Compwers & Sm~fures Vol. 59, No. 4, pp.

613-621, 1996
Copyright 0 1996 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
0045-7949/96 sl5.00 + 003

EXACT DYNAMIC STIFFNESS MATRIX OF A

BENDING-TORSION COUPLED BEAM
INCLUDING WARPING

J. R. Banerjeet, S. Guot and W. P. HowsonS

TDepartment of Mechanical Engineering and Aeronautics, City University, Northampton Square,
London EClV OHB, U.K.
fSchoo1 of Engineering, University of Wales College of Cardiff, Newport Road, Cardiff CF2 IXH, U.K.

(Received 14 October 1994)

Abstract-It is known that an allowance for warping stiffness can change the natural frequencies of
thin-walled open section beams substantially. The purpose of this paper is to investigate the magnitude
of such changes by using an exact member theory. This is achieved by formulating an exact dynamic
stiffness matrix for a typical beam member from established theory and linking this to a new and
convenient procedure which extends the well-known Wittrick-Williams algorithm to ensure convergence
upon any desired natural frequency. Numerical results are given for both single and continuous beams
of the channel section for which some comparative results are available in the literature. The effect of
warping stiffness on the natural frequencies is discussed and it is concluded that substantial error can be
incurred if the effect is ignored.

I. INTRODUCTION natural frequencies, by tracking the zeros of the

determinant of the structure stiffness matrix, is prone
Coupled bending-torsional vibration of beams has
to missing roots when the element formulation leads
been investigated in recent years by (amongst others)
to a transcendental eigenvalue problem, as in the
Hallauer and Liu [I], Friberg [2] and Banerjee [3,4].
present case.
These investigations have shown that the coupling
Natural frequencies are given for single and con-
between the bending and the torsional displacements
tinuous beams which have a thin-walled open cross-
in the free vibratory modes of a beam occurs when
section. The effect of warping on natural frequencies
the mass centre (centroid) and the shear centre (centre
is discussed and conclusions are drawn.
of twist) of the beam cross-section are non-coinci-
dent. However, the torsion theory used by these
authors assumes that the beam twists according to the 2. DYNAMIC MEMBER STIFFNESS MATRIX
Saint-Venant theory and thus no allowance is made
for warping of the beam cross-section. Such an A straight uniform beam member of length L and
assumption can lead to large errors when calculating of channel section is shown in Fig. 1, with the mass
natural frequencies, particularly when considering axis and the elastic axis (i.e. the loci of the mass centre
beams of a thin-walled open section. Bishop et al. [5] and the shear centre of the cross section) being
demonstrated this convincingly when they extended separated by x,. In the right handed coordinate
the governing differential equations of a bending- system of Fig. 1, the elastic axis, which is assumed to
torsion coupled beam to include warping and then coincide with the Y-axis, is permitted flexural trans-
solved them to obtain comparative results for natural lation h(_y, t) in the Z-direction and torsional ro-
frequencies of various single members. tation I(Ib, t) about the Y-axis, where y and t denote
The work presented in this paper extends this distance from the origin and time, respectively. The
approach by recasting the equations in the form of a governing differential equations of motion of the
dynamic member stiffness matrix. This enables the coupled bending-torsional vibration of the beam with
usefulness of the finite element technique to be com- the inclusion of warping stiffness are given by [5]
bined with the accuracy of exact member theory to
solve both single members and assemblies of mem-
bers in a well established way. A new procedure is
then presented, based on the Wittrick-Williams al-
gorithm [6,7], for converging with certainty upon any
required natural frequency. Such a procedure is very where E1, GJ and ET are, respectively, the bending,
useful, since the conventional method of determining torsionalandwarpingrigidityofthebeamcross-section;

613
614 J. R. Banerjee et al.

The solution of the differential eqn (6) can be

obtained by substituting the trial solution W = eSc to
give the characteristic equation

Let

7 =s2. (12)

Substituting eqn (12) into eqn (11) gives

Fig. 1. Coordinate system and notation for a thin-walled

open-section beam. i,74-73-(~W~B+~~)72+1287+C&JT=0. (13)

It can be shown [5] that all four roots of eqn (13)

m is the mass per unit length; Z, is the polar mass
are real, two of them negative and two positive,
moment of inertia per unit length about the Y-axis
and they lie within four distinct intervals, namely
(i.e. an axis through the shear centre) and primes and
(--co, -J&), (-,/&t-9, (O,,/&) and (J&,
dots denote differentiation with respect to position y
+ co). A simple computer program can be written to
and time t, respectively.
find these roots by an iterative method to any desired
If a sinusoidal variation of h and $, with circular
accuracy. Let these four roots be 7,) 72, - 7) and - 74
frequency w, is assumed, then
where 7, (j = 1,2, 3,4) are real and positive. Then the
eight roots of the characteristic eqn (11) are
h (y, t) = Hb)sin it and $(JJ, t) = Y b)sin wt,
(3) a, -a, 8, -p, iy, -iy, is, -is, (14)

where H(JJ) and Y(r) are the amplitudes of the where i =J(-1) and
sinusoidally varying vertical displacement and tor-
sional rotation, respectively. GL=Jz,, p=Jz,; y47,; s=J74. (15)
Substituting eqn (3) into eqns (1) and (2) gives
It follows that the solution of eqn (6) is of the
EIH”” - mo2H + mx,w’Y = 0 (4) form I51

ETY”‘” - GJY” - I,w’Y + mx,o’H = 0. (5) W(r) = C, cash u< + C, sinh cc< + C, cash /I<

+ C, sinh /I< + C, cos rc + C, sin yt

Equations (4) and (5) can be combined into one
equation, by eliminating either H or Y, to give the
+ C, cos St + C8 sin St. (16)
eighth-order differential equation

Equation (16) represents the solution for both

the bending displacement H(t) and the torsional
rotation Y (5). Thus
+i,D2+cl,i,}W=0, (6)
H(5) = A, cash u< + A, sinh cc< + A, cash p<
where
+ A, sinh pt + A, cos yt + A, sin ~5
W=Hor Y (7)
+ A, cos St + A, sin St. (17)
D = d/d<
(8) Y(5)=A:coshcr~+A~sinhcc~+A~cosh/I<
r =ylL

and +A:sinhP< +Afcosyt +A,*sinyt

fA~cos6~ +ABsin6<, (18)

1, = mw2L4/EI; i, = I,02L2/GJ;
where A,-A, and A :-AZ are two different sets of
1, = El-/GJL2 (9) constants.
The constants A, and A: (j = 1,2,3, . . . ,8) can be
c = 1 - mx:jI,. (10) related through eqns (4) and (5) as
 Exact dynamic stiffness matrix of a bending-torsion coupled beam 615

Fig. 2. Sign convention for positive shear force (S), bending motion (M), torque (T) and bi-moment (II).

A: = klx,)A,; A: = (k,lxJA,; +A4p3cosh/?t +A,y3sinyr-A,y3cosy<

A: = (k,Jx.M; A$ = (k&,)4; +A,?j3sin65 -A,G3cos65} (23)

A: = (k,lx,)A,; A: = (k,lx,)A,;
T(5) = (GJ/~)W(U - L’Z”“(5))
A: = (kdx,)A,; A,* = (k,lx,)A, (19)
= (GJ/x,L){A,e, sinh a< + A,e, cash u<
where

k, = (I, - a4)/1,; k, = (1, - /?‘)/A,; + A3eS sinh /I< + A4es cash j<

k, = (I, - y”)/& ; k, = (1, - S’)/&. (20) - A,e, sin y< + A6e, cos y< - A,e, sin SC

Following the sign convention of Fig. 2, the ex-

pressions for the bending rotation O(t), the bending + Aged cos SC} (24)
moment M(t), the shear force S(r), the torque T(r)
and the bi-moment B(c) can be obtained from eqns
B(c) = (W/L2.)‘f”‘(~) = (ET/x,L2){A,a2k, cash at
(17)-(20) as follows:

S(c) = H’(<)/L = (l/L){A,c( sinhcrt + A,a cash a<

+A,B sinh /?t + A4B cash B{ - A,y sin yt

+ A4fi2kp sinh 85 - A,y2k, cos yc

+Asy cosyr -A,6 sin65 + A,6 cosd<}

-A,y’k,sinyt -A,d2k,cos6<
(21)

M(c) = -(EZ/L2)H”(l) = -(EZ/L2){A,a2 cash c$ -A,a2k, sin 6<} (25)

+A2a2 sinh err + A,p2 cash /3<

where a prime, now denotes differentiation with
respect to 5 and
+ A4B2 sinh /?5 - A,y2 cos y5

-Aby sin yt - A,a2 cos a< - A,6* sin St} e, = ~(1 - l,a2)k,; eS = j?(l - ,I,/?2)ka
(26)
e,=y(l +&y*)k,; e,=6(1 +1,6’)k, I
(22)
S(r) = -(l/L)M’(r) = (EZ/L3)(A,a3 sinh crt
The end conditions for displacements and forces of
the beam (see Fig. 3) are, respectively,

Atendl(t=O), H=H,, @=@,, Y=Y, and yj=y;

(27)
Atend2((=1), H=H,, O=@,, y=y’, and y’=y;

At end 1 (5 = 0), S = S,, M=M,,T= -T, and B= -B,

(28)
Atend2(C=l),S=-S2, M=-M,, T=T, and B=B,

CAS
59,4-c
616 J. R. Banerjee et al.

Fig. 3. End condition for forces and displacements of the beam.

Substituting eqn (27) into eqns (17)-(21) gives

1 0 1 0 1 0 1
u B Y
0 0 0 0
z L L
k k
1 0 ks 0 -I 0 -k,
x, X, XZ %

0 -$ 0 pJ 0 yk, 0
01 J&L x,L
ci,. sh, ch,3 Sh,4 c, s, Cb
QShrr
--
aCh, BshS
__-
,-%I, -rs, rc, - 6S6
L L L L L L L
k,
- Ga k,
- Six k,
-- CM k, ‘hS kycy k, S, ka Cd
xz X, -% XZ TX x, X,
ak,
_
sh, ak,
-
chm Bk,
__
sh,9 b%
-
ch, - yk, S, yk, - 6k, S,
x,L &L x,L %L x*L x,L x,L

X A
(29)
4
-46

A7

i.e.

U=BA, (30)

where

C,=cosha; C,,,=coshfi; C,=cosy; C,=cos6;

(31)
S,,=sinha; ShS=sinhb; S,=siny; S,=sind; >
 Exact dynamic stiffness matrix of a bending-torsion coupled beam 617

Substituting eqns (28) into eqns (22)-(25) gives

S, 0 Pa3 0 0 -pv3 0 -P63

W -Qu2 0 -Qb2 QY2 0 QS2 0
T, 0 - RR 0 0 - RP, 0 - RA
4 - vvvv, 0 - vv, vv, 0 Vvs 0
s, = - Pa 3Shz - Pci 3Chn - W3Sh&4 -pY3S, pY3C, -Pm, PST,
M2 Qa ‘GE Qa 2&,a Qb 2GS -Qr'c, -Qr'S, -QS2C, -QS2S,
T2 RP,S,, J+c,Gz RP~4,s - RP,S, RKc, - RF, S, RA C,
B2 _ vv,G, vv, &I, vv,CtIs - vv,c, - vv,s, - vv, c, - vv, s,

A,
A2

A3

A4
X (32)
‘45
A6

A7

_Aa _

i.e.

F=DA, (33)

where

(34)

and

P = EI/L3; Q = EI/L2; R = GJIL; V = ET/L’. (35)

Equations (30) and (33) give

F=KU (36)

i.e.

S, 4.1 K.2 K1.3 Kc Ku KM Ku K.6

M K2.2 K2,3 K2.4 K2.5 K2,6 K2,7 K2.s

T, K3.3 K3.4 4.5 K3,6 K3.7 K3.8

4 K 4,4 &,s K4,6 K4.7 K4.8

(37)
s, = Symmetric KS &,a KSJ &a
M2 x6.6 &,T &.a
T2 KT.T &,a
_B2 _ K a,8
618 J. R. Banerjee et al.

corresponding to IK*I= 0. However, when K* is

developed from exact member theory the determinant
is a highly irregular, transcendental function of w.
d Additionally, several natural frequencies may be
close together or coincident, while others may excep-
tionally correspond to D* = 0. Thus any trial and
--1 error method which involves computing II<*1 and
noting when it changes sign through zero, can miss
roots and hence, natural frequencies. This danger can
Fig. 4. Cross-section and property details of the channel
be overcome by use of the Wittrick-Williams algor-
section beam of the example 1 (d = 0.058 m, b = 0.1 m, ithm [6] which has received wide attention in the
f = 0.00125 m, L = 1.28 m, EI = 0.974 x IO5 N-m’, literature [7]. The algorithm states that
GJ = 11.21 N-m*, ET = 35.4 N-m4, M = 2.095 kgm-‘,
I, = 0.725 x lo-’ kg m, X, = 0.03771 m).
J = J, + s{K*}, (40)
where
where J is the number of natural frequencies of the
K=DB-' (38) structure
exceeded by some trial frequency, CO*, JO is
the number of natural frequencies which would still
is the required dynamic member stiffness matrix be exceeded if all the members were clamped at their
ends so as to make D* = 0, and s{K*} is the sign
count of the matrix K*. s{K*} is defined in Ref. [6],
3. WITTRICK-WILLIAMS ALGORITHM
and is equal to the number of negative elements on
The dynamic structure stiffness matrix, K*, when the leading diagonal of the upper triangular matrix
assembled from the member stiffness matrices, yields obtained from K*, when w = CO*, by the standard
the required natural frequencies as solutions of the form of Gauss elimination without row interchanges.
equation The knowledge of J corresponding to any trial
frequency makes it possible to develop a method for
K*D*=O, (39) converging upon any required natural frequency to
any desired accuracy. However, while s {K*} is easily
where D* is the vector of amplitudes of the harmon- computed, the value of JO is some times more difficult
ically varying nodal displacements and K* is a func- to determine as in the present case. A procedure for
tion of CO,the circular frequency. In most cases the the calculation of JO i.e. the sum total of the
required natural frequencies correspond to 1K* 1,the number of clamped<lamped natural frequencies of
determinant of K*,being equal to zero. Traditionally each member in the structure which lie below the
the required values have been ascertained by simply trial frequency, CO*, is described below. (Note that JO
tracking the value of 1K* 1and noting the value of w can also be interpreted as the number of natural

Table 1. Dynamic stiffness coefficients of equation (37) for the beam of

Fig. 4 at two frequencies with the effect of warping included

Stiffness Numerical values

coefficients 80.0 (rad SK’ I 300.0 (rad s -’ )

K,,, = KS,,= 550,941.3903 465,23 18658

4.2 = -Ku = 355.535.7197 339,847.2948
k, = Ku = 241.2456804 3578.736035
6, = -Ks,B = 43.30717418 658.6952940
_ 559,544.0824 - 590,59 1.3582
4.5
4, = -45 = 351,372.2151 366,876.7301
K,,, = 4, = 84.089075 18 1352.277424
6.a = -45 = -25.60944021 -407.1925764
K2,2 = Ke.6 = 304.106.3761 300.429.4504
K,,, = -G = 43.62973036 663.6993907
K,,= K,.= 10.07846293 155.2425938
K; = “‘” 152,389.1463 155,185.8978
K2;, = - K,,6 = 25.88107357 411.4713151
Ku = 46 = - 7.562383541 - 119.4356184
4.3 = K,., = 190.9340106 -- 114.2460529
Kl.4= - K,.a= 126.7833125 70.59598752
4.7 - 220.7807082 - 336.2820117
K,,, = - &,, = 133.1052844 167.8715483
Ku = Ks.8= 111.6086658 98.47638602
K4.8 55.52853213 65.62121120
 Exact dynamic stiffness matrix of a bending-torsion coupled beam 619

Table 2. Natural frequencies of the channel section beam (see Fig. 4) with Free-Free (F-F), Clamped-Free
(C-F), Simply-Supported (S-S) and Clamped-Clamped (C-C) end conditions

Natural frequency (Hz)

Support Frequency Error
condition number Warping included Warping ignored (%)

F-F 1 22.04 19.90 11%

2 152.08 30.69 396%
3 412.23 46.28 791%

C-F 1 25.37 7.66 231%

2 98.57 22.99 329%
3 149.40 38.38 289%

s-s 1 67.13 15.34 338%

2 263.67 30.71 759%
3 275.80 46.07 499%

c-c 1 149.39 15.36 873%

2 410.56 30.71 1237%
3 624.60 46.08 1255%

frequencies of the structure corresponding to D* = 0, boundary conditions (other than the clamped-
which lie below the trial frequency w*.) clamped one) which admits a simple solution from
which the solution for the clamped-clamped case can
Calculation of JO be deduced. This is most easily achieved by imposing
From the definition of J,, it can be seen that [6,7] pin-pin supports at the ends of the member which
prevent bending displacements and torsional ro-
tations but allow the cross-section of the member to
Jo=cJm> (41)
warp.
If the stiffness matrix for this single member struc-
where J,,, is the number of natural frequencies of a
ture is K+, the number of root exceeded by w* is
component member (element), with its ends clamped,
given by eqn (40). (Note that K+ for the member with
which has been exceeded by o*, and the summation
pin-pin support is a 4 x 4 matrix which can be
extends over all component members.
obtained by deleting the appropriate rows and
In some cases it is possible to determine the value
columns of the dynamic stiffness matrix of eqn (37).)
of J,,, for a structural member symbolically, using a
Thus following the arguments given above
direct approach [8] which gives an analytical ex-
pression for J,. However, this is impractical in the
present case due to the algebraic complexity of the Jpp = J,,, + s{K+}, (42)
expressions. Instead, the same result (i.e. the calcu-
lation of J,,,) is achieved by an argument based on eqn where Jpp is the number of natural frequencies of the
(40) to apply the Wittrick-Williams algorithm [6] in member with pin-pin supports, which lie below the
reverse. The procedure is similar to the one originally trial frequency w*.
put forward by Howson and Williams [9] and is It then follows directly from eqn (42) that
described as follows.
Consider a component member which has been
isolated from its neighbours by clamping its ends. J,,, = Jpp --s{K+}, (43)
Treating this member as a complete structure, it is
evident that the required value of J, could be evalu- where s{K+} is readily obtained by performing
ated if its natural frequencies were known. Unfortu- Gauss elimination on K+ and counting the negative
nately, this simple structure can rarely be solved terms on the leading diagonal of the triangulated
easily. We therefore seek to establish a different set of matrix. Evaluation of Jpp is rather more difficult, but

A B C D

Fig. 5. Cross-section and property details of the channel section beam of the example 2 (d = 0.0889 m,
b = O.l524m, r =0.0071 m, EI =0.1704 x lO’N-m’, GJ =0.314 x 104N-m*, ET =0.1337 x 104N-m4,
m = 17.61 kg m-l, I, = 0.1342 kg m, X, = 0.05626 m).
620 J. R. Banerjee et al.

fortunately it relates to the boundary conditions where o, _ and w, + are given by using the - and +
which yield a simple exact solution, as explained sign of eqn (50). Clearly wi_ < wi+ and, both w,_
below. and wi+ increase monotonically with i. Therefore, Jpp
For the pin-pin support, the boundary conditions can be calculated from
are defined for y = 0 and y = L (i.e. r = 0 and < = 1)
as Jpp = i_ + i, (51)

H =d2Hldr2=0 and Y =d2Yyldr2=0. (44) where im (i,) is the highest value of i for which
IX_ (wi+ ) lies below w*. Once Jpp is known from eqn
(51), J, can be calculated from eqn (43).
These conditions are satisfied by assuming sol-
utions for bending displacement H(r) and torsional
rotation Y(l) of the form [lo] 4. APPLICATION OF THE DYNAMIC STIFFNESS MATRIX
AND RESULTS

H = Ci sin(in{) and Y = Di sin(inc) Two examples are given to demonstrate the effect
of warping on the natural frequencies of a single
(i = 1,2, 3,. . co), (45) beam and an assembly of beams. In both cases the
cross-section is that of a channel section.
where C, and Di are constants. The first example considers a single beam whose
Having established the physical principles, the sol- cross-section and other details are shown in Fig. 4.
ution could proceed by substituting eqns (45) and This example was selected because comparative re-
their derivatives into eqns (4) and (5). However, it is sults are available in the literature [5]. The results are
more convenient to work with eqn (6) in which one presented in two parts. Firstly, the 20 independent
of the variables has been eliminated. Thus making the dynamic stiffness coefficients of eqn (37) for the beam
equivalent assumption that are calculated using double precision arithmetic in
FORTRAN at frequency values of 80 and
W = CT sin(i7rl;) 300 rad s-r, respectively, and listed in Table 1 as a
(46)
checking aid to those who may wish to code the
theory. Secondly, a set of results comprising the first
and substituting this into eqn (6), and after simplifica-
three natural frequencies of the beam for various
tion, we obtain
boundary conditions are given in Table 2, both with
and without, an allowance for warping. The theory of
(bf-W2)(t:-c?)-(l -c)w4=0, (47) Ref. [3] was used to obtain the results for the case
when warping was ignored. It is clear from the results
where in Table 2 that warping has a profound influence on
the natural frequencies of the beam. Indeed, the
bf = i4n4/1,* and tt = i2n2(l + i2n2A,)/li (48) errors are unacceptably large for all the four bound-
ary conditions investigated. The first three (elastic)
with natural frequencies of the beam with effect of warping
included, were obtained for the free-free (F-F) case
1; = mL4/EI(=I,/w2) by Bishop et al. [5] as 21.64, 149.57 and 405.43 Hz,
respectively. These frequencies agree quite well with
the ones shown in Table 2, the discrepancies being
and A,* = ZmL2/GJ( =I,/02). (49)
less then 2%. Table 2 also shows that the error as a
result of ignoring the warping stiffness for the
Equation (47) is a quadratic equation in w2 and clamped-clamped (C-C) case is much greater than
yields two positive values of w for each value of i as the other cases, as expected.
follows [ 111: The second example examines a channel section
continuous beam as described in Fig. 5. For conven-
Oi- Ywi+ = {(bf + rf)/2c}“2 ience the beam is assumed to be supported through
its shear centre and warping is allowed at B, C and
x [1 T (1 - 4cb3tj/(bf + tf)2}“2]“2, (50) D but is fully constrained at A. The first three natural

Table 3. Natural frequencies for the structure defined in Fig. 5

Natural frequency (Hz)
Frequency Error
number Warping included Warping ignored (%)
1 6.127 5.462 12.2%
2 19.34 16.34 18.3%
3 32.71 26.14 25.1%
 Exact dynamic stiffness matrix of a bending-torsion coupled beam 621

frequencies with and without an allowance for warp- exact vibration modes. J. Sound Vibr. 85, 105-113
ing are shown in Table 3. The percentage errors for (1982).
P. 0. Friberg, Coupled vibration of beams-an exact
the three frequencies when the warping effect is dynamic element stiffness matrix. Int. 1. numer. Meth.
ignored, are 12.2, 18.3 and 25.1%, respectively. Once Engng 19, 479-493 (1983).
again the necessity to allow for warping in members J. R. Banerjee, Coupled bending-torsional dynamic
which are prone to this form of deformation is quite stiffness matrix for beam elements. Inr. J. numer. Meth.
Engng 28, 1283-1289 (1989).
clear. J. R. Baneriee. A FORTRAN routine for commutation
of coupled-bending-torsional dynamic stiffness matrix
of beam elements. Ado. Engng Software 13, 17-24
5. CONCLUSIONS (1991).
5. R. E. D. Bishop, S. M. Cannon and S. Miao, On
An exact dynamic stiffness matrix has been devel- coupled bending and torsional vibration of uniform
oped for a bending-torsion coupled beam which beams. J. Sound Vibr. 131, 457-464 (1989)
takes into account the effect of warping stiffness. The 6. W. H. Wittrick and F. W. Williams, A general algor-
application of this theory utilizes a new extension of ithm for computing natural frequencies of elastic
structures. Q. J. Mech. appl. Moth. 24, 263-284 (1971).
the Wittrick-Williams algorithm which ensures con- F. W. Williams and W. H. Wittrick, Exact buckling and
vergence upon any required natural frequency. It has frequency calculations surveyed, J. Srrucr. Engng ASCE
been shown that large errors may be incurred in the 109, 169-187 (1983).
calculation of natural frequencies of thin-walled open 8. W. P. Howson, A compact method for computing the
eigenvalues and eigenvectors of plane frames. Adu.
section beams or beam assemblies, when the effect
Ggng Sofrware 1, iSl-190 (1979):
of warping is ignored. The results obtained using 9. W. P. Howson and F. W. Williams. Natural freauencies
the present theory agreed very well with published of frames with axially loaded Timoshenko members,
results. J. Sound Vibr. 26, 503-515 (1973).
10. S. P. Timoshenko, D. H. Young and W. Weaver, Jr,
Vibration Problems in Engineering Wiley, New York
REFERENCES (1974).
11. J. R. Banerjee. and F. W. Williams, Clamped-clamped
1. W. L. Hallauer and R. Y. L. Liu, Beam bending- natural frequencies of a bending-torsion coupled beam.
torsion dynamic stiffness method for calculation of J. Sound Vibr. 176, 301-306 (1994).