• 沒有找到結果。

Analysis of Torsional Vibration Systems by the Extended Transfer Matrix Method

N/A
N/A
Protected

Academic year: 2021

Share "Analysis of Torsional Vibration Systems by the Extended Transfer Matrix Method"

Copied!
6
0
0

加載中.... (立即查看全文)

全文

(1)

Yuan Mao Huang

Professor.

C. D. Horng

Research Assistant. Department of Mecfianical Engineering, Nationai Taiwan University, Taiwan, Republic of Cfiina

Analysis of Torsional Vibration

Systems by the Extended

Transfer Matrix Method

This study applies the extended transfer matrix method and Newton-Raphson tech-nique with complex numbers for torsional vibration analysis of damped systems. The relationships of the vibratory amplitude, the vibratory torque, the derivatives of the vibratory angular displacement and the vibratory torque between components at the left end and the right end of the torsional vibration system are derived. The derivatives of the vibratory angular displacement and the vibratory torque are used directly in the Newton-Raphson technique to determine the eigensolutions of systems that are compared and show good agreement with the available data.

Introduction

The well known Holzer method [1] that is a simple and systematic approach to determine the frequencies and mode shapes of systems was used originally for analysis of an un-damped system. Den Hartog and Li [2] improved the Holzer method by using complex numbers. Pestel and Leckie [3] intro-duced the transfer matrix method with the point and field trans-fer matrices. Eshleman [4] developed a technique for the solu-tion of torsional vibrasolu-tion problems encountered in design, de-velopment and diagnosis of engine systems. Sankar [5] used the extended transfer matrix method to analyze an undamped system. Doughty and Vaf aee [6] used complex numbers in the transfer matrix with the Newton-Raphson technique to deter-mine the eigensolutions of a damped system. Dawson and Da-vies [7] used the extended transfer matrix with the Newton-Raphson technique to analyze the torsional vibration for un-damped systems. Doughty [8] considered the systems of prop-erly counting for the shafting inertia in the torsional vibration analysis by using a Rayleigh-type approximation for the dis-placement between stations. Neilson and Barr [9] studied im-balance effects in rotor systems under high angular accelera-tions. Yun and Lee [10] studied the dynamic analysis of flexible rotors subjected to the torque and the force. Lu [11] analyzed the transfer matrix method for the nonlinear vibration of the rotor-bearing system. Kang and Chung [12] studied the transfer matrix method for the general periodic torsional vibration of rotor systems. Huang [13] analyzed the engine excitation on a transmission with a damper for the converter operation. Doughty [14] analyzed the steady state torsional response with the viscous damping.

Although Doughty and Vafaee [6] used complex numbers with the Newton-Raphson technique to calculate the eigensolu-tions of damped systems, they used the conventional transfer matrix method. A complicated multiplication should be per-formed before the Newton Raphson technique is applied. The multiplication results in a nominal equation that is well known as the eigenvalue equation. The term number of the equation increases and the terms become complicate if the number of rotors in the system becomes huge. Dawson and Davies [7] used the extended transfer matrix with the Newton-Raphson technique. Nevertheless, their study was limited to undamped systems. The damping effect exists in the most of real systems. It is important to include the damping effect to better predict the torsional vibration of systems. Therefore, this study uses Contributed by the Technical Committee on Vibration and Sound for publica-tion in the JOURNAL OF VIBRATION AND ACOUSTICS . Manuscript received Sept, 1998. Associate Technical Editor: D. J. Inman.

complex numbers to extend the extended transfer matrix method with the Newton-Raphson technique to analyze the torsional vibration for damped systems. This method eliminates the oper-ation of the inverse matrix because the derivatives of angular displacement and the torque are used directly with the Newton-Raphson technique to determine the eigenvalues of torsional vibration systems.

Method of Approach

A mass elastic system with both ends free is shown in Fig. 1. The symbols / , k, c and d are the mass moment of inertia of a disk, the stiffness of a shaft, the internal damping coefficient and the external damping coefficient, respectively.

It is assumed that disks are rigid bodies with infinite stiff-nesses, that shafts are slender with negligible mass moments of inertia, that the damping effect is proportional to the relative vibratory angular speed, and that the damping coefficient is constant. The equations of motion for a disk and a shaft as shown in Fig. 2, respectively, are

0f- = 0f

r f - r f = /, ^ 0 , . + rf<^ (1)

rfr dt

and Tf = rf+i

Tf = A:,(0f;, - 0f) + c, 4 {®t^ - ©f) (2)

dt

The angular displacements and torques of stations can be ex-pressed as [6]

f 6\ + je\

9'2+jd'2 I 0:, + iO'n \ " J 't J { 0 ( 0 } = { > e = {Q{t)]e ( - f f + , ; u ) ( (3) and Tl + JT'z , r:, + JT'„ _

> e i~o-^juj)l {T{t)}e (~a + jij)l (4)

{T(t)] = .

(2)

cl

D

-k i C2 k2 C l ki -jm W7 'W' d i d2 d i cn-1 kn-1 dn-1 ^ dn I I I2 l i I n - 1 In

Fig. 1 Mass elastic system witli botli ends free

and the vibratory torque. The superscripts r and ; indicate the real part and the imaginary part, respectively. The extended state vector of a component can be expressed as

9' r' 36' dr' dB' dr' d9' dr' 09' dr' da da da da dui du> du du

(8)

The field transfer matrix, F, and the point transfer matrix, P, are defined as functions of/, k, c, d, a, u> [ 3 ] . Differentiating the state vector with respect to a and ui, respectively, and com-bining with Eq. ( 7 ) , the relationship of the extended state vec-tors for a shaft and a disk, respectively, are

T^ e;R

eh

ki , Ci

Fig. 2 Schiematic diagrams of a disl< and a sliaft

Zi

r z 1

dZ < da dZ . i ^ t ^ , L

, ~

1+1 OF da OF 0 0"^ F 0 0 F X . 1

r z 1

dZ da \ dZ (9) and

eigenvalue, \, respectively. Substituting Eqs. (3) and (4) into Eqs. (1) and (2) yields

0f = ef

Tf-T'^ = Id-<T+jt^ydi+jdi(-a +JLj)e'- (5) and

Zf

' z • dZ < da dZ . dijj, R > =

/

P 0 dp da ^ 0 duj 0 " 0 P X . ' z • dZ da dZ (10)

The relationship between the right side of the {i + 1 )th compo-nent and the left side of the (th compocompo-nent is

T'^^T^

T^ = kiei, - ef) + c:(~a+jujKeu^ - ef) (6)

The state vector of a component is defined as

Z-<^:}=< T (7)

r z ]

dZ < da dZ duj J R

>

i + i TOO' 3 ^ T- n d^ ' '

f 0 ^

duj X .

' ^ 1

dZ da dZ

which is a column matrix of the vibratory angular displacement as

(11)

where T is the extended transfer matrix. The elements in T are the functions of /, k, c, d, a, to. Using the transfer matrix method, the relationship between the right side of the wth com-ponent and the left side of the first comcom-ponent can be obtained

N o m e n c l a t u r e A = matrix product a, b, e, g, h = element in the matrix

c = internal damping coeffi-cient

d = external damping

coef-ficient

/ = mass moment of inertia

k = stiffness

T = torque matrix; transposed

transfer matrix or torque / = time

Z = state vector

Z = extended state vector

a = absolute value of real part of eigen- ij = ith row and y'th column element of

value

X. = eigenvalue of matrix 0 = angular displacement vector

0 = angular displacement magnitude

T = torque magnitude

a = phase angle

6 = external forced circular frequency ui = imaginary part of eigenvalue Subscripts e = external i = ith component matrix Superscripts c = cosine component i = imaginary part

L = left side of component

R = right side of component r = real part

(3)

da dr'' da

del

da dr' da

ar

du dr' duj

dei

dio

dr_

du a-i\ 041 «51 « ? ! flsi 0121 «12 «22 «32 «42 fl52 fl62 ^72 082 092 Ol02 a i 3 «23 «33 «43 ^ 5 3 063 073 083 O93 O1O3 fl]12 O1I3 O122 O123 fll4 024 034 O44 0S4 064 074 084 O94 O104 0 [ | 4 Ol24 0 0 0 0 O i l 021 031 041 0 0 0 0 0 0 0 0 012 0 0 0 0 Ol3 022 023 032 042 0 0 0 0 033 043 0 0 0 0 0 0 0 0 Ol4 024 034 044 0 0 0 0 0 0 0 0 0 0 0 0 On 02I 031 041 0 0 0 0 0 0 0 0 012 0 0 0 0 0 0 0 0 Ol3 022 023 032 033 042 043 0 0 0 0 0 0 0 0 Ol4 024 034 044, dO'' da dr' da

ml

da dr' da dW_ dr' duj

dei

dri

duj (12) Z j = AZ, (13)

The elements of the matrix are functions of / , k, c, d, a and w. If the extended state vector at the left side of the first compo-nent is

Z t = {1 0 0 0 0 0 0 0 0 0 0 0 } ' ' and Eq. (14) is substituted into Eq. (12), we can get

(14)

1"

da du)

dT'„ dT'„ .da dui

and substituting Eq. (15) into Eq. (17) yields

O6I Oioi Ogl O121 -fl21 - 0 4 1 (17) (18) T „ = 041 da dr', da dui dK du> da 9fl41 ^ da dan du! da2\ dw = 061 = Ogl — Oioi (15) If the right side of the «th component is free, the vibratory torque of the «th component is zero. Therefore, it can be written

and

T',Xcr, W) = 0

r'„(o-, w) = 0 (16)

Equation (16) is known as the eigenvalue equation of a damped system.

Using the Newton-Raphson method with

By using the trial and error method with assumed initial values of a, w, the accurate eigenvalues can be obtained when T'^, and r i in Eq, (17) converge.

If the system shown in Fig. 1 is the equivalent mass elastic system of a multi-cylinder engine, 6 stands for the excitation frequency and a, stands for the phase angle of the cylinder with respect to the first firing cylinder of an engine, the external torque can be expressed as

T . = T .^'•<*'-"/'

= (Te, cos «; COS 6t - Tel sln a, sin St)

+ jiTei COS a, sin 6t - T^i sin a, cos 6t) (19) The real part of the torque is

^[(T'ci + JT'ei)e''"] = Tel COS «/ COS 6t - T^ sin a, sin 6t

= Tei COS 6t + r% sin 6t

Hence, we have

T% = Tel c o s a,- = T'ei

and

(4)

Table 1 Mass moments of inertia of disi<s and stiffnesses of shafts for the first undamped system

tfw^^vwwv^ri^^fwvwww^'vwwwwvwwimn^vwm^^n

Station no. I (N-m-sQ k (MN-m/rad) 1 2 3 4 5 6 7 8 9 71.882 41.387 41.387 41.387 41.387 71.822 4401.1 1763.2 162.28 1.3614 1.3614 1.3614 1.3614 1.3614 2.1523 1.4789 0.8417 - Tei Sin Ui = (21)

similarly, the real part of the angular displacement is

R[(0'e, + 7^;:,)e'*'] = O'ei COS 6t - 0', sin 8t (22)

If a system is subjected to an external torque, the relationship between the right side of the nth component and the left side of the first component can be expressed as

T L 1 J > = fli ^1 e, g, / , «2 ^2 62 g2 fl tti h, e , g j fj ^4 ^4 e4 g4 / 4 , 0 0 0 0 1 (23)

where the elements in the matrix are functions of / , k, c, d, 6 and T(,. If both ends of a system are free, rli and T'„ are all equal to zero. Hence, we have

and fl20'l + 626*'i + /i2 = 0 04^'1 + e^O'i + hi = Q a,^'i + eid\ + h, = Q', (24) (25) The vibratory angular displacement of the first component can be determined from Eq. (24). Therefore, the vibratory angular displacements of all components can be determined.

Q^ 1.2 ^ 0.8 ? 0.4 > -0.4 •B -0.8 2:-1.2

first mode «currQnt

AHolzer X current •second mode ^ j^ n H d z e r 4 5 6 7 S t a t i o n niimber

Fig. 3 Relative vibratory angular displacements of Vt\6 first undamped system

Table 2 Mass moments of Inertia and stiffnesses of ttie second un-damped system

VVVWVWVVSlv^VVVMyvVIIVWVyV|WWVVVWVVI<VV«VVVVWV»VVWVVV>n«VVVVWf¥VV»VI^^

station no.

I (N-m-sQ

k (MN-m/rad)

1 2 3 4 5 0.680 0.680 1.360 2.260 0.136 0.170 0.290 0.079 0.090 ^.^ !!!', « . „ «

-Table 3 Natural frequencies and relatively vibratory angular displace-ments of the second undamped system

1st mode 2nd mode .?.fd.5}.o.'?.6 ^t*!.!?}.'?.'?.^ current MATUB current MATLAB current MATLAB current MATLAB ffl atio no. 1 2 3 4 235 rad/s ti 1 0.78 0.55 -0.81 235 512 512 839 839 956 rad/s rad/s rad/s rad/s rad/s rad/s

relative vibratory angular displacement 1 1 I 1 1 1 0.78 -0.05 -0.06 -1.82 -1.82 -2.65 0.57 -0.63 -0.63 -0.47 -0.47 0.89 -0.83 0.09 0.06 10.16 10.21 -0.04 956 rad/s 1 -2.65 0.88 -0.04 -0.89 -0.91 0.15 0.13 -160.1 -161 0.11 0.11 Results

The first system with the mass moments of inertia of disks and the stiffnesses of shafts of a six-cylinder engine with a flywheel and a converter listed in Table 1 is used for analysis. It is assumed that the system is undamped with both ends free. There is a rigid mode with an eigenvalue equal to zero for a system with both ends free. The calculated natural circular frequencies of the top four modes are 1001, 1312, 2353 and 3677 rad/s. The first mode and the second mode obtained by using the Holzer method are 1030 and 1330 rad/s, respectively. The top four natural circular frequencies obtained by using the MATLAB software program are 1003, 1336, 2354 and 3811 rad/s, respectively. The results of the relative vibratory angular displacements for the first and the second mode shapes are shown in Fig. 3.

The mass moments of inertia of disks and the stiffnesses of shafts of the second undamped system with both ends free are shown in Table 2. The results of eigensolutions of the second undamped system obtained from this analysis and from the MATLAB software program are shown in Table 3.

The mass moments of inertia, the stiffnesses and the damping coefficients of the third damped system obtained from Doughty

Table 4 Mass moments of inertia and stiffnesses of the third undamped system •W(WVWWVWWV«V*VI«VW I k e d station no.

i

2 3 0.4 — — 6 0.3 0.5 8000 10000 10 0 0 0

and Vafaee [6] with both ends free are shown in Table 4. The calculated results of the eigensolutions and the data obtained from Doughty and Vafaee [6] are shown in Table 5.

The fourth damped system with the mass moments of inertia, the stiffnesses and the damping coefficients of an engine with a generator obtained from Doughty and Vafaee [6] are shown in Table 6. Both ends of the damped system are free. The results of the eigensolutions for the system are shown in Table 7.

The mass moments of inertia and the external damping coef-ficients of disks, and the stiffnesses and the internal damping

(5)

Table S Results of the eigensolutions for the third damped system Table 8 Mass moments of inertia, stiffnesses and damping coefficients of the forced system

current calculated

results

Doughty and Vafaee's

results

stat-ion

no.

I

2

3

first mode second mode first mode

A*~ A ~ A."

-12.1+J161.3 -19.6+J246.3 -12,l+jl61.3

rad/s rad/s rad/s

second mode

X.=

-19.6+J246.3

rad/s

relative vibratory angular displacement

real imag real imag real imag

1.0 1.00 1.0

0.03 0.05 -1.21 0.33 0.03 0.05

-0.77 -0.14 0.79 -0.36 -0,77 -0.14

real imag

1.00

-1.21 0.33

0.79 -0.36

Table 6 Mass moments of inertia and stiffnesses of the fourth damped system tfW^tfWV^WWVWVWWWArtrfWVW^Wl^Art'VW^A^^^^AA'VVMnnAfWW^MWArVt^WmKWM^rtMAA^'

Station

no.

I (N-m-s^)

k

(MN-m/rad) C (N-m-s/rad)

d

(N-m-s/rad)

1

2

3

4

5

6

7

8

9

10

11

12

6

4.5

4.5

6

6

4.5

4.5

6

3.5

100

20

350

16

16

16

13

16

16

16

25

150

0.5

10

-330

550

Table 7 Results of the eigensolutions for the fourth damped system

Doughty

and

Vafaee's

results

current calculated

results

station

no.

1st mode 2nd mode 3rd mode 3rd mode

A; — A. A, A*

•1.6+j67,4 -0.3+J376.7 -8.0+j743,2 -8.0+J743.2

(rad/s) (rad/s) (rad/s) (rad/s)

1

2

3

4

5

6

7

8

9

10

11

12

real

1.00

1.00

1.00

0.99

0.98

0.98

0.97

0.96

0.95

0.95

-0.32

-0.38

relative vibratory angular displacement

imag real imag real imag real

0.00

-0.00

-0.00

-0.00

-0.00

-0.00

-0.00

-0.00

-0.00

-0.00

-0.01

-0.01

1.00

0.95

0.86

0.73

0.53

0.34

1.31

-0.08

-0.21

-0.23

-0.01

0.02

0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 -0.00 0.00

1.00

0.79

0.46

0.06

-0.45

0.00

-0.00

-0.01

-0.02

-0.02

-0.77 -0.02

-0.97 -0.01

-1.02

-0.92

-0.89

86.2

-4.65

0.00

0.02

0.02

-42.1

2.39

1.00

0.79

0.46

0.06

-0.45

-0.77

-0.97

-1.02

-0.92

-0.89

86.2

-4.65

imag

0.00

-0.00

-0.01

-0.02

-0.02

-0.02

-0.01

0.00

0.02

0.02

-42.1

2.39

station

no.

I

(N-m-s^)

k

(MN-m/rad) C (N-m-s/rad)

d

(N-m-s/rad)

1

2

3

4

5

1.7

2.5

2.3

1.4

3.7

0.30

0.30

0.35

0.22

0.40

5600

8200

12000

4000

-0.0

0.0

0.0

9.9

-coefficients of shafts in a system subjected to external torques

are shown in Table 8. If the disks of the system are subjected

to the external excitation torques,

TAt) = {

0

20

25

0

0

> cos 27/ + <

0

15

-20

0

0

sin 54/ (26)

it is known from Eq. (21) that the original coefficients of cos

27/ and the sign change of the original coefficients of sin 54/

in Eq. (26) should be used in the matrix. The angular

displace-ment of a components is

0(t) = U, cos 27/ + Vt sin 27/

+ U2 cos 54/ + V2 sin 54/ (27)

The results of the vibratory angular displacements of the

com-ponents and the data obtained by Doughty [14] are shown in

Table 9.

Discussion

The derivatives of the vibratory angular displacements and

the vibratory torques are used directly in the Newton-Raphson

technique to calculate the eigenvalues. The results are based on

the first order of approximation in the Newton-Raphson method.

If rJi and T'„ in Eq. (17) are less then lO"*" N.m, the values are

assumed to be converged.

This method eliminates the operation of the inverse matrix

because the derivatives of angular displacement and the torque

are used directly with the Newton-Raphson technique to

calcu-late the eigenvalues of systems. In addition to the zero elements,

there are 80 elements in the transfer matrix in Eq. (12).

Never-theless, only six elements, ajii <5!4i, flei. as\, flioi and ant, are

needed for Eq. (18) to calculate the eigenvalues. It is not

neces-sary to calculate all elements of the transfer matrix. Therefore,

using the Newton-Raphson technique with complex numbers in

the extended transfer matrix method should be better than using

the conventional transfer matrix method and then applying the

Table 9 Results of the coefficients of trigonometric functions

Angular Displacement (10 ) rad

station

no.

1

2

3

4

5

current calculated

results

Doughty's

results

U,

2.47

1.93

-1.31

-5.42

-16.3

U2

-1.48

-0.16

0.89

1.10

-0.18

V,

1.33

1.03

0.60

0.22

-1.69

V2

4.33

0.50

-4.39

-3,61

2.23

U,

2.47

1.93

-1.31

-5.42

-16.3

U2

-1.48

-0,16

0.89

1.10

-0,18

V,

1.33

1.03

0.60

0.22

-1.69

V2

4.33

0.50

-4.39

-3.61

2.23

(6)

Newton's equation to determine the eigenvalues of a system. However, there is no computer running time available from the conventional transfer matrix method. In the meantime no computer time can be recorded in the personal computer for comparison.

There are some disadvantages of using this methodology. An initial eigenvalue is required to calculate one eigenvalue of a system. Improperly assumed initial values of a and w may result in the overshooting that leads to difficulty in determining some eigenvalues. For an undamped system, the residual torque con-verges quickly. Since the deviation of eigenvalues for the damped system from the system with the damping coefficients assumed to be zero is small for the most cases, the eigenvalues obtained for the system with the damping coefficients assumed to be zero are the best assumed initial values for damped sys-tems. Only one eigenvalue can be obtained at one time. If there are three eigenvalues for a system, it is necessary to assume three different initial eigenvalues. However, if only the natural frequency near the operation frequency is concerned for a steady running system, such as a turbine generator system, this method-ology is sufficient and adequate to analyze the torsional vibra-tion of the system.

For the first and the second undamped systems, the imaginary parts of the eigenvalues are zero, and no phase angle exists. The calculated results of the first and the second systems agree very well with the results obtained by using both the Holzer method and the MATLAB software program. The calculated results of the third and fourth damped systems show very good agreement with those obtained by Doughty and Vafaee [6] and Doughty [14]. Therefore, the methodology can be used for not only undamped systems but also damped systems subjected to external torques. It can also be used to analyze the torsional vibration for both steady systems and dynamics systems. The application range of the extended transfer matrix method with the Newton-Raphson technique is improved.

Acknowledgments

The authors would thank the National Science Council of the Republic of China for the Grant NSC80-0401-E002-21 to

complete this research project and to Grant D. Huang for com-ments and revisions made on this manuscript.

References

[ 1 ] Holzer, H„ Die Berechnung der Drehschjwingungen, Springer, Berlin, 1922.

[2] Den Hartog, J. P., and Li, J. P., "Farced Torsional Vibrations with Damping; An Extension of Holzer's Method," ASME Journal of Applied Me-chanics, Dec. 1964, pp. 276-280.

[3] Pestel, B.C. and Leckie, F. A., Matrix Methods in Elasto Mechanics, McGraw-Hill, New York, 1963.

[4] Eshleman, R. L., "Torsional Response of Internal Combustion Engines," k^yK, Journal of Engineering for Industry, Vol. 96, May 1974, pp. 4 4 1 -449.

[51 Sankar, S., "On The Torsional Vibration of Branches System Using Extended Transfer Matrix Method," ASME Journal of Engineering for Industry, Series B, Vol. 101, Oct. 1979, pp. 546-553.

[6] Daughty, S., and Vafaee, G., "Transfer Matrix Eigensolutions for Damped Torsional System,'' ASME JOURNAL OF VIBRATION, ACOUSTICS, STRESS,

AND RELIABILITY IN DESIGN, Vol. 107, Jan. 1985, pp. 128-132.

[7] Dawson, B., and Davies, M., "An Improved Transfer Matrix Proce-dure," International Journal for Numerical Methods in Engineering, Vol. 8, 1974, pp. 111-117.

[8] Doughty, S., " A Rayleigh-Type Inclusion of Shaft Inertia in Torsional Vibration Analysis," ASME Journal of Engineering for Gas Turbines and Power, Vol. 116, Oct. 1994, pp. 831-837.

[9] Neilson, R. D., and Barr, A. D. S., "Imbalance Effects in Rotor Systems under High Angular Accelerations," ASME Proceedings of International Gas Turbine and Aeroengine Congress and Exposition, paper 21091, 1994, pp. 1-10.

[ 10] Yun, Jong-Seop, and Lee, Chong-Won, "Dynamic Analysis of Flexible Rotors Subjected to Torque And Force," ASME Proceedings of the 14th Biennial ASME Design Technical Conference on Mechanical Vibration and Noise, DE v60, 1993, pp. 331-338.

[11] Lu, Songyuan,' 'Transfer Matrix Method for Nonlinear Vibration Analy-,sis of Rotor-bearing System," ASME 13th Biennial Conference on Mechanical Vibration and Noise, DE, Vol, 37, 1991, pp. 7 5 - 8 3 .

[ 12] Kang, Yuan, and Chung, Yung Being, "Transfer Matrix Method for the General Periodic Torsional Vibration of Rotor Systems,'' ASME Proceedings of the 2nd Biennial European Joint Conference on Engineering Systems Design and Analysis, Part 7 (of 8) PD, Vol. 64, No. 7, 1994, pp. 195-202.

[13] Huang, Y. M., "Performance of the Damper in a Power Train," ASME paper No. 84-DET-160, 1982.

[14] Doughty, S., "Steady-State Torsional Responnse with Viscous

Damp-ing," ASME JOURNAL OF VIBRATION, ACOUSTICS, STRESS, AND RELIABILITY IN

數據

Fig. 2 Schiematic diagrams of a disl&lt; and a sliaft
Table 1 Mass moments of inertia of disi&lt;s and stiffnesses of shafts for  the first undamped system
Table 7 Results of the eigensolutions for the fourth damped system

參考文獻

相關文件

The disadvantage of the inversion methods of that type, the encountered dependence of discretization and truncation error on the free parameters, is removed by

Reading Task 6: Genre Structure and Language Features. • Now let’s look at how language features (e.g. sentence patterns) are connected to the structure

In this chapter we develop the Lanczos method, a technique that is applicable to large sparse, symmetric eigenproblems.. The method involves tridiagonalizing the given

We have also discussed the quadratic Jacobi–Davidson method combined with a nonequivalence deflation technique for slightly damped gyroscopic systems based on a computation of

/** Class invariant: A Person always has a date of birth, and if the Person has a date of death, then the date of death is equal to or later than the date of birth. To be

ˆ If the dual CD method reaches the iteration limit and primal Newton is called, we see that the number of Newton iterations needed is generally smaller than if the Newton method

synchronized: binds operations altogether (with respect to a lock) synchronized method: the lock is the class (for static method) or the object (for non-static method). usually used

This reduced dual problem may be solved by a conditional gradient method and (accelerated) gradient-projection methods, with each projection involving an SVD of an r × m matrix..