNAMICS OF MACHINERY
13.1.1 Definitions
A mass m rotating at a radius r with an angular speed gives rise to a centrifugal force F = 2 m r . At a given value of the angular speed, the magnitude and direction of the centrifugal force is determined by the product m r . In balancing it is called unbalance [13.1]. The unbalance is defined by the vector quantity
u = mr ,
(13.1)
whose magnitude u = m r is measured in units of g mm (gram millimeter ) or kg m . Unbalance is independent of the rotational speed, hence implies that the radius is constant. In a rotor, the radius is measured with respect to the shaft axis, i.e. the line connecting the centers of the bearing journals. It is the geometric axis of rotation attached to the rotor. In most rotors the condition of unbalance does not change noticeably up to the operating speed. These rotors are referred to as rigid rotors. Flexible rotors operating below one third of the first bending critical speed are considered as rigid. For such rotors the unbalance may be specified as a fixed value, irrespective of the rotational speed. They may be balanced at any desired speed up to the operating speed. Unbalance compensation is carried out by adding (or subtracting) material to the unbalanced rotor. For a disc-shaped rigid rotor, this is done in such a way that the sum of the centrifugal forces produced by the initial unbalance and the correction mass becomes zero. The product of the correction mass with the correction radius must be equal to the initial rotor unbalance. The added mass must be placed in the opposite angular position, while the material removing is done in the same angular position as the initial unbalance. The above considerations apply only to discs mounted perpendicular to the rotation axis. Several correction masses can be used (due to the particular form of the rotor) providing that the vector sum of their individual unbalances counteracts the initial rotor unbalance. As a rotor spins, the unbalanced mass tends to pull the rotor toward the bearings on the side the unbalance is located. The point of maximum orbit radius,
205
where the vibration sensed by a fixed non-contacting probe has maximum amplitude, is called the high side or high spot of motion. The point located at 180 0 phase angle is named the low spot. At very low speeds, the high spot will be in phase with the unbalanced (heavy) mass. The intersection of its radius with the disc rim locates the heavy spot. In a flexible rotor, the phase of the measurable high spot will lag the unknown position of the heavy spot. As speed increases, the high spot of these rotors begins to lag the heavy spot. At the critical speed, the phase lag is 90 0 and well beyond the critical speed, the phase lag increases to 180 0 (Fig. 2.11). The high spot can be marked on a spinning disc by holding and gradually moving a piece of chalk (or marking pen) close enough to cause a mark (streak) on the disc rim. It can be located with a dial indicator and a stroboscope, or with proximity transducers and a strip chart recorder or oscilloscope. On small machines a seismic velocity pickup and a vibration analyzer can also be used. The high spot is the part of the shaft that would first contact a closely fitting labyrinth seal and would leave a mark on a shaft at this location. A reference mark called key phasor is also prescribed on the disc to define the reference with respect to which phase angles are measured. For a rigid rotor having significant length, there are four types of rotor unbalance i.e. static, couple, quasi-static and dynamic [13.2], [13.3].
206
DYNAMICS OF MACHINERY
is referred to as pure dynamic unbalance or swash unbalance. Unlike the static unbalance, couple unbalance becomes apparent only when the part is rotated and can be identified by comparing the magnitude and phase readings at the rotor support bearings. Rotors supported between bearings will typically reveal equal amplitudes of vibration but phase readings will differ by 180 0 . This method of detecting couple unbalance does not apply to overhung rotors. In a rotor with overhung discs at both ends, the distance between bearings is smaller than the distance between the correction planes so that couple unbalance is of primary importance as compared with the static unbalance. Unfortunately, only a few balance problems will be pure static or pure couple. Most balance problems will be a combination of static and couple unbalance, further classified as quasi-static and dynamic unbalance.
c
Fig. 13.1
207
Quasi-static unbalance can often be compensated to a satisfactory level by carrying out a simple single-plane balancing to the end having the higher vibration level, and by making the mass corrections in a nearby reference plane not coinciding with the plane through the rotor center of mass.
208
DYNAMICS OF MACHINERY
Fig. 13.2
A phase reference is marked on the disc. Then the disc is spun preferably at the operating speed and the high side is marked on it. The measured response is expressed by the complex displacement z 0 whose magnitude is equal to the measured amplitude and the phase is equal to the angle between the high side and the reference timing mark. Next, a trial mass mt is placed in any position (at 270 0 in Fig. 13.2) and, with the disc running at the same speed as before, the new high side is marked and the response amplitude is measured. The response is expressed by the complex
209
displacement z1 . It represents the effect of both the original unbalance and the added trial mass. The vector difference z is then the effect of mt alone. If mt is rotated counterclockwise by an angle , its response vector z will rotate by and will be parallel and opposite to z 0 . If it is increased in the ratio z 0 z to equal the original unbalance, it will balance the disc. Hence the correction mass (when mt is removed) is mc = mt z0 z . This procedure is a variant of the more general influence coefficient method [13.6] presented in the following.
z0 = z0 0 = z0 R + i z0 I = z0ei 0 ,
(13.2)
where z 0 is the displacement amplitude measured with a non-contacting probe and 0 is the phase angle between the high spot and a reference timing mark. This response is produced by a centrifugal force u 2 , where the vector unbalance u = m r is expressed as the product of the rotor mass m and the unbalance eccentricity r , of magnitude r and phase lag angle m . This angle shows the radial plane where the unbalance u is located with respect to the timing reference mark on the shaft, as measured in a direction opposite to rotation. The rotor synchronous displacement response may be expressed in terms of a complex influence coefficient a multiplied by the rotor unbalance u
z0 = a u .
(13.3)
It is assumed that the influence coefficient a is a function of rotor speed only. This implies that, if a small correction mass is placed on the rotor, the influence coefficient at a particular speed will not change (which is not true in nonlinear systems).
210
DYNAMICS OF MACHINERY
A trial mass mt is placed on the rotor at a given radius rt and at a (sometimes unknown) phase angle t relative to the reference timing mark. The trial complex unbalance is
ut = mt rt e
i t
(13.4)
The rotor response, measured at the same speed as the initial response, is
z1 = a ( u + ut ) ,
(13.5)
i 1
or
z1 = z11 = z1R + i z1I = z1e
(13.6)
z = z1 z0 = z ei .
The initial unbalance vector is
(13.7)
u=
z0 z 0 z = ut = 0 ei ( 0 ) ut . z a z
(13.8)
If the trial mass is removed, it is desired to find the right location where to place a compensating mass mc in order to balance the rotor. The placement of the new mass creates a z vector which is equal to the original z 0 vector, but acts in the opposite direction. This is achieved if the balance correction u c is equal and opposite to the rotor initial unbalance vector u
uc = u = u e i =
where
z0 i ( 0 ) i t e ut e = uc ei c , z
(13.9)
uc =
and
z0 z ut = 0 mt rt = mc rt , z z
(13.10)
c t = = 0 1800 .
The compensation mass
(13.11)
z0 m, (13.12) z t may be placed at the same radius rt as the trial mass mt , but should be shifted by mc =
an angle = 0 1800 in the positive direction (if t is not known).
211
If the trial mass is left in the rotor (when it is welded onto it), the compensating mass produces a trim balance correction. The trim balance correction vector is given by z (13.13) utrim = ( u + ut ) = 1 . a If the rotor response z1 is much larger than the original response amplitude, not only should the trial mass be removed, but also the trial run to get z1 should be repeated with either a reduced trial mass or a change in its angular position.
Example 13.1
A rotor of 450 kg has a steady-state response amplitude of 75 m at a phase angle of 270 0 . A trial mass of 5 g is placed at a radius of 225 mm at a position of 30 0 from the timing mark against rotor rotation. The steady-state response with the trial mass on the rotor has an amplitude of 50 m at 170 0 . Determine the correction balance [13.7].
Solution
The initial response is (Fig. 13.3)
z0 = 752700 = 0 i 75 .
The amplitude of the trial unbalance is
ut = mt rt = 5 225 = 1125 g mm .
a=
z
ut
212
u=
DYNAMICS OF MACHINERY
Fig. 13.3
z0
mt =
213
Example 13.2
A rotor has a steady-state response amplitude of 3.4 mm s at a phase angle of 116 0 . A trial mass of 2 g is fixed to the rotor at the same angular position as the reference mark. With the trial mass, the vibration velocity level is 1.8 mm s at a phase angle of 42 0 . Determine the position and the magnitude of the compensating mass necessary to balance the rotor [13.8].
Fig. 13.4
Solution
The initial response is (Fig. 13.4)
DYNAMICS OF MACHINERY
Example 13.3
To balance the rotor statically, a machine was run up to its operating speed and a vibration velocity level of 15 mm s was measured at a phase angle of
55 0 , after which the machine was stopped. A trial mass of 5 g was fixed to the rotor at the same angular position as the reference mark. Then the machine was run up to its operating speed again. The new vibration velocity level was 18 mm s at a
phase angle of 170 0 . Determine the position and the magnitude of the compensating mass necessary to balance the rotor [13.8].
Solution
The initial response is (Fig. 13.5)
mc =
z0 15 mt = 5 = 2.69 g 27.87 z
215
Fig. 13.5
(or at 00 , 900 , 180 0 ) on the rotor, at the same radius. Geometrically, Sieberts construction (Fig. 13.6) can be used to evaluate the amount and angular location of the required balance correction. A circle with the center in O is drawn having the radius proportional to the amplitude of response to the original unbalance. The relative angular positions A, B, C of the trial mass mt are marked on the original circle. Using the trial mass positions as the centers, circles are drawn (at the same scale) having the radius equal to the corresponding
216
DYNAMICS OF MACHINERY
trial run amplitude AT, BT, CT. The three trial run circles intersect at point T. The line OT is drawn. The correction mass is
mc = mt OT OA
(13.14)
and its location is determined by the angle between the vector OT and the vector opposite to OA, i.e. OC in this case. It must be placed to have the same phase lag with respect to the first trial mass position.
Example 13.4
Sieberts construction from Fig. 13.6 is obtained using the following measurement data. Original response amplitude OA = OB = OC = 20 m .
AT = 30 m - test run with trial mass at 0 0 . BT = 15 m - test run with trial mass
at 90 0 . CT = 43 m - test run with trial mass at 1800 .
Fig. 13.6
and must be placed at 22 .30 clockwise with respect to the second trial mass position.
217
(13.15)
(i , j = 1,2 )
influence coefficients, a trial unbalance is placed in each plane and the new resulting amplitudes of motion are measured. Next, a trial mass producing an unbalance u t1 is added to the first correction plane and the resultant readings z11 and z21 at both bearings are recorded. They can be expressed using the influence coefficients as
( ) z 21 = a 21 ( u 1 + u t1 ) + a 22 u 2 .
(13.16)
Subtracting equations (13.15) from (13.16), the influence coefficients a11 and a21 may be determined as
a11 = z11 z10 , u t1 a 21 = z 21 z 20 . u t1
(13.17)
Finally, the trial mass is removed from the first correction plane and a trial mass is added to the second correction plane. The resultant readings at both bearings are again recorded. If the first trial mass is assumed to be left in place, and a second trial mass producing an unbalance u t 2 is added to the second correction plane, the resultant readings z12 and z 22 are recorded at both bearings. They can be expressed using the influence coefficients as
218
DYNAMICS OF MACHINERY
( ) ( ) z22 = a 21 ( u 1 + u t1 ) + a 22 (u 2 + u t 2 ) .
z 22 z 21 . u t2
(13.18)
Subtracting equations (13.16) from (13.18), the influence coefficients a12 and a22 may be determined as
a12 = z12 z11 , ut2 a 22 =
(13.19)
If the first trial mass is removed, then z11 and z21 should be replaced by the original vibration readings z10 and z 20 . To balance the rotor, correction masses should be placed in planes 1 and 2 to generate vibrations equal in magnitude but opposite in direction to z10 and z 20 . The balance corrections uc1 and uc 2 are given by
1 u10 + u t1 uc1 a11 a12 z12 = u +u = . t2 a21 a22 z 22 u c 2 20
(13.20)
Solving for the final balance correction gives u c1 = a12 z 22 a22 z 12 a11 a 22 a21 a12 , u c2 = a 21 z 12 a11 z 22 a11 a 22 a21 a12 .
(13.21) The values of uc1 and uc 2 represent the trim balance corrections required if both trial masses are left in place. If the computations are carried out using the original vibration readings z10 and z 20 , then the computed balances will correspond to the total original balance correction required in the rotor
(13.22)
Example 13.5
For a machine with a rigid rotor supported in two bearings, vibration velocity levels and phase angles have been measured as shown in Table 13.1. The
219
trial mass mt = 2.5 g was mounted on the rotor in turn, in the bearing plane 1 and bearing plane 2, at the same radius and angular position. Calculate the correction masses and their positions [13.8].
Table 13.1
Trial mass None 2.5 g in Plane 1 2.5 g in Plane 2 Solution 7.2 mm/s 4.9 mm/s 4.0 mm/s
Plane 2
z10 z11 z12
z 20 z 21 z 22
114 0
79 0
220
Fig. 13.7
DYNAMICS OF MACHINERY
z11 z10 = 1.8224 + i 10.5823 = 10.738180.20 , z12 z10 = 4.5787 + i 10.0325 = 11.027965.50 , z 21 z20 = 3.0462 + i 10.0642 = 10.515173.10 , z22 z 20 = 1.4227 + i 1.0075 = 1.7433144.7 0 .
The correction unbalances are
Example 13.6
The initial vibration readings on a rigid-body rotor are
221
The first trial mass is removed, and a trial balance mass of u t 2 = 12 g is placed at a relative phase angle of 180 0 at plane 2. The resulting vibration readings at planes 1 and 2 are
The vibration readings at planes 1 and 2, due to the placement at plane 1 of the trial mass (at a unit radius) of unbalance u t1 = 0 i 10 , are
z11 = 3.2134 + i 4.9482 , z 21 = 3.0111 i 3.3442 .
The vibration readings at planes 1 and 2, due to the placement at plane 2 of the trial mass (at a unit radius) of unbalance u t 2 = 12 + i 0 , are
z12 = 5.0159 + i 3.6443 , z 22 = 9.8910 + i 3.2138 .
a11 = 0.2714 i 0.7118 = 0.7618290.90 , a21 = 0.0495 + i 0.2831 = 0.287480.10 , a12 = 0.0926 + i 0.3349 = 0.3474105.50 , a22 = 0.3374 i 0.5053 = 0.6076303.7 0 .
The correction unbalances (13.22) are
222
DYNAMICS OF MACHINERY
Example 13.7
The run-out readings taken on a rotor during coast down at a very low speed are 0.5 m2720 and 0.4 m1230 at probe 1 and 2, respectively. They are not produced by the unbalance, therefore should be subtracted in the balancing calculation. The initial vibration readings at design speed are
z10 = 1.81480 0.52720 = 2.1205136.7 0 = 1.5440 + i 1.4535 , z20 = 3.61150 0.41230 = 3.20441140 = 1.3033 + i 2.9273 .
The vibration readings at planes 1 and 2 due to the placement of the trial mass (at a unit radius) of unbalance u t1 = 2.4500 + i 4.2435 at plane 1 are
z11 = 1.11780 0.52720 = 1.2396154.30 = 1.1167 + i 0.5382 , z21 = 2.0980 0.41230 = 1.646292.10 = 0.0606 + i 1.6451 .
The vibration readings at planes 1 and 2 due to the placement of the trial mass (at a unit radius) of unbalance u t 2 = 3.7536 i 3.1497 at plane 2 are
z12 = 2.1 980 0.52720 = 2.5978 96.90 = 0.3098 + i 2.5792 , z22 = 3.71020 0.41230 = 3.3296 99.50 = 0.5513 + i 3.2837 .
The influence coefficients (13.17) and (13.19) are
223
a11 = 0.2054 + i 0.0179 = 0.20611750 , a21 = 0.3534 i 0.0888 = 0.3644 165.90 , a12 = 0.3406 i 0.0141 = 0.3409 177.60 , a22 = 0.1643 + i 0.0429 = 0.1698165.30 .
The correction unbalances (13.22) are
u trim2 = u c 2 u t 2 = 3.54116.30 .
The trim balance mass in plane 2 is 3.54 g and should be placed at
116.30 counterclockwise.
u1 = u s ud , u2 = u s + ud .
Solving for u s and u d we obtain
us = u1 + u 2 u u , ud = 2 1 . 2 2
(13.23)
(13.24)
The u s components at the two planes are acting in the same radial direction and generate a centrifugal force at the center of mass. They are equivalent to a 2 u s unbalance applied at the rotor center of mass. The u d components are acting 180 0 out of phase to each other and create a couple which is a free vector.
224
DYNAMICS OF MACHINERY
The static and couple corrections can be carried out simultaneously, but they are independent of each other. The corresponding displacements, z1 and z 2 , can be resolved into the static (in-phase) component and couple (out-of-phase) components
zs = z1 + z2 , 2 zc = z2 z1 2
(13.25)
where z s defines the cylindrical mode and z c define the conical mode.
Example 13.8
The original readings on a two-disc rotor are 8 m1300 and 6 m300 at probe 1 and 2, respectively (Fig. 13.8). Determine the response to the original static and couple unbalance [13.2]. Solution The vibration readings in the two planes are
z1 = 5.1423 + i 6.1284 , z 2 = 5.1622 + i 3.0000 .
Fig. 13.8
225
The product e is the tangential velocity of the centre of mass. In geometrically similar rotors, with the same peripheral velocity, the same stresses in the rotor and the same specific bearing loads are produced if the value e per is kept constant (assuming rigid bearings). The balance quality grades G are based on this relationship. The G number is the product of the specific unbalance and the angular velocity of the rotor at maximum operating speed. It is constant for rotors of the same type G = e per = constant . (13.28) Balance quality grades are separated by a factor of 2.5. However, G numbers of intermediate value may be used to satisfy special requirements. For example, a standard pump impeller has a recommended balance quality grade of
226
DYNAMICS OF MACHINERY
6.3. Special conditions may require a better balance quality of G 4.0 to meet the installation requirement in an area with low structure-borne noise limits [13.13]. The balance quality grades are designated according to the upper limit of the product e given in millimeters per second (for measured in radians per second). Plotted against the maximum operating speed, n, the upper limits of e per are shown in Fig. 13.9.
Example 13.9
How large is the permissible specific residual unbalance e per in a rotor of the balance quality grade G 6.3 for a service speed n = 3000 rpm ? Determine the permissible residual unbalances in each of the correction planes if the rotor is symmetrical and has 40 kg ? Solution For n = 3000 rpm on the horizontal axis in Fig. 13.9, moving vertically to the line G 6.3 , then horizontally to the left to the e per axis we obtain e per 20 m (or 20 g mm kg ). This value can also be calculated. If G 6.3 means that the permissible tangential velocity of the center of mass is 6.3 mm s , then
e per =
v per
3000
30
6.3
For a rotor of mass m = 40 kg , the total permissible residual unbalance is u per = e per m = 20 40 = 800 g mm , hence 400 g mm in each correction plane.
227
228
229
The result can be used even for flexible rotors. Using a finite element model of the rotor, the first modes of bending vibration are calculated, usually all modes below the machine trip speed and the mode just above the trip speed. Then, the worst unbalance distribution for each mode is considered, dividing the total unbalance into suitable individual unbalance components located so that to produce a maximum response in the respective mode of vibration. The calculated amplitudes of the unbalance response are then compared to limit values given by guidelines and standards.
230
DYNAMICS OF MACHINERY
Consider a rotor with a single major mass M (Fig. 13.10) and with an arbitrary unbalance (13.29) u k = m k ek . The displacement of the major mass is
w1 = 11 F1 + 1k Fk
(13.30)
where 11 is the deflection at station 1 due to a unit force at 1, 1k is the deflection at station 1 due to a unit force at station k, Fk is the sum of external and inertia forces acting at station k and F1 = M 2 w1 . For synchronous motion, neglecting external damping and other external forces at the major mass station
Fk = mk (wk + ek ) 2 = 2 (uk + mk wk ) .
The deflection at the major mass is
(13.31)
(13.32)
It is assumed that the balance correction mass mk is small in comparison to the major mass M.
Solving for the deflection at the major mass station for a series of unbalances u k w1 = 1 2 M 11
1k uk 2
(13.33)
The balancing requirement that ensures zero displacement at the major mass station is (13.34) 1k uk = 0 . The bearing force reactions are given by
Fb1 + Fb 2 = 2 [ M w1 + mk (wk + ek )] .
(13.35)
231
2 (M w1 + mk ek ) = 0 .
(13.36)
uk = 0 .
(13.37)
This is recognized as the first requirement for rigid body balancing. The third balancing requirement is obtained by summing moments about the first bearing
(13.38)
l k uk = 0 .
(13.39)
This is recognized as the second requirement for rigid body balancing. In summary, the requirements for flexible rotor balancing may be stated as two equations of rigid body balance (13.37) and (13.39), plus a flexible rotor balance requirement (13.34). In the N plane method, only one balance correction mass is needed to reduce the amplitude at the major mass station or shaft antinode to zero. The balance correction ub1 placed at the major mass station is
ub1 =
11
1k uk
(13.40)
Although the amplitude at the major mass station has been reduced to zero, the transmitted bearing forces are nonvanishing. In order to eliminate the transmitted bearing forces due to unbalance, as well as reduce the rotor amplitude of motion while passing through the first critical speed, two additional balance planes are required. Let ub 2 and ub3 be two additional balance corrections by masses placed on the rotor. The balance corrections are given by
ub1 + ub 2 + ub3 = u k , l 1ub1 + l 2ub 2 + l 3ub3 = l k u k ,
(13.41)
ub1 +
1 12 ub 2 + 13 ub3 = 1k uk . 11 11 11
232
DYNAMICS OF MACHINERY
233
Modal balancing is based on the assumption that the distributed unbalance in the rotor (Fig. 13.11, e) can be represented as a sum of modal unbalances (Fig. 13.11, f, g, h), which are proportional to the eigenforms, but lying in different axial planes. At the same time, the shaft deflection at a particular speed (Fig. 13.11, i) has a modal expansion by virtue of which it can be resolved into a sum of terms proportional to the eigenforms (Fig. 13.11, j, k, l). Running at a critical speed, the deflected shape almost coincides with the corresponding eigenform, the other terms in the expansion being vanishingly small. Due to the orthogonality of mode shapes, a particular unbalance distribution ui (x ) = ai i ( x ) , proportional to the ith eigenform i (x ) , can only excite the lateral deflection wi (x ) = bi i (x ) , i. e. the deflection which also takes the ith eigenform. Thus unbalance excitation and bending response shapes are the same for each term in the series. This property enables the serial elimination of unbalance terms ui (x ) by compensating balance masses. In practice, this means running the rotor up to very near its first critical speed and adding unbalances (correction masses) u1k (k = 1, 2 ,...) at particular locations x k on the rotor. The rotor deflection is thereby reduced enough to allow running through the first critical speed and almost up to the second. The process is repeated with another set of balancing masses u 2 k (k = 1, 2 ,...) . Note that the individual sets of balancing masses only affect the bending caused by the corresponding eigenform and no other. This leads to a systematic balancing procedure with condition equations for the sets of balance masses [13.18]. From this it is found that balancing out N critical speeds or N eigenforms requires at least N planes and therefore the addition of at least as many balancing masses (known as the N plane method). If there are, for example, two critical speeds in the operating range and one just over, then at least 2 + 1 = 3 balance planes are necessary. Let the three correction planes be denoted I, II, III (Fig. 13.11, m). To compensate the unbalance at the first critical speed, a single mass producing u12 should be added at the antinode of the first eigenmode (Fig. 13.11, n). To compensate the unbalance at the second critical speed, two masses producing the unbalances u21 and u23 should be added in antiphase near the antinodes of the second eigenmode (Fig. 13.11, o). To balance at the third critical speed, three masses producing u31 , u32 , u33 are located at the antinodes of the third eigenmode (Fig. 13.11, p). This way the rotor is not balanced as a rigid rotor at low speeds.
234
DYNAMICS OF MACHINERY
The end result of such balancing is subject to small errors, as higher order unbalance terms, or eigenforms remain uncompensated. The accuracy can be much improved if the rotor is balanced as a rigid rotor at low speed. This means that two more balance planes are required (using the N + 2 plane method), as shown in Fig. 13.11, q, where the five planes are indicated as I, II, III, IV, V. The individual sets of unbalance masses are then statically in balance, i.e. the sum and the static moment of the unbalances are zero. The locations of balance masses are shown in Figs. 13.11, r, s, t. Their magnitudes and angular positions are measured in the neighbourhood of the critical speeds nk1 , nk 2 , nk 3 . One set of masses affects only the deflection of one particular eigenform. Each set is both statically and dynamically compensated. The method includes balancing the rotor at low speeds as a rigid rotor. The sets of masses no longer affect the balance of the rigid rotor.
235
References
13.1. ISO 1925, Mechanical vibration, Balancing Vocabulary, 2001. 13.2. * A practical guide to in-place balancing, IRD Mechanalysis, Technical Paper No.116, 1981. 13.3. Schneider, H., Auswuchttechnik, VDI Taschenbcher T29, Dsseldorf, 1972. 13.4. Kellenberger, W., Elastisches Wuchten, Springer, Berlin, 1987. 13. 5. Thearle, E. L., Dynamic balancing of rotating machinery in the field, Trans. ASME, vol.56, 1934, p.745-753. 13.6. Chen, W. J. and Gunter, E. J., Introduction to Dynamics of Rotor-Bearing Systems, Trafford Publ., Victoria, Canada, 2005. 13.7. Somervaile, I. J., Balancing a rotating disc: simple graphical construction, Engineering, Feb 1954. 13.8. * Static and dynamic balancing, Brel & Kjaer Application Note No. 17227. 13.9. Kellenberger, W., Balancing flexible rotors on two generally flexible bearings, Brown Boveri Review, vol.54, no.9, Sept 1967, p.603-617. 13.10. Dimarogonas, A. D. and Haddad, S., Vibration for Engineers, Prentice Hall, Englewood Cliffs, NJ, 1992. 13.11. Gunter, E. J. and Jackson, Ch., Balancing of rigid and flexible rotors, Handbook of Rotordynamics, Ehrich, F. F. ed., McGraw-Hill, New York, 1992. 13.12. ISO 1940-1, Mechanical vibration - Balance quality requirements for rotors in a constant (rigid) state, Part 1: Specification and verification of balance tolerances, 2003. 13.13. ISO 1940-2, Mechanical vibration - Balance quality requirements of rigid rotors, Part 2: Balance errors, 1997. 13.14. Dimentberg, F. M., Theory of balancing flexible rotors, Russian Engineering Journal, vol.11, 1964. 13.15. Bishop, R. E. D. and Parkinson, A. G., Vibration and balancing of flexible shafts, Applied Mechanics Reviews, vol.21, no.5, May 1968, p.439-451. 13.16. Darlow, M. S., Smalley, A. J. and Parkinson, A. G., Demonstration of a unified approach to the balancing of flexible rotors, ASME Paper 80-GT-87, ASME Gas Turbine Conference, March 1980. 13.17. Bishop, R. E. D. and Gladwell, G. M. L., The vibration and balancing of an unbalanced flexible rotor, J. Mech. Eng. Sci., vol.1, no.1, 1959, p.66-77.
236
DYNAMICS OF MACHINERY
13.18. Kellenberger, W., Should a flexible rotor be balanced in N or (N + 2 ) planes?, ASME Journal of Engineering for Industry, vol.94, 1972, p.548-560. 13.19. Den Hartog, J. P., Mechanical Vibrations, 4th ed., Dover, New York, 1984. 13.20. Gunter, E. J., Barrett, L. E. and Allaire, P. E., Balancing of multimass flexible rotors, Proc. 5th Turbomachinery Symp., A&M University, Cllege Station, Texas, Oct 1976. 13.21. ISO 11342, Mechanical vibration Methods and criteria for the mechanical balancing of flexible rotors, 1998. 13.22. SR ISO 1925: 1995, Echilibrare. Vocabular. 13.23. SR ISO 1940-1: 1994, Vibraii mecanice. Condiii de calitate pentru echilibrarea rotoarelor rigide. Partea 1: Determinarea dezechilibrului rezidual admisibil.
14.
RECIPROCATING MACHINES
There are two groups of vibration phenomena of practical importance in reciprocating machines: a) vibrations transmitted to the foundation by the machine as a whole, and b) torsional vibrations in the crankshaft and in the shafting of the driven machinery. They are produced by the periodic accelerations of the crank mechanism and the periodic variations in cylinder gas pressure. Only the vibratory motions of the machine frame and the developed reaction forces are studied herein. An example is also given of fault diagnosis by vibration measurement for an auxiliary diesel engine. Related topics are the acoustic resonances and pulsation control of the piping system of reciprocating compressors.
238
DYNAMICS OF MACHINERY
Let P be the pressure force on the piston, which is variable with the time (or with the crank angle = t ). The gas pressure pushes the piston downward and presses upward against the cylinder head so that when the entire engine is considered, the resultant in any direction would be zero.
a
Fig. 14.1
However, the force P results in a torque about the crankshaft, called the gas-pressure torque. Indeed, the piston force P (force F1 ) is transmitted to the piston pin C (or through the piston rod to the crosshead). Neglecting friction, the force F1 is held in equilibrium by the forces F 2 = P tan and F3 = P cos . The
forces F1 , F 2 , F3 of Fig. 14.1 are acting on the piston pin (or crosshead). The reaction force to F 2 acts to the right on the guide or frame (Fig. 14.1, b). The force equal and opposite to F3 is a compression in the connecting rod. It is transmitted through the connecting rod to the crank pin as force F 4 . By shifting this force parallel to itself to O we obtain a force F 5 = F 4 and a torque T p which is the driving torque of the gas pressure
239
Tp =
P d = P y tan . cos
F6 = P
(14.1)
The force F5 = P cos is taken up by the main bearings at O and can be resolved into a vertical component F7 = P tan . The four forces transmitted to the stationary parts of the engine are: a) P upward on the cylinder head; b) P tan to the right on the cylinder or the crosshead guide; c) P downward on the main bearings at O; and d) P tan to the left on the main bearings at O (Fig. 14.1, b). The total resultant force on the frame is zero, but there is a resultant torque P y tan acting clockwise. By the law of action and reaction, this torque must be equal and opposite to the driving torque d P cos on the crankshaft (acting anticlockwise in the direction of rotation). Thus the gas pressure in the cylinder do not causes any resultant force on the engine frame, but produces only a torque about the longitudinal axis. and a horizontal component
has a counterclockwise reaction torque on the frame, hence of opposite direction with respect to the gas pressure torque (Fig. 14.2, b).
240
DYNAMICS OF MACHINERY
a
Fig. 14.2
There are also inertia forces of the rotating parts (the crank and the parts revolving with it). These can be reduced to zero by counterbalancing the crankshafts. Figure 14.3 shows the finite element model of a crankshaft in which the counterweights can be seen.
Fig. 14.3
241
Fig. 14.4
s = DC = DA'+ A' O BO CB ,
s = l + r r cos l cos = r ( 1 cos ) + l ( 1 cos ) .
(14.3)
AB = l sin = r sin
sin = r sin , l
cos =
r2 1 sin = 1 2 sin 2 l
2
12
1 r2 sin 2 . (14.4) 2 l2
242
DYNAMICS OF MACHINERY
s = r ( 1 cos ) +
or
r2 sin 2 , 2l r2 4l
s = r ( 1 cos ) +
( 1 cos 2 ) ,
(14.5)
r2 r r s= r + cos t + cos 2 t . 4l 4l
The velocity of the piston is
v=
a=
(14.6)
Note that one term varies with the same frequency as the rotation; this is called the primary term. The term which varies at twice the frequency of rotation is called the secondary term. The importance of the secondary term is established by the crank-connecting rod ratio r l . For a connecting rod of finite length, the motion of the piston is periodic but not harmonic.
(14.7)
where k G is the radius of gyration of the rod about the piston pin.
243
In the dynamically equivalent rod, m1 undergoes translation while m 2 is subjected to translation and rotation.
By proportioning the connecting rod with a slight extension beyond the crank end, it is possible to make h = l . The mass m 2 then coincides with the crank pin and its motion is that of pure rotation. The two concentrated masses are then expressed by the equations
c m1 = m 1 , l
m2 = m
c . l
(14.8)
We will analyze here only the simpler case where the connecting rod is replaced by concentrated masses expressed by equations (14.8). The translating mass m t in C is the sum of the piston mass and the portion of the connecting rod m 1 . The rotating mass m r is composed of the remaining portion m 2 and any unbalanced mass of the crankshaft assigned at this position, both of which will be assumed to be balanced by a counterweight.
244
DYNAMICS OF MACHINERY
Ti = mt a y tan =
. r = mt r 2 cos t + cos2 t (l cos + r cos t ) tan l Replacing
tan r sin t , l
(14.10)
cos 1.0 ,
(14.12)
1 ( 1 + cos 2 t ) , 2
.
1 sin 2 t , 2
(14.13)
equation (14.12), giving the inertia torque about the crankshaft, reduces to
r 1 3r sin t sin 2 t sin 3 t Ti = mt r 2 2 2 2l 2l
(14.14)
The engine-frame torque differs from the reverse of the crankshaft torque by the magnitude of the so-called residual couple. This is an inertia couple due to the connecting rod. It corrects for the error in the angular acceleration of the connecting rod which is introduced when the common assumption is made that the mass of the connecting rod is borne at the piston pin and crankpin in inverse proportion to the distances of these points from the center of gravity of the connecting rod. The residual couple of the connecting rod usually is negligible in in-line engine cylinders but is taken into account in radial engine dynamics. If the pressure variation throughout the cycle of the machine is known, it is possible to evaluate the gas-pressure torque (14.1) as a function of the crank angle . This calculation is based on the pressure-volume diagram for a typical cylinder, obtained from the pressure-volume card obtained experimentally.
245
From the diagram of the cylinder pressure vs. crank angle (Fig. 14.6, a) the resulting gas-pressure torque can be calculated as a function of crank angle, hence time (Fig. 14.6, b).
It is usually convenient to make a harmonic analysis of this function in the following form:
T p = b0 + a1 2 sin + a3 2 sin
+ b1 2 cos
+ a1 sin + b1 cos +
(14.15)
In a two-stroke machine, the cycle is complete in a single revolution, and only integer orders occur. In a four-stoke machine, the cycle requires two revolutions, and, in general, half-integer as well as integer orders will occur. The coefficients of all orders up to j = 18 for a number of representative engine cycles have been tabulated (F. P. Porter 1943). Neglecting gravity, the total torque delivered to the crankshaft is the sum of the inertia torque (14.14) and the gas-pressure torque (14.15). The total torque on the frame is that due to inertia plus the negative of (14.15) due to gas pressure. The total forces on the frame are due only to inertia. In a more detailed analysis [14.4], not limited by the approximation (14.4), the magnitude of the inertia force (14.9) has the form
(14.16)
246
DYNAMICS OF MACHINERY
b1 = 1 ,
b2 = +
and
3 1 b 4 = 3 + 5 + ... , 16 4
(14.17)
r . l
(14.18)
It contains only even higher order components. The inertia torque (14.14) has the form
(14.19)
1 1 a 2 = + 4 + ... , 2 32
(14.20)
It contains integer harmonic components of all orders. The major unbalance occurs at twice rotational speed.
247
first crank ( 1 = 0 ) .
Based on equations (14.9) and (14.14), the inertia unbalance of a counterbalanced multi-cylinder engine consists of a vertical force of magnitude
Fi = mt r 2 r cos ( t + ) + l cos 2 ( t + )
j j j =1 n
(14.21)
sin 3 t + j .(14.22)
Even if these forces and moments add up to zero, it is possible to have a pitching moment about a horizontal axis perpendicular to the center line of the engine. This moment can be found by summing the moments of the Fy forces about the first cylinder. If the distance from the jth cylinder to cylinder 1 is c j , the pitching moment about the first cylinder becomes
248
M z = mt r 2
DYNAMICS OF MACHINERY
c
j =1
cos t + j +
r c j cos 2 t + j , (14.23) l
Table 14.1
where c1 = 0 .
Table 14.1 [14.5] illustrates the forces and couples developed by some multi-cylinder machines for different crank arrangements and numbers of cylinders. It applies to machines having the same bore and strike for each cylinder. For compressors at which the bore and strike of the cylinders are not all the same, Table 14.1 should not be used. The unbalanced forces and couples should be computed for each cylinder and the results superposed. In multi-cylinder engines and compressors, the net forces and torques are modified by the cancellation of harmonics among events in the different cylinders. This cancellation is achieved by the arrangement of cylinders, positioning of cranks
249
of the crankshaft, ignition sequence, etc. A general treatment of engine balancing [14.6] is beyond the aim of this presentation. The main results can be summarized as follows [14.3]: Multi-cylinder in-line four-stroke-cycle engines, firing at equal intervals and with parallel lines of stroke are unbalanced with respect to harmonics of order N 2 , whenever N is an integral multiple of the number of cylinders n. The reciprocating inertia force contains no odd harmonics after the first. Half-integer harmonic orders occur only for the gas-pressure torque. The magnitudes of the unbalanced harmonics are n times the magnitudes of corresponding harmonics for one cylinder. The reciprocating inertia forces act in planes normal to the crankshaft and distributed along the length of the shaft. The unbalanced harmonics are n parallel forces which alternate in phase in the lines of stroke of each cylinder at frequencies i n 2 , where i is an integer. The resultant of these forces is at the centroid of the parallel forces. For a balanced-type crankshaft (rear half is a mirror image of the front half, implying an even number of cylinders), this point is at the middle of the crankshaft. The harmonics of the moment of the resultant force with respect to a reference line are called rocking moments, since the resultant force may produce pitching vibrations of the engine on its mounts (or on the foundation). The total engine-frame torque due to both gas-pressure reactions and dynamic unbalances is obtained by adding the magnitudes of corresponding harmonics of the gas-pressure torque reaction and inertia couple due to inertia torque of reciprocating parts. Their effect is to produce rolling vibrations of the engine frame. A six cylinder four-cycle engine of 0, 120, 240, 240, 120, 0 deg. crank shaft has all forces balanced and all moments balanced. Consequently, a V-12 engine made up of two in-line 6s would also be balanced. An eight-cylinder in-line engine (0, 180, 90, 270, 270, 90, 180, 0 deg.) is completely balanced. The balance of the engine presupposes that the engine parts, including the engine frame, are rigid bodies which do not deform under the influence of a balanced force system. Actually, the engine is composed of elastic bodies and the balanced forces may cause deformations which are vibratory in nature. The effects of elasticity of engine parts is most pronounced in the case of large marine in-line engines. Some of the assumptions upon which a theoretical treatment is based often are not satisfied exactly in the practical cases. A basic assumption is that each cylinder produces an identical pressure-time history. Irregularities arising from variations in ignition, fuel distribution to the cylinders, irregular valve operations, etc., violate this assumption. They usually excite the fundamental harmonic of the gas-pressure cycle, which is quickly noticeable and usually referred to as engine roughness [14.3].
250
DYNAMICS OF MACHINERY
251
truncated waveform yields a number of discrete frequencies, sometimes sum and difference components. A similar problem occurs with looseness of the engine frame bolting to the subbase. Strong second-order (2 X ) harmonics occur when some bolts lose their tightness. Increase of the clearance between the main bearing shell and housing can determine a periodic variation of the crankshaft support stiffness. Stiffness nonlinearities produce Mathieu-type vibrations with subsynchronous spectral components X 2 or X 3 , and their multiples. Usually the X 2 component is accompanied by integer multiples 1X, 3 X 2 , 2X, 5 X 2 of decreasing amplitude (with increasing order). Piston slap is an impact generated by the reversal of the side force acting on the piston, especially near the top dead center. In large diesel auxiliary engines, large impact forces occur due to the relatively large piston mass and the piston to liner clearance, which in some cases overpass combustion generated forces. These impacts produce an increase of all harmonic components in the frequency spectrum, especially low order. One cannot neglect excitations induced by the driven (electrical) machine or other neighboring engines.
Fig. 14.8
a) A pitching moment due to reciprocating inertia forces. It is represented as a couple acting in the vertical plane containing the lines of stroke of the cylinders,
252
DYNAMICS OF MACHINERY
usually referred to as M 1V (external couple, first order, 1X) and M 2 V (external couple, second order, 2X). the first major unbalanced component (10X ) . b) A resultant vertical force, usually small, with the tenth harmonic F10
c) A rolling moment, due to both the gas-pressure torque reaction and the inertia torque of reciprocating parts, having the first significant components: the fifth-order harmonic M 5 R ( 5X ) due to gas pressure and the tenth-order harmonic M 10 R ( 10 X ) due to both gas-pressure reaction and inertia torque. Higher order harmonics are of negligible magnitude. Measurement points were chosen on the mounting feet, on the engine frame at the crankshaft level, at the top edge of the frame and on the cylinder covers, at several locations along the engine (between the coupling flange end and the turbocharger end). Apart from overall level measurements, frequency spectra of the peak-topeak vibration velocity were plotted, in the frequency range 4-100 Hz. A characteristic spectrum, plotted at 750 rpm crankshaft speed and partial load, is shown in Fig. 14.9
253
c
Fig. 14.10 (from [14.7])
254
DYNAMICS OF MACHINERY
Worthy of note are the large amplitude components 2X and 2X. The first is produced by the unbalanced pitching moment M 2 V , the second corresponds to the ignition rate in five cylinders. Diesel engine-electric generator alignment was checked during measurements to minimize the misalignment-induced 2X spectral component. The harmonic 5X can be produced by the rockers or by the unbalanced rolling moment M 5 R . The 7X component corresponds to the impact rate of the camshaft (two cams/cylinder for valves plus one cam/cylinder for injection). The X 2 component, which corresponds to the ignition rate in a cylinder and to the camshaft speed, is associated with irregularities in the injection timing. In order to explain the relatively large amplitude of components 2X and 2X, measurements were performed at various engine speeds, at a point near the cylinder covers, with the engine in idling condition. At 500 rpm (Fig. 14.10, a) the component 2X has the highest level, while 2X is slightly lower. At 600 rpm (Fig. 14.10, b), the component 2X occurs at 25 Hz, with a level three times higher than at 500 rpm. At 680 rpm (Fig. 14.10, c) 2X occurs at 28.3 Hz, having a level 2 times lower than at 600 rpm, while the component 2X increases to 5 mm s . At 750 rpm (Fig. 14.10, d), 2X decreases to 8.5 mm s , while 2X, which now occurs at 25 Hz, increases to 12 mm s , having the highest level. It was concluded that, at 25 Hz, a structural resonance occurs, excited at 600 rpm by the component 2X and at 750 rpm (the rated speed), by the component 2X. Measurements have shown that the relative level of the 2X and 2X components decreases at points located in the middle of the engine, both at the engine mounting feet level and at the top edge of the frame (Fig. 14.11). The relatively low frequency of the recorded vibrations and the amplitude maps plotted along the height and the length of the engine excluded consideration of lateral resonances of the flexible engine frame, in the so-called H-shape and X-shape modes of vibration. It was concluded that the resonance is a rigid body resonance of the whole engine-generator set on the flexible engine room floor structure. The main cause of engine vibration does not lie, in this case, in the engine itself, but in the insufficient stiffness of its supporting structure. Location of the engine with the middle on a local transverse stiffener was not sufficient to prevent the pitching vibrations induced by the unbalanced moment M 2 V . Supplementary stiffeners have been added, to transmit the load from the engine base to adequate bulkhead plating and deck stiffeners, which eliminated the resonance condition.
255
In some engines, loosening of the engine bolting to the subbase yielded higher 1X and 2X components, with a series of higher multiples, which made identification of vibrations stemming from the engine itself more difficult (Fig. 14.12).
Fig. 14.11
Fig. 14.12
The effect of other malfunctions was studied, mainly by simulating various faults. Irregularities in combustion produce large X 2 , 2X and 7X harmonic components (Fig. 14.9). In fact, all measured engines were maladjusted, compression checks showing differences up to 15% between cylinders for the same engine. When the injection was shut off in the cylinder next the measurement point (Fig. 14.13) or in the adjacent cylinder (Fig. 14.14), the X 2 component decreased at a level corresponding to measurements at idleness. A broken valve spring, producing a recognizable clattering noise, gave rise to larger 2X and 5X components in the frequency spectrum measured at a point near the respective cylinder. Excessive valve clearance produced deliberately during measurements had the same effect. Other faults such as worn main bearing halfshells, a damaged roller in the fuel injection pump, misalignment preloads caused by offset camshaft bearing housings, and problems with the nozzle needle of the fuel injection valve gave minor or unnoticeable changes in the frequency spectrum. In some cases the diagnosis was made difficult because the faults caused the engine to run unevenly and spectrum averaging attenuated some changes in the frequency spectra.
256
DYNAMICS OF MACHINERY
Fig. 14.13
Fig. 14.14
257
correct modeling is a condition for the accurate analytical prediction of the mechanical natural frequencies, associated mode shapes, and cyclic stresses in the compressor cylinder-manifold system, in order to avoid coincidences between the mechanical and acoustical resonant frequencies.
Figure 14.16 shows various components of an ideally simple two cylinder system. Besides the cylinders, there are the suction and the discharge bottle which often have internals in the form of choke tubes and baffles (Fig. 14.17).
258
DYNAMICS OF MACHINERY
In addition, the discharge bottle is usually restrained by wedge clamps of finite stiffness. The bottles are flexibly connected to the cylinders via the nozzles. One end of the nozzle is connected to the cylinder by a bolted flange joint, and the other end of the nozzle penetrates the bottle with a welded joint (branch connection). Additional suction and discharge piping, additional stages, connected units, etc., distort the ideal features of Fig. 14.16 and make the mechanical analysis of this system even more complex.
259
crank end internal cylinder pressure times the area of the crank end head. The internal cylinder pressure force is plotted versus crank angle in Fig. 14.18. Compressor mechanical unbalance forces These forces are caused by accelerations of the piston assembly and crankshaft throws and are transmitted to the compressor frame via the bearing reactions. This type of force is determined from the manufacturers balance data and pressure-volume cards.
The compressor inertial force versus crank angle (time) is nearly sinusoidal (broken line in Fig. 14.18). As a result, depending on the connecting rod length to stoke ratio, the amplitude of the second harmonic is about 20% of the fundamental. There are no significant harmonics above two times crank shaft speed. The gas force plus the inertial force represents the total load (solid line). Cylinder stretch forces Figure 14.19 illustrates, schematically, an additional important excitation force - the cylinder stretch. Under the action of compression loads within the cylinder, each cylinder will tend to stretch and shrink once per revolution of the crankshaft. Thus, there is a strong first order component of cylinder stretch motion, with the various cylinders out-of-phase with each other due to the crank angle phasing. The magnitude of cylinder stretch can often be 0.25 mm in peak-to-peak amplitude and occasionally reaches 0.4 mm. This movement can, in some cases, cause excessive stresses in the attached nozzles and bottles. First order cylinder stretch has sometimes been a significant contributor in breaking nozzles and baffles.
260
DYNAMICS OF MACHINERY
Shaking forces in piping Acoustic shaking forces act in the axial direction of pipe runs causing vibration in the axial direction of that pipe run. But maximum vibration actually occurs in the transverse direction of adjoining piping that runs perpendicular to the piping where the shaking force is acting. Figure 14.20 shows the action of shaking forces on common adjoining piping configurations, including spans with pinned to nearly fixed ends, L-bends and U-bends.
261
The magnitude and shape of the flow pulses through the compressor valves are determined by the physical, geometrical and mechanical characteristics of the compressor (rotational speed, bore, stroke, loading, compression ratio, etc.). The velocity of the piston is approximately sinusoidal in shape, due to the finite ratio of connecting rod length to crank radius. Since the flow is based on the product of the piston velocity and the piston swept area, the shape of the discharge
262
DYNAMICS OF MACHINERY
flow curve at the piston face is of the same shape as the piston velocity curve. Two simplified examples are shown in Figs. 14.36 and 14.37.
Figure 14.36, a shows the discharge valve flow versus time for a singleacting cylinder. During compression, the suction and discharge valves are closed. When the pressure in the cylinder reaches the discharge back pressure, the discharge valve opens, and the flow versus time (or crank angle) wave through the valve has the shape of the corresponding portion of the piston velocity curve (a quarter of a sinusoid for l r = ). As the cylinder reaches the top dead center, the discharge valves close, and the flow returns to zero.
263
A frequency analysis of the flow wave is shown in Fig. 14.36, b. Due to the repetitive action of the compressor cylinder, excitation is generated only at discrete frequencies, which are integer multiples of the running speed. The highest harmonic amplitude occurs at 1 running speed (for a single cylinder end), with the levels decreasing at higher harmonics. For a double acting cylinder ( l r = 5 and no valve loses), the flow versus
time diagram contains two flow slugs slightly different and not 180 0 apart in time (Fig. 14.37, a). The cylinder produces flow excitation at all integer harmonics of running speed as shown in Fig. 14.37, b. Assuming no interaction between the piping (i.e., no reflected acoustic waves), the pressure wave out of the cylinder takes on a shape, as a function of the crank angle, as shown in Fig. 14.38.
Figure 14.39, a shows the p-V diagram for a cylinder which is not affected by pulsations, while Fig. 14.39, b shows a diagram distorted due to pulsation. For this type of p-V card, the discharge pressure is higher than desired, and the suction pressure is lower. The valve opening and closing times are also distorted. The capacity is lower than calculated for the ideal case, resulting in decreased efficiency.
a
Fig. 14.39 (from [14.11])
264
DYNAMICS OF MACHINERY
Actual dynamic pressure data taken from a natural gas compressor is shown in Fig. 14.40 as pressure-volume (p-V diagram) and pressure-time data.
Notice that the ideal p-t wave (Fig. 14.38) and the actual p-t wave (Fig. 14.40) are definitely non-sinusoidal, which results in pressure pulsations at the higher harmonic frequencies, as seen in the overlaid frequency spectrum. Comparing the p-V card with an ideal p-V diagram in Fig. 14.39 one can notice the dynamic character during the discharge (top of the curve). This distortion of the p-V diagram
265
comes from acoustic resonances of the discharge piping as the pressure pulses are reflected back into the cylinder. Strong acoustic responses of the piping can distort the p-V card resulting in compressor overloading. The complex interaction which occurs between the piping and compressor can cause a variety of actual p-V cards among different cylinders and compressors. Considering the phasing of pressure waves resulting from multiple compressor cylinders operating with head-ends and crank-end pockets, the shape and frequency content of pressure-time waves can become very complex. The interaction of the piping with the cylinders further distorts the picture since these pressure pulses can excite acoustic natural frequencies (resonance).
14.3.3.2 Acoustic resonance phenomena
The flow pulses caused by the reciprocating action of the compressor create pressure pulses or waves that move through the piping system. As the disturbance propagates through the medium, portions of the gas are alternately compressed or expanded from the equilibrium state. The wavelength of the pressure wave is
a , f
(14.24)
where a is the acoustic velocity and f = 2 is the frequency. This equation describes the spatial distribution of pressure maxima and minima of the acoustic wave.
Standing waves
In order for acoustic or pulsation waves to reinforce and result in resonance, reflections of acoustic waves are necessary. Full reflections occur at closed or open ends. An acoustic compression pulse is reflected by a closed end as a compression pulse; an open end will reflect it as a rarefaction pulse. Partial reflections occur at pipe section discontinuities. Pulsations can cause pressure forces at a restriction such as reducers, elbows, pipe caps, orifices or partially closed valves. The superposition of an incident wave and a reflected wave, being the sum of two waves travelling in opposite directions, will give rise to a standing wave. Acoustic standing waves are like the natural modes of mechanical vibrating systems. They are defined by a natural frequency and two distinct mode shapes, one for pressure and one for velocity. The acoustical response in the piping is a function of both the mechanical properties of the compressor, the thermo physical properties of the gas, and the acoustical circuit defined by the attached piping. When a particular harmonic of running speed is near or coincident with an acoustical natural frequency, the acoustic
266
DYNAMICS OF MACHINERY
response (dynamic pressure amplitude) is amplified. These resonances can be simple organ pipe type resonances or complex modes involving all of the piping. Figure. 14.41 provides the pressure and velocity mode shapes for the second mode of a closed-closed pipe.
At the pressure nodes, pulsation pressures are minimal. If a pressure transducer were placed at these nodes, no pulsation pressure would be detectable. The velocity mode shape is shown by the bottom trace in Fig. 14.41. The velocity amplitude is a maximum at the pressure nodes (maximum kinetic energy) and zero at the pressure maxima (maximum potential energy), except at the piston surface, which is not a true velocity node due to piston motion. Because of the resonant condition, the gas velocity at the velocity maxima may be greater than the piston velocity. When pipe lengths are equal to multiples of the wavelength, resonance can occur. In addition, pipe resonances can occur when the lengths coincide with one half or one quarter of the wavelength, with the right combination of end conditions (open or closed). Most practical piping is open at least at one end. The compressor discharge line ends in a bottle or a manifold. Likewise, the suction pipe almost always begins with an open end. According to API 618 [14.8], if the diameter reduction is two-to-
267
one, a contraction can be treated as a closed end. If a pipe is connected to another pipe having a diameter that is at least twice as large, it can be considered to be openended. Closed-closed modes are occasionally encountered in pulsation control bottles and acoustic filters.
Half wave resonance
For half wave resonances, both end conditions must be the same, i.e. openopen or closed-closed. The first three pressure mode shapes are shown in Fig. 14.42. Resonances occur at multiples of the half-wave frequency. The acoustic resonance frequency is
f = na , 2L
(14.25)
where L is the effective length of pipe and n = 1, 2, 3,... The length should be corrected for entrance and exit effects [14.11].
For quarter wave resonances, end conditions must be opposite, i.e. one open end and one closed. The first three pressure mode shapes for an open-closed pipe are depicted in Fig. 14.43. The acoustic resonance frequency is
f = na , 4L
(14.26)
where L is the effective length of pipe and n = 1, 3, 5,... (odd integers). The length should be corrected for end effects.
268
DYNAMICS OF MACHINERY
A quarter-wave resonance can cause erroneous dynamic pressure measurements when a pressure transducer is connected to a main line with a short nipple and a valve forming a quarter-wave stub. When its length is tuned to the pulsations in the main line, the needle of the pressure gage will wobble or indicate severe pressure variations which do not actually exist in the main line [14.11].
Acoustic resonances
The existence of quarter and half-wave modes alone do not constitute resonances. Resonance occurs when a compression wave is generated at a frequency equal to an acoustical natural frequency. The build up in amplitude occurs because a reflected wave arrives at the proper time to reinforce the wave at the compressor. The arrival of the reflected wave is dependent upon the path length of the piping elements. Therefore, the standing wave pattern amplitude is reinforced so that the actual maximum pulsating wave amplitude is substantially greater than the induced level. Since the large pressure amplitudes are the element of most concern, the pressure antinodes are the areas of concern. Figure 14.44 illustrates how the response of an acoustic system varies with the excitation frequency. The figure shows a piston operating in a pipe with a closed end opposite the piston. Since this system behaves as a closed-closed pipe, the resonant frequencies are n a 2 L , which for the given dimensions occur at 20, 40, 60, 80, and 100 Hz. The plot in the figure represents the pressure amplitudes at point A, located at the piston, as the piston excitation frequency is varied from 0 to 100 Hz. As the frequency increases, the resonant amplitudes decrease. This is because the lowest modes, starting with the fundamental, have the most energy and, thus, are more dangerous. The responses at the nonresonant frequencies are small but nonzero. At points A and C, the peak pressures at resonance are approximately equal, while at point B, the only resonances observed are at 40 and 80 Hz, which are the even-numbered modes, for which point B is a pressure node.
269
If no damping is present, the pressure fluctuations at the antinodes would, theoretically, be infinite. Actual piping systems have acoustic damping as a result of the following mechanisms: a) viscous fluid action (intermolecular shearing), b) transmission (lack of total reflection) at line terminations, junctions, diameter changes, and c) piping resistance (pipe roughness, restrictions, orifices). Therefore, damping of acoustic modes may be accomplished by placement of resistance elements, such as an orifice, which will work most effectively at velocity maxima. Note that acoustic resonances of piping systems for constant-speed compressors can usually be adjusted to detune them from the compressor harmonic frequencies by locating the resonances between harmonics and avoid pulsation amplification. However, for a variable speed compressor, detuning resonances becomes impossible and requires an acoustical filter.
Lumped acoustic elements
Certain components of the piping system may be viewed as lumped elements. Lumped acoustic elements are: a) the acoustic compliance, represented by a volume which acts as a stiffness or storage element and opposes a change in applied pressure; b) the acoustic inertance, characterizing a mass of gas contained in a relatively small diameter pipe which, when forced into motion, has the property of opposing a change in volume velocity; and c) the acoustic resistance, an orifice which dissipates energy when the gas is forced through the smaller diameter opening. These acoustic elements are directly analogous to the mechanical stiffness,
270
DYNAMICS OF MACHINERY
mass and damping elements, or to the electrical capacitance, inductance and resistance. Moreover, the acoustic volumetric flowrate is analogous to the mechanical displacement and the electrical current, while the acoustic pressure is analogous to the mechanical force and the electrical voltage. Systems consisting of only the two reactive elements (a gas inertia and compliance, a mass-spring system, an L-C circuit) are one degree of freedom oscillators. The equations of their natural frequencies are essentially equivalent.
14.3.3.3 Pulsation control in piping systems
The control of pulsations can be accomplished by judicious use of filters and tuned absorbers. Both of these types of acoustic elements amplify pulsations at their resonant frequencies; however, by their careful design and application, they can be used to attenuate energy.
271
A filter is designed such that its resonance is located at a frequency where the pulsation energy does not exist, and the maximum attenuation frequencies are located where the pulsation energy does exist. A tuned absorber uses a resonant component to take energy from the main system and relocate the resonance where it is controllable. The attenuation characteristics of four acoustic components are given in Fig. 14.45. The fundamental acoustic properties of piping components defined previously can be used to describe methods of controlling pulsations in piping systems. These methods include: a) use of side-branch resonators (Helmholz resonators), b) use of a surge volume for compressor cylinders, c) use of baffles and choke tubes within surge volumes, and d) use of dissipative components such as orifice plates, perforations, etc.
Helmholz resonators
The side-branch resonator (Fig. 14.45, a) can be an effective dynamic absorber in an acoustic system. It is a choke-volume system which, attached to a pipe, creates an antiresonance. For long neck resonators
fr =
a 2
A . L +1 2 A
(14.27)
Its use should be limited to constant speed systems where the resonator is tuned to a major frequency of pulsation. The resonator will pull pulsation energy out of the main line. However, the pulsations will be amplified in the resonator. It must be mechanically restrained to prevent vibrations in the cantilever mode. Acoustic resonances of the nozzles to a filter bottle in a reciprocating gas compressor generally have a strong response, because the pulsating flow from the cylinder flows directly into the nozzle. The nozzle and cylinder are similar to the elements of a Helmholtz side branch resonator, where the cylinder internal passages and clearance pockets form the volume and the gas in the nozzle is the oscillating mass. The nozzle resonant frequency can be estimated from the equation for the Helmholtz resonator (14.27). Because the cylinder generates strong pressure pulses over a range of harmonic frequencies, the probability is high that one of the harmonic frequencies will match the nozzle resonant frequency. In high flow, multi-cylinder compressor stages where several cylinders discharge into a surge volume or filter bottle, nozzle resonances may be near the bottle passbands simply because of typical dimensions.
Surge volumes
A surge volume (Fig. 14.45, c) can be quite effective in attenuating pulsations of a compressor, particularly if it can be located near the discharge flange.
272
DYNAMICS OF MACHINERY
Although it has qualities similar to a filter, it is not a true filter. The maximum attenuation of the outlet pulsations occur away from the resonant frequency. The attenuation factor is approximately
pin = pout
where the expansion ratio m =
1+
1 1 m , 4 m
(14.28)
D2 d2
1 + 2
V
(14.29)
where
j =
Aj L j + 1 2 Aj
(14.30)
and a acoustic velocity, V chamber volume, L j - choke tube lengths, A j - choke tube area. Economic considerations limit the size of surge bottles, and therefore impose practical limits on the degree of the overall acoustic attenuation which can be achieved. While the surge volume reduces the outlet pulsations, the pulsations within the volume can be at resonance, so that it must be adequately supported. Reactive acoustic filters An acoustic filter is designed to reflect as much of the incident energy as possible and, thereby, transmit as little as possible. One of the simplest low-pass filters is a pipe with either a constriction or an added enlarged section. They are commonly used in the design of automobile mufflers, gun silencers, and soundabsorbing plenums used in the ventilating systems. Another filter is a bottle with an internal baffle and a choke tube (Fig. 14.45, d). One of the most common types of acoustic filters used in reciprocating equipment involves the use of two volumes joined by a relatively small diameter pipe, to create a volume-choke-volume filter. Figure 14.46 shows various forms of the volume-choke-volume filter. If properly designed, the filter components behave as basic lumped acoustic elements, i.e. the bottles behave as acoustic compliances and the choke acts like an acoustic inertia. These lumped characteristics are valid as long as the excitation frequencies are below the open-open resonant frequency of the choke tube and the closed-closed resonant frequencies of the bottles.
273
These devices have pulsation response characteristics at low frequencies like that shown in Fig. 14.47. At frequencies above its resonance frequency, known as the Helmholtz frequency f H , transmitted pulsation levels drop off rapidly.
The Helmholtz resonant frequency of an ideal filter with no piping attached is given by
fH =
a 2
Ac Lc
1 + 1 , V1 V 2
(14.31)
where Lc = L c + 0.6 d c , d c - the choke diameter, L c - the length of choke tube, A c - the area of choke tube, V 1 - the volume of primary bottle and V 2 - the volume of secondary bottle. For equal volumes, the Helmholtz frequency is approximately [14.11]
274
DYNAMICS OF MACHINERY
fH =
a 2
c
V1
(14.32)
c =
Ac . Lc + 1 2 Ac
(14.33)
In addition to the Helmholtz resonance of a two-chambered filter, other internal resonances of elements such as compressor internals, nozzles, etc., may exist and will have the effect of passing particular frequencies. They produce peaks (pass bands) in the high frequency portion of the filter frequency response (Fig. 14.45, d). The number of pass bands occurring can be minimized by making the effective length of the choke equal to the lengths of the bottles. In low mole weight gas systems, reactive filters are impractical due to the high speed of sound values. Pulsation control can be accomplished through the use of ample surge volumes and resistive or pressure drop elements. Dissipative devices The most frequently employed dissipative device is the orifice plate. Forcing the flow through small openings, substantial pressure drops can be obtained. Orifices are the cheapest pulsation control device and the one most amenable to quick fix solutions. If a pulsation problem is uncovered in the field, it is relatively easy to add an orifice to the system (usually at a flange), compared to adding an accumulator or acoustic filter. Of course, the likelihood that simply placing an orifice in the most available location will solve the problem is not very high. Their sizing and placement has to be guided through an accurate acoustic simulation of the entire system. The effectiveness of dissipative devices is frequency-dependent. Their performance is better at high frequencies. Orifices are most effective when placed at or near the location of a velocity antinode in the mode shape of the mode to be attenuated. In order to provide some appreciable damping to the system, the orifice diameter should be not larger than one-half of the pipe diameter [14.15].
275
support structure to change its natural frequency or to revise the piping layout to change the location where the harmonic force is applied.
14.3.4.1 Design separation margins for natural frequencies
Field experience has confirmed that a 10% shift of the natural frequency away from resonance results in acceptable vibration levels (a reduction by a factor of five to ten, depending on damping) when the primary cause of the high vibration was resonance. Allowing for 10% uncertainty in natural frequency predictions, a 20% design separation margin is recommended. The operation of double acting compressors results in significant inherent forces at one and two times rotational speed. Resonance could be avoided at these frequencies if the predicted natural frequencies should be at least 20% above two times the compressor rotational speed. The two design separation margin guidelines adopted by the API 618 standard are: a) minimum predicted natural frequencies should be greater than 2.4 times maximum compressor rotational speed, and b) predicted natural frequencies should be separated at least 20% from frequencies with significant excitation forces.
14.3.4.2 Piping span natural frequencies
Actual piping span natural frequencies deviate from the theoretical beam natural frequencies, since the configurations that exist in typical plant piping have boundary conditions that differ from ideal values. Nevertheless, ideal beam theory gives a valuable starting point for understanding piping vibration behavior. Simplified natural frequency formulas can be used to evaluate the piping system with a minimum of detailed computer analyses. Straight piping spans For a straight uniform piping span, the natural frequency can be calculated using the following relationship f =
EI , Al 4
(14.34)
where: f is the span natural frequency, E Youngs modulus, density, A pipe cross section area, I cross section moment of inertia, l span length, frequency factor. By substituting the material properties for steel, E = 30 106 lb in 2 and
276
DYNAMICS OF MACHINERY
where k = I A is the radius of gyration, inches, and L is the length of span, ft (!). Note that this equation does not include the weight of the fluid and the insulation. The frequency factors, , for calculating the first two natural frequencies for ideal straight piping spans are given in terms of the overall span length in Fig. 14.48.
277
Piping bends The natural frequencies of selected pipe configurations with piping elbows (L-bends, U-bends, Z-bends, and three dimensional bends) were calculated using the ANSYS finite element program to generate frequency factors for the first two modes of vibration. Values are given in Fig. 14.48 for bends with equal span length and a total length L, calculated using a curved beam (elbow) element at corners. Frequency factors for a range of bend aspect ratios are published in reference [14.17]. Effect of concentrated masses Applying Rayleighs method, the first natural frequency of a beam with a concentrated mass can be calculated by
fw = f P 1+ W
(14.36)
where f w is the pipe span natural frequency with concentrated weight, f pipe span natural frequency without concentrated weight, P concentrated weight, W weight of beam span, - weight correction factor. Weight correction factors to be used in calculating the natural frequencies of ideal piping spans for weights at the maximum deflection locations are given in Fig. 14.49. If two weights are located in one span, Dunkerleys formula can be used to calculate the effect of the second weight. The frequency for one weight P1 is f1 = f 1+ P1 W . (14.37)
If the second weight in the span is considered by itself, the equation is f2 = f 1+ P2 W . (14.38)
The frequency for the span with both weights is given by the following equation 1 f 12 . (14.39) +2 = 1 1 1 + f12 f 22 f 2 Correction factors for non-ideal end conditions are suggested in [14.17].
278
DYNAMICS OF MACHINERY
279
With the above approach, clamp spacings can be selected which ensure that the piping spans will be resonant above some selected frequencies. Table 14.2 [14.13] gives the recommended maximum clamp spacing for minimum natural frequencies from 10 to 50 Hz. For complicated piping configurations a finite element analysis is required. These include flange flexibilities, flexibility of structures on which pipe supports are mounted, branch connection flexibilities, dynamic pipe-soil interaction, compressor frame flexibility, etc.
Table 14.2
In cases where high vibrations are noticed, the engineer must have some simple criteria to judge the severity of vibrations. Screening criteria have been developed to eliminate the necessity of a comprehensive analysis of every span in the piping system. API 618 Design Vibration Guideline The piping system design vibration criteria adopted by API Standard 618 is shown in Fig. 14.50. The figure is based on: a) a constant allowable vibration amplitude of 0.5 mm peak-to-peak for frequencies below 10 Hz, and b) a constant allowable vibration velocity of approximately 32 mm s peak-to-peak for frequencies between 10 and 200 Hz. In fact, no single vibration guideline can completely account for the wide variation in geometry and supporting of actual compressor and piping systems. The adopted design vibration limit balances between typically acceptable vibration levels for large slow speed versus smaller high speed compressor piping systems.
280
DYNAMICS OF MACHINERY
Vibration displacement amplitude as a function of stress The severity of piping span lateral vibration displacement amplitudes can be assessed by comparing the maximum resonant vibration-induced dynamic stresses to an allowable fatigue limit stress. The low cycle fatigue curves for carbon steel given in the ASME USAS B31.7-1969 (Fig. 14.51) can be used to obtain an acceptable fatigue limit stress [14.20].
281
The ANSI/ASME Code OM3-1987 [14.21] uses this stress versus cycles-tofailure curve as a basis for specifying criteria for evaluating the vibration-induced stresses in nuclear power plant piping for preoperational and startup testing. The code defines the allowable fatigue stresses as 0.8 times the allowable alternating stress intensity at 10 6 cycles which is 13,000 psi zero-to-peak (89.5 MPa). The vibration-induced dynamic stresses in a piping span vibrating at resonance has been shown to be related to the maximum vibration amplitude in the span [14.18]. The relationship is given by the equation below
= Kd y
D (SCF ) , L2
(14.40)
where dynamic stress, psi, K d deflection stress factor, y maximum vibration amplitude measured between nodes (normally at supports), mils, D outside pipe diameter, inches, L span length, ft, SCF stress concentration factor ( 1 psi = 6.895 kPa , 1ft = 0.3048 m , 1in = 25.4 mm ). The deflection stress factor K d is a function of the boundary conditions and the vibration mode shape at resonance. The deflection stress factors for the first two modes of the ideal Bernoulli-Euler beams and the piping configurations with elbows are given in Fig. 14.48 for equal leg lengths and in reference [14.17] for various values of the leg length ratio. The allowable vibration amplitude can be calculated based on the fatigue limit using the relationship [14.18] L2 a , ya = (14.41) (SCF )(SF ) K D d where a allowable stress, psi, K d deflection stress factor, SCF stress concentration factor and SF safety factor. If the API 618 allowable of 13,000 psi zero-to-peak is used as the endurance limit combined with a stress concentration factor of 4.33, a safety factor of 2, and a stress deflection factor of 3000 (applicable for a fixed-fixed pipe), the allowable vibration in peak-to-peak mils can be calculated. Equation (14.41) becomes
ya =
L2 . D
Rule of Thumb
(14.42)
This can be used conservatively as a screening criterion for straight runs of piping or for piping with bends. Note that the pipe diameter is measured in inches, while the span length is measured in feets. This criterion is overly conservative for cantilever beams. If the measured vibrations exceed the screening criterion, the vibration induced stresses are not necessarily excessive, and more detailed calculations using computer programs are required.
282
DYNAMICS OF MACHINERY
Vibration velocity amplitude as a function of stress In a piping span vibrating at resonance, it is also possible to relate the maximum stress to the measured velocity [14.16]. In order to develop a closed-form solution of the dynamic stress as a function of the velocity, the radius of gyration has to be expressed as a function of the outside diameter of the pipe. A comparison of the radius of gyration for different sizes of pipe versus the diameter shows that, for a significant range of pipe sizes, this is approximately 0.34 D0 , where D0 is the outside pipe diameter. By making this substitution for the radius of gyration, the stress in an ideal beam can be expressed as a constant K v , referred to as the velocity stress factor, multiplied by the maximum velocity measured in the piping span, times the stress concentration factor
= K v v SCF .
(14.43)
where dynamic stress, psi, v is the maximum velocity in the pipe span, in sec . Some velocity stress factors are given in Fig. 14.48 for ideal straight spans and piping bends with equal legs. Values for different leg length ratios are presented in reference [14.17]. The actual velocity is a function of the fatigue limit and is given in equation (14.44) where a safety factor (usually 2) is included to account for system unknowns
v=
K v SF SCF
(14.44)
Based on an allowable fatigue limit of 13,000 psi zero-to-peak, a maximum velocity stress factor of 318, a stress concentration factor of 5, and a safety factor of 2, the allowable zero-to-peak velocity is equal to [14.17]
va =
(14.45)
(14.46)
Dynamic strain criteria For typical piping with an ultimate tensile strength of less than 80,000 psi, the fatigue limit from ASME B31.7 is 26,000 psi peak-to-peak. Since the stress is equal to the dynamic strain times the modulus of elasticity, the allowable strain would be 866 10 6 in in . If a stress concentration factor of 4.33 and a safety factor of 2 is used, a safe allowable strain reading for a gage mounted near the area of high stress concentration would be 100 10 6 in in or 100 microstrain. The guidelines for the interpretation of the strains are the following [14.19]:
283
Strain:
Since the span natural frequency is an inverse function of the square of the span length, the most effective way to solve a mechanical resonance is to add pipe restraints, such as piers, supports or clamps to shorten the vibrating span. Many times, temporary bracing with hydraulic jacks, wooden beams and wedges can be used to confirm that a support at a particular location will reduce the vibrations. Some of the general guidelines which can be used in selecting modifications to detune the mechanical resonances are outlined below [14.12]: 1. Pipe supports and clamps should be installed on one side of each bend, at all heavy weights, and at all piping discontinuities. 2. The support and clamp stiffness should be adequate to restrain the shaking forces in the piping to the desired amplitudes and should be greater than twice the basic span stiffness in order to effectively enforce a node at the support location. 3. Vents, drains, bypass, and instrument piping (appurtenances) should be braced to the main pipe to eliminate relative vibration between the small-bore piping and the main piping. 4. Restraints, supports, or gussets should not be directly welded to pressure vessels or the piping unless they are subjected to the appropriate heat treatment. It is more desirable to add a saddle-type clamp around the pipe and weld the braces to the clamp. 5. To resist vibration, the piping clamps should have contact with the pipe over 180 degrees of the circumference. Rubber or gasket-type material can be used between the clamp and the pipe to improve the contact. 6. The piping span natural frequency should not be coincident with the excitation frequencies.
14.3.4.5 Design procedure using reactive pulsation control
The use of reactive filtering in conjunction with control of mechanical natural frequencies results in a safe margin between significant pulsation-induced forces and mechanical natural frequencies. The procedure for designing reactive filters consists of the following steps: 1. Determine choke diameter and approximate length based on allowable pressure drop.
284
DYNAMICS OF MACHINERY
2. Determine volume-choke-volume filter design to filter all harmonics of running speed. Generally, the filter frequency is set at 50-80 percent of 1 running speed for heavy gases, or between 1 and 2 running speed for lighter gases. 3. Perform pulsation simulation to determine pulsation levels and acceptability of filter design. Determine maximum frequency f p of significant
( )
( )
4. Determine minimum allowable mechanical natural frequency ( f m ) based . Set f m 1.5 f p . 5. Locate vibration restraints near all concentrated masses (e.g., valves).
6. Use pipe support span tables (Table 14.2) to determine additional support locations based on f m . 7. Determine minimum stiffness (k) of each support: k 2 lateral span 2 48 E I stiffness = ( l = support span). l3 Use of this acoustic filtering concept in conjunction with control of minimum piping mechanical natural frequencies provides a high level of confidence that resonance will be avoided.
References
14.1. Den Hartog, J. P., Mechanical Vibrations, 4th ed., Dover, New York, 1985. 14.2. Thompson, W. T., Vibration. Theory and Applications, George Allen & Unwin, London, 1966. 14.3. Magrath, H. A., Rogers, O. R., and Grimes, C. K., Shock and vibration in aircraft and missiles, Ch. 47 in Shock and Vibration Handbook, C. M. Harris and Ch. E. Crede, eds., McGraw-Hill, New York, 1961. 14.4. Crandall, S. H., Rotating and reciprocating machines, Ch. 58 in Handbook of Engineering Mechanics, W. Flgge, ed., McGraw-Hill, New York, 1962. 14.5. Richart, F. E. Jr., Hall, J. R. Jr. and Woods, R. D., Vibrations of Soils and Foundations, Prentice Hall, Englewood Cliffs, N.J., 1970. 14.6. Maas, H. and Klier, H., Krfte, Momente und deren Ausgleich in der Verbrennungskraftmaschine, Die Verbrennungkraftmaschine, Neue Folge, Band 2, Springer, Wien, 1981. 14.7. Rade, M., Diagnosis of an auxiliary diesel engine vibration problem with signature analysis, Machine Vibration, vol.1, 1992, p.58-63.
285
14.8. * Reciprocating Compressors for Petroleum, Chemical, and Gas Industry Services, ANSI/API Standard 618, 5th ed., 2007. 14.9. Bloch, H. P., Compressors and Modern Process Applications, Wiley, New York, 2006. 14.10. Lifson A. and Dube, J. C., Specifying reciprocating machinery pulsation and vibration requirements per API-618, American Gas Association Distribution/Transmission Conference, Las Vegas, Nevada, May 4-6, 1987. 14.11. Wachel, J. C. et al, Vibrations in Reciprocating Machinery and Piping Systems, Engineering Dynamics Incorporated, Technical Report EDI 85-305, 2nd ed., 2nd Printing, 1988. 14.12. Wachel, J. C. and Tison, J. D., Vibrations in reciprocating machinery and piping systems, Proc. 23rd Turbomachinery Symposium, Texas A&M University, College Station, Texas, 1994, p.243-272. 14.13. Atkins, K. E., Pyle, A. S. and Tison, J. D., Understanding the pulsation and vibration control concepts in the new API 618 Fifth Edition, 2004 Gas Machinery Conference, Albuquerque, New Mexico, Oct. 4-7, 2004 14.14. Corbo, M. A. and Stearns, Ch. F., Practical design against pump pulsations, Proc. 22nd International Pump Users Symposium, Turbomachinery Laboratory, Texas A&M University, Feb.28-March 3, 2005, p.137-177. 14.15. Price, S. M., and Smith, D. R., Sources and remedies of high-frequency piping vibration and noise, Proc. 28th Turbomachinery Symposium, Texas A&M University, College Station, Texas, 1999, p.189-212. 14.16. Wachel, J. C., Piping vibration and stress, Proc. Machinery Vibration Monitoring and Analysis Seminar , Vibration Institute, April 1981. 14.17. Wachel, J. C., Morton, S. J. and Atkins, K. E., Piping vibration analysis, Proc. 19th Turbomachinery Symposium, Texas A&M University, College Station, Texas, 1990, p.119-134. 14.18. Wachel, J. C, Displacement method for determining acceptable piping vibration amplitudes, International Pressure Vessels and Piping Codes and Standards, PVP-vol.313-2, ASME 1995, p.197-208. 14.19. Wachel, J. C., Field investigations of piping systems for vibration-induced stresses and failures, Pressure Vessels and Piping Conference, Orlando, Florida, June 27 July 2, 1982, ASME Bound Volume No.H00219, 1982. 14.20. * Nuclear Power Piping, USAS B31.7-1969 ASME Code, New York, 1969. 14.21. * Preoperational and Initial Startup Vibration Testing of Nuclear Power Plant Piping Systems, ANSI/ASME Operations & Maintenance Standards/Guides Part 3, ASME, New York, 1990.
286
DYNAMICS OF MACHINERY
14.22. Wachel, J. C. and Smith, D. R., Vibration troubleshooting of existing piping systems, Engineering Dynamics Incorporated Report 91903, July 1991. 14.23. * Nivele admisibile de vibraii pentru conducte din instalaii chimice i rafinrii, Ministerul Industriei Chimice, NTR 11230-85, ICITPR Ploieti, iulie 1985.
Index
Accelerometers 70 Acceptance region plots 111 Acoustic resonance 265, 268 Amplitude demodulation 28 probability density 18 Angular contact bearings 4 API 618 279 API standards 186 Axial compressors 145 Balance quality grades 225, 228 Balancing of rotors 203 in N+2 planes 229 Bearing frequencies 6 wear 14 Blake Severity Chart 192 Bode plots 108 Brinelling 16 Burning 56 Cage damage 15 Cascade plots 109 Case crushing 57 Centrifugal compressors 142 fans 144 pumps 141 Cepstrum analysis 35, 69 Condition indicators 59 monitoring 115 Connecting rod 242 Contact ratio 42 Couple unbalance 205, 223 Coupled industrial machines 170, 178 Cracked shafts 138 Crank mechanism 241 Creeping 16 Crest factor 17, 62 Cylinder-manifold system 256 Cylinder stretch forces 259 Demodulation 63 Denting 15 Diagnosis process 120 Diesel engine 251 Discharge bottle 258 Discoloration 56 Dresser Clark Chart 194 Dynamic unbalance 207 Eddy current transducer 88 Electrical machines 151 Envelope analysis 28 Enveloping 62 Energy operator 62 Errors 54 Fault diagnostics 121 Feature extraction 59 Flaking 14 Fluid induced instabilities FM0 63 FM4 64 Fretting 16 Frosting 56 Full spectrum plots 104
127
Galling 55 Gas pressure excitation 237 torque 238 Gas turbines 150, 172, 180 Gear errors 54 Gears 39, 185 Ghost frequencies 48 Glazing 15 Grooving 16 Half spectrum plots 104 Half-wave resonance 267 Heavy spot 205 Helmholtz resonator 271 High spot 205 Hydraulic machines 172, 181 Identations 14 Industrial buildings 187 Inertia torque 239
288 Influence coefficient method ISO 1940 225 ISO 7919 176 ISO 10816 168 Kurtosis 22, 61 Laser alignment 159 Looseness 135, 137 Lumped acoustic elements 209, 217
DYNAMICS OF MACHINERY Reciprocating compressors 152, 256 machines 174, 237 Residual unbalance 227 Reverse dial indicator method 156 Rigid rotors 226 Rolling element bearings 1 Root mean square 17, 61 Rotor-bearing clearance 136 Rubbing 130 Run-to-failure maintenance 117 Scoring 55 Scuffing 55 Shaft alignment 155 Shock pulse method 30 Shaking forces 260 Sieberts construction 215 Single cylinder engines 237 plane balancing 208 Skewness 21 Skidding 2 Smearing 14 Spalling 14, 57 Spike energy 25 Spinning 16 Standard deviation 21 Standing waves 265 Static unbalance 205, 223 Statistical moments 21 Steam turbines 146, 169, 177 Surge volume 271 Tapered roller bearings 7 Three-trial-mass method 215 Time base plots 101 Time-frequency analysis 72 Time synchronous averaging 59 Tooth deflection 46 engagement 45 fracture 58 wear 47 Transducers 85 Trend plots 107 Two-plane balancing 217 Unbalance 121 couples 244, 246 force 243, 246 tolerances 225 VDI 2056 166
269
M6A 66 M8A 66 Machine deterioration 115 Maintenance strategies 117 Mass unbalance 204 Misalignment 123 Modal balancing 232 Mode shape plots 106 Modulation amplitude 49 effects 48 frequency 51 Multi cylinder engines 246 Multiplane balancing 229 NA4 NB4 65 67
Oil debris analysis 67 Oil whip 128 whirl 128 Orbits 101 Orifice plate 274 Piping vibration 274 Pitting 56 Plots 101 polar 108 Predictive maintenance 118 Preventive maintenance 117 Proximity probes 88, 97 P-V card 263 Pulsation analysis 261 control 270 excitation mechanism 261 Quarter-wave resonance 267 Quasi-static unbalance 206 Radial preload 123 Rathbone Chart 164 Reactive acoustic filters
272
INDEX VDI 2059 196, 197, 198 Vector balancing 208 Velocity pickups 91, 97 Vibration intensity 188 limits 163 magnitude 80 measurement 75 peak-to-peak 83 severity 81, 195 charts 164 Volume-choke-volume filter Wear 56
289
272