Compurersd StnrcruresVol. 29, No. 5, pp. 875-889, 1988 lxl45-7949/88 $3.00 + 0.

00
Printed in Great Britain. Pcrgamon Press pk

DYNAMIC ANALYSIS OF FINITELY STRETCHED

AND ROTATED THREE-DIMENSIONAL
SPACE-CURVED BEAMS

M. IURA and S. N. ATLURI

Center for the Advancement of Computational Mechanics, Georgia Institute of Technology,
Mail Code 0356, Atlanta, GA 30332, U.S.A.

(Received 3 September 1987)

Abstract-The problem of transient dynamics of highly flexible three-dimensional space-curved beams,

undergoing large rotations and stretches, is treated. The case of conservative force loading, which may
also lead to configuration-dependent moments on the beam, is considered. Using the three parameters
associated with a conformal rotation vector representation of finite rotations, a well-defined Hamilton
functional is established for the flexible beam undergoing finite rotations and stretches. This is shown to
lead to a symmetric tangent stiffness matrix at all times. In the present total Langrangian description of
motion, the mass-matrix of a finite element depends linearly on the linear accelerations, but nonlinearly
on the rotation parameters and attendant accelerations; the stiffness matrix depends nonlinearly on the
deformation; and an ‘apparent’ damping matrix depends nonlinearly on the rotations and attendant
velocities. A Newmark time-integration scheme is used to integrate the semi-discrete finite element
equations in time. Several examples of transient dynamic response of highly flexible beam-like structures,
including those in free flight, are presented to illustrate the validity of the theoretical methodology
developed in this paper.

1. INTRODUCTION an equilibrium configuration, provided that no dis-

tributed external moments are assumed to exist. Iura
This paper is concerned with the large defo~ation and Atluri [i], on the other hand, have shown that the
dynamic analysis of finitely stretched and rotated use of any three independent components of the finite
three-dimensional elastic space-curved beams, ex- rotation tensor, as rotational variables, leads to a
tending model developed by Iura and Atluri[l] for symmetric tangent stiffness matrix, not only at the
such problems, in the static case. equilibrium but also the nonequilibrium con-
The model used is based on Timoshenko’s hypoth- figuration, even if the distributed external moments
eses; the effects of stretching, bending, torsion and exist in the problem.
transverse shear are taken into account. For sim- The large deformation dynamics of a continuum
plicity, however, the cross-sectional warping effects body have, in the past, been formulated with the use
are neglected. These kinematic assumptions have of the total Lagrangian formulation, the updated
been employed also by Antman and Jordan [2], Lagrangian formulation, the Eulerian formulation,
Reissner [3] and Simo and Vu-Quoc [4] to develop a the Euler-Lagrangian formulation and the moving
thr~-dimensional beam theory. In these references, coordinate formulation [7-lo]. Among these formu-
the existence of prescribed external moments has lations, the inertia effects are readily taken into
been postulated a priori. Iura and Atluri [I] have account in the total Lagrangian formulation. Here,
utilized the variational method to derive the con- therefore, we employ the total Lagrangian formu-
sistent boundary conditions in which the external lation and show the capability of the present formu-
moments arise as a consequence of the applied lation to simulate the dynamic behavior of finitely
external forces. Iura and Atluri [l] have observed that stretched and rotated beams.
the conservative moments (using the definition of In this paper the elasto-static model for finitely
Schweizerhof and Ramm [S] as to whether the load is stretched and rotated space-curved beams, developed
conservative or not) are generally configuration de- by Iura and Atluri [1], is extended to the case of
pendent. Argyris et al. [6] have employed the same dynamic analysis. In Sec. 2 we summarize the kine-
definition for external moments. Using the rotational matic relations of the present model briefly. The
degrees of freedom referred to fixed axes of a global principle of virtual work for elastodynamics is intro-
Cartesian system, Argyris et al. [6] have derived a duced, in Sec. 3, to derive the linear momentum
nonsymmetric tangent stiffness matrix at the element balance (LMB) and angular momentum balance
level. Simo and Vu-Quoc [4] have concluded that, in (AMB) conditions and the associated boundary con-
the context of a classical formulation of rotations, the ditions. Depending on the form of virtual variations
tangent stiffness matrix becomes symmetric only at of the rotational parameters considered, the LMB
876 M. IURAand S. N. ATLURI

and AMB conditions take on different but equivalent Let Y” be a convected orthogonal curvilinear
forms, as in the static problem. A well-defined func- coordinate system. The coordinates Y” are taken in
tional for Hamilton’s principle is obtained by using the cross-section, while the coordinate Y3 is taken
one form of rotation parameters, or the components along the beam axis, as shown in Fig. 1. The un-
of the finite rotation tensor. deformed base vectors at an arbitrary point in a
In Sec. 4, the finite element formulation is utilized cross-section of the beam are given, in terms of the
for deriving the semi-discrete equations of motion. undeformed unit base vectors E, at the beam axis, by
The rotation variables used herein are taken as the
Lagrangian components of conformal rotation A, = E,, (la)
vector [ll]. Without using the four quarternion or
Euler parameters, the singularity, associated with the A, = - Y2K3E, + Y’K,E, +g,,E,, (lb)
finite rotation vector, can be avoided with a simple
manipulation. As shown in the existing literature where
[lo, 121, the resulting mass matrix is no longer con-
stant due to the effects of finite rotations. Even g,= 1 - Y’K2+ Y’K,.
though no external damping effects are accounted for
in the formulation, the nonlinear terms of the velocity The quantities K, are the components of initial
of rotational components appear in the semi-discrete curvature, and the K3 is the initial twist, satisfying the
equations. following relations:
A variety of time integration schemes has been
proposed by many investigators [13]. In this paper, E,,,,=KxE,,,, K=K,,,E,,,, @a, b)
we use the Newmark algorithm to integrate the
resulting semi-discrete equations. Although the stab- where ( ), 3 = d( )/dL where L is the parameter of the
ility and convergence conditions for the nonlinear length of an arc along the line of origin of the
dynamic problems have not been established yet, the coordinate system Y”, in the reference configuration.
Newmark family of algorithms has received wide Let e, be the maps of the base vectors E,,, after a
attention. purely rigid rotation, denoted by the finite rotation
To demonstrate the validity and the applicability tensor R, alone: that is, e, = RE,,,. In general, due to
of the present beam theory, numerical examples are the shear deformation, the unit vector tangential to
presented in Sec. 5. After we confirm the accuracy of the deformed beam-axis does not coincide with the
the beam model developed herein, we investigate the vector e3. From the definition of covariant base
configuration dependency of the external moments in vectors, the deformed unit tangent vector, denoted by
both planar and nonplanar problems. e,, may be written as:

2. PRELIMINARIES ~3=~mI13E,n, (34

2.1. Fundamental hypotheses
~“II3=@;+u”l,)/& (3b)
The fundamental hypotheses used are itemized as
follows: g =J[(u’13)2+(~213)2+(1 +u313)21, (3c)

(1)’ The plane cross-sections of the beam remain where u( =u”‘E,,,) is the displacement vector at the
plane and do not undergo any change of shape during beam axis, S; the Kronecker delta, and ( ) I3 denotes
the deformation. a covariant differentiation by using the metric tensor
(2) The cross-sections are constant along the beam E,= E,.E,.
axis, which remains a smooth space-curve throughout According to the hypothesis (l), the displacement
the deformation. vector at an arbitrary material point is represented as

It should be noted that no simplification is made

U=u+ Y’(e,-E,). (4)
in the present formulation; not only the rotatory
inertia but also the Coriolis and the centrifugal effects
are accounted for. The deformed base vectors at an arbitrary point in a
Throughout this paper, the summation convention cross-section of the beam are given as
is adopted; and the Latin indices will have the range
1, 2 and 3, and the Greek indices the range 1 and 2. a, = e,, (5a)

2.2. The geometry of the undeformed and deformed a3 = (g sin /I, - Y2k3)e,
beam
We summarize, for completeness, the kinematic +(g sinP2+ Y’k,)e,
relations of the present beam model developed by
Iura and Atluri [l]. + (g/I3 - Y’k, + Y2k,)e3 (5b)
 Analysis of three-dimensional space-curved beams 877

Before the /

After the
defor -mation

Fig. 1. Kinematic scheme for highly flexible space-beam analysis.

where tensor, F the deformation gradient tensor, ( )r a

transpose, and dA = d Y’ dY2. The components of T
sin B, = (R.E,).(a” II3&,,), (6a) and M are given by

(6b) T=T”‘e,,,, M=M”e,,,, Pa, b)

(6~) T”= t3mg,,dA, M’= t33Y2g,,dA 6% 4

in which Q is the permutation symbol. The par-

ameters /?, denote the angles of shear deformations.
(94
The vector k, defined by k = k,e,,,, satisfies the
following differential relation:

e,,,=k xe,.
M’ = (t” Y’ - t’l Y2)g0 dA, (90
(7) J
2.3. Strain-stress relationships where tm are the stress tensors defined by
According to Atluri [14], the stress resultants and
moments are defined as tmR= sq’q
1 j .e I* (10)

The (-) in the contravariant tensor components is

T = gOA3.(S, .FT) dA, (84
s used to emphasize that these are not components in
the convected coordinates Ym. The conjugate strain
tensors y,,,xare expressed as
M = Y”e, x [g,A’.(S,.F’)]dA, (8b)
s
= a;e,
Ynu? - A;E,,. (11)
where A” is the reciprocal basis of A,,,,
S, (= ST A, A,) is the second Piola-Kirchhoff stress For one-dimensional beams, we assume that the
878 M. IURA and S. N. ATLURI

following constitutive relationships hold: (EVW), defined as [ 161

t3” = Gy,, , t3’ = &J,~, (1% b) EVW=SP~.6UdV+SP,.~ods,

where G is the shearing modulus and E the Young

modulus. Substituting eqns (11) and (12) into eqns (9) + [j-P/6”
dS,l::L, (16)
and utilizing eqns (5) leads to &

T’ GA, 0 0 0 0 - Cl, h
T2 GA, 0 0 0 Cl2 h2
T’ EA EI, -Er, 0 h,
Ml = (13)
EI,, -EZ,, 0
M2 Sym. EI22 0
M3 GJ

where
where Pb is the vector of body force defined per unit
h, = g sin &, h3 = gB3 - l/u, (1% b) volume of the undeformed beam, denoted as V; PC
the vector of distributed surface traction defined per
unit area of the undeformed cylindrical surface of the
A = g,,dA, A,,= kA, (14~9d) beam, denoted as SC; and P, the vector of distributed
i
surface tractions at the end cross-sections denoted as
I- P S,. The quantity I is the length of the beam axis
I,= Y”g,,dA, Z,,= Y’Y*g,dA, (1% f) before the deformation, and S, is a part of boundary
J J
on which mechanical boundary conditions are pre-
scribed. Since we are concerned with a conservative
II, = ( Y2)2g,, dA, I,, = (Y’)‘g dA, (Ws h) system of forces, the external forces are expressed as
5 5

Pb = PXE,, PC = PiEi, P, = P:E,. (17a-c)

J= pg,dA, p =(Y’)2+(Y2)2, (14ij)
5
Introducing eqns (17) into eqn (16) yields
1; = ki - K, (1W

The factor k is the shear-correction factor [15].

The strain-energy function per unit length along
the beam axis is obtained as

Ws=;GA,(h,)2+fGA,(h2)2+fEA(h3)2 where

+fEI,,(&)2+f EZr,(&)2 +f GJ$d2 q = q’E,, 4 = $E,, (1% b)

+ EZ, h& - E12h3c2 - EI,,&L2 m = m,e, x E,, fi = Rtije, x E,, (19c, d)

- GZ,h,E3 + G12h2E3. (15)

&=IP&dA +~P:IS,,xSlld.r, (19e)
It should be emphasized that the widely accepted
strain-energy function, expressed by eqn (15), is
qj= P!dS,, (19f)
based on the constitutive eqns (12a, b). If we intro- I
duce the well-known constitutive equations such that
S3a= Ge3a and S:3 = Ee where e,,,,, denotes the
mmj= Y’Pjg,dA + Y’P!lS,, x S,ld.s, (19g)
dreen strain tensor, the i&rlting strain-energy func- s s
tion is substantially different from the present one.

2.4. Dejinition of external moments A,= YdP;dS,, (19h)

s
Argyris et al. [6] have defined a conservative
moment as a moment generated by a couple of two in which S is the position vector at the undeformed
equal and opposite conservative forces acting on a cylindrical surface, and ( ),, denotes a differentiation
rigid lever. In this paper, the definition of the external with respect to the coordinate S taken along the
moments are obtained from the external virtual work bounding curve of the cross-sections. The vector
 Analysis of three-dimensional space-curved beams 879

(al problem is written as [16]

P

‘2+
I E
-1
-4 -3
_
a
I. 0

1
I
-- Y3
‘*[6T - IVW + EVWJ dt = 0,
s0

where T is the kinetic energy of the beam and IVW

the internal virtual work, defined as
(20)

(214
Before the deformation

IVW = Sf6cudV, (21W

w
(b) s

E in which p is the density in the reference state and (‘)

a differentiation with respect to time. The subsidiary
E
-2
conditions for eqn (20) are the strain-displacement
-*. relationships, geometrical boundary conditions and
P2 \
the conventional conditions that the variations of
E @b :
-1 displacements at t = t, and t = t2 are equal to zero.
4 At first, following Atluri [14], we introduce a tensor
/p 6R.Rr = (Se x I) as a rotational variation to derive
Y3
% the AMB condition. Then, using eqn (20), we obtain
the LMB and AMB conditions, expressed as
After the deformation
Fig. 2. Configuration dependency of prescribed external r, 3+ q = t, (for abitrary 6u), (22a)

M,, + (X + u),,~ x T + m = fi,

S+ is the rotational variation and defined by (for arbitrary a+), (22b)

SI# x Z = 6R+ Rr [14] where I is an identity tensor.
It follows from eqns (19c, d) that the external where x is an undeformed position vector at the
moments m and iir defined herein are dependent on origin of coordinates Y” and
the deformation. Figures 2(a) and (b) show the
configuration dependency of the external moments. L, = A,i + J,L,, (234
Before the deformation, as shown in Fig. 2(a), the
equivalent load at point 0 (or the origin of the H, = J.e, x li + I,. W, (23b)
coordinates Y”) is the force f( = 1P 1E,) only. After
the deformation, as shown in Fig. 2(b), the equivalent I, = J&:e$ - Jmswp, (234
loads at point 0 are not only the forcef( = 1P )E,) but
also the moment m( = (PIE, x ae,). This example WxI=fi.Rr, (23d)
indicates that the external moments are dependent on
the deformation.
A,, = pg, dA, J. = pY”g, dA, (23e, f)
Argyris et al. [6] have also pointed out the
s s
configuration dependency of external moments. They
have derived a nonsymmetric tangent stiffness matrix
Jma= pY”Yfig,, dA. (23g)
by using the rotational degrees of freedom referred to
s
fixed axis of a global Cartesian system. In this paper,
however, the resulting tangent stiffness matrix is The associated boundary conditions at both the end
always symmetric, as shown later, as long as any cross-sections are obtained as
three independent components of R are employed as
the rotation parameters. The present result agrees T=#, M=l on S,, (24a, b)
with that discussed by Schweizerhof and Ramm [5].
6u = 0, 84 = 0 on S,, (24~ d)

where S, is a part of boundary on which geometrical

3. THE EQUATIONS OF MOTION boundary conditions are prescribed.
Iura and Atluri [1] have introduced three com-
3.1. The principle of virtual work ponents of R denoted by ai, as a set of rotation
The principle of virtual work for the elastodynamic variables. The advantage of the use of ai in dynamic
880 M. IURA and S. N. ATLURI

problems is that a well-defined functional is obtained the so-called ‘static method’ (i.e. using the first prin-
for Hamilton’s principle, as shown later. To obtain ciples of force and moment balance). Noting that the
the AMB and the boundary conditions for ai, we LMB conditions remain unchanged under the ex-
consider, at first, the tensor equation of the AMB change of rotation parameters, the basic equations
condition corresponding to 64. The inner product for 6u and 6a’ take on different forms, but equivalent
between the AMB condition and the variation S4 is to those for 6u and S+.
written as
3.2. Hamilton’s principle
{M,+(x+u),,x T+m-a}.d~ When the potential energy rcPis obtained, Hamil-
ton’s principle for elastodynamic problems is ex-
= C:@R.Rr), (25) pressed as [ 161
where
(32)
C= Q'e,e,
+ Q'e,e,
+ Q3e2e,
+ mqEje, - J$eU- J,sZaeg, (264 where the subsidiary conditions are the geometrical
boundary conditions and the conventional conditions
Qmem=M,,+(x +u),3X T. at t, and t2 cited before. It is not always possible,
however, to construct the potential energy, especially
Since ai are taken as the Lagrange components of R, in a finite rotation beam theory. When the external
(=R,E,E,), 6R = Rjk;,E,EkGa’, where ( );iis a differ- moments defined in eqns (19c, d) are applied on the
entiation with respect to ai. The right-hand side in beam, the use of 4 as a rotational variable makes
eqn (25) is rewritten, in terms of c?, as it difficult to construct the potential energy (Iura
and Atluri [l]). Vu-Quoc [lo] has also indicated that
C:(GR.Rr) = C:(Rzi.Rr)aai. (27) the potential energy does not exist even at the
equilibrium configurations as long as externally dis-
The AMB condition for S+ is represented by tributed moments exist. Note that the variation of
rotational variable used by Vu-Quoc [lo] is the same
c=c= (28) as that used by Atluri [14].
To obtain a well-defined functional, Iura and
while the AMB condition for 6a’ is expressed as Atluri [ 1] have introduced the three components a i of
R as rotational variables. As shown in Sec. 3.1, the
C:(R,,.R’) = 0. (29) resulting equations of motion are equivalent to those
associated with another variable Cp.When using ai as
Since R,;Rr is a skewsymmetric tensor, eqn (29) is rotational variables, the potential energy is ‘obtained
equivalent to eqn (28); the AMB conditions for &$ as PI
and 6a’ are equivalent to each other.
In a similar fashion, the tensor equations for the
nP= [W,(u,cx”)-q.u-m,R,(a”)]dL
boundary conditions are given by s

(M-rii)=(M-1)r 1 - j0” ‘u + 6z~jRjU(am)]~:f,. (33)

(for arbitrary 64) on s

0) Introducing eqn (21a) and eqn (33) into eqn (32),
I
Hamilton’s principle yields the LMB condition in
(M - I):(R: i. Rr) = 0 I eqn (22a), the AMB condition in eqn (29) and the
mechanical boundary condtion in eqn (30b).
(for arbitrary 6a’) J (3% b)

GR.RT = 0 (for 4) 4. FINITE ELEMENT FORMULATION

on S,, (30c,d)
da’= 0 (for ai) > 4.1. Rotation parameters
As discussed in the previous section, the significant
where advantage of the use of ai is that a well-defined
functional is obtained for Hamilton’s principle. In
M = M’e3ez + M*e,e, + M3e2e,, (3W this section, emphasis is placed on the definition of
the present rotation parameters a’ to avoid the singu-
P = i2,Eie,. (3lW larity associated with finite rotation representations.
It is well known that no three-parameter represen-
As seen in eqns (22), the equations of motion for tation of R can be both global and nonsingular [17];
6u and S# are the same as those derived a priori from for this reason the four quaternion or Euler par-
 Analysis of three-dimensional space-curved beams 881

ameters have been introduced to describe the large polated by

overall motions [lo, 18, 191. To avoid using the four
parameters, the conformal rotation vector has been ui = ubNfi (41a)
introduced [1 11. The modification of the Rodrigues
vector leads to the conformal rotation vector defined a’= a;Nfl, (41b)
by
where a; and ai denote the nodal displacement and
rotation components, respectively, and Np are the
@* =4tanTe, (34) shape functions defined by

where e is a unit vector satisfying R.e = e and w a N’=l-L/l,, N2=L/I,, (42a, b)

magnitude of rotation about the axis of rotation
defined by e. where 1, is an element length before the deformation.
Using the conformal rotation vector, we define the For later convenience, the following notations are
present rotation parameters ai such that introduced:

@* = a’E,. (35) d = {u;}, r = {afp}. (43a, b)

Then the Langrangian components of R are ex- Introducing eqns (41) into Hamilton’s principle and
pressed by performing partial integrations with respect to time,
we have
1
___
R”=(4_&)2 [{(aO)’ - akak}6,, &
[A&N”N86u$ + J,~(R,,,,&’
s0
+ 2(a’aj - tbkaOak)], (36)
+ R,,,dik)Ris,,NvGa~ + GA,h,Gh,
where

cr, = (16 - a’a’)/8. (37)

Because of singularity, the Rodrigues vector, defined

by 8 = 2 tan(w/2)e, is valid only in the range of + (EI, h, - EZ,,&)bf;
]w 1-c x. As shown in eqn (34) however, the con-
formal rotation vector is valid even at ]w 1= n. - (Er, h3 + El,,& )S&
Therefore, with this simple manipulation, the finite
rotations are described in terms of the only three - (GI,h, - GI,h,)& - GZ,k;6h,
rotation parameters.
The main idea to avoid the singularity of the + GI,I$h, + (El,& - EZ21;,)6h,
conformal rotation vector is the following [ 111: when
the angle o reaches a value such that -qiNr6u’,-m~jRj~;kNQ6ak,]dL

w=A+C, (38) - [ q’N”Guk + fi,R,;,NvGa!$~$ = 0, W)

.x7
with L being a small positive value, we introduce a where
new rotation parameter 8’ defined by
Sh,= Rijk(6\+uhNP,)Nq6at
8’ = - 16a’/(akak). (39)
+ R,N’% 38uz, Wa)
The corresponding velocity and acceleration are also
S~=~Q~(,(R,~~,R,~~ Nf,Nq+ R,,,i,,R,,NT,
defined by

di = - 16(ai’- aic$/2)/(akak), (4Oa) + R,,,i;.Rks,a~Nf,Nv)Ga~. W)

;i = -16($ - ~‘&,/2 - dic&)/(akak). (40b) Integrating eqn (44) over the beam length and noting
that Sd and Sr are arbitrary, we obtain the following
semi-discrete equations of motion:

4.2. Semi-discrete equations of motion

M(li F, r) + Cc+, r) + KM r) =f(r), (46)
As a standard finite element discretization, the
displacement and rotation components are inter- where M is linear with respect to (w.r.t.) 2 but
882 M. IURAand S. N. ATLURI

nonlinear w.r.t. P and r, C is nonlinear w.r.t. r’ and where DK is defined by

r, K is nonlinear w.r.t. d and r, and f is nonlinear
w.r.t. r. It should be stressed that the vector C is K(dv+ , + Adv, , , r W+, + Ar$J+, )
not derived from the damping effects but from the
nonlinear effects of finite rotations. = K(dv+ , , r W+, )

+DK(d$)+,,r$)+,)
4.3. Time-integration scheme
The Newmark algorithm is employed herein to x {Ad~+,,Ar~+,}‘+O(A*). (51)
integrate the semi-discrete equations of motion in eqn
(46). In a linear problem, this algorithm has received In a similar way, DM, DC and Df are obtained from
a wide attention because of its unconditional a consistent linearization. Note that DM and DC are
stability. nonsymmetric matrices, while DK and Df are sym-
Let ( )N be the value at time t = tN. We postulate metric matrices.
that the solution {d,,,, rn+,} satisfies the semi- The initial values of acceleration and velocity at
discrete eqn (45), i.e. each time step follow from eqns (48):

l- 28 . .
np+, = -+&ON--2p qN9
+K(d,+,,r,+,)=f(r,+,). (47)
cjf)+ , = bN+Af{(l -y)ti,+ynf?+,}. (52b)
According to the Newmark algorithm, the acceler-
ation and velocity at time t = fN+ , are approximated The i + 1 iterative values are also evaluated from
by eqns (48) as follows:

1
oti+
N+I
I) _
-
fi(r) -A(@ N+1 )9 (W
N+’ + /3(At)2

l-2/? ..
-- Ma) WI
2/3 qN,

Ii N+I-- h,V+At((l --Y)fh+Yfh+,}, (48b) These procedures expressed in eqns (52) and (53) are
the same as those of Vu-Quoc[lO].
where 0 : d or r, and A( ) is an incremental value, and The iterations continue until the appropriate con-
fi and y are constants. Substituting eqns (48) into eqn vergence criterion is satisfied.
(47), we obtain the nonlinear algebraic equations in
terms of d,, , and rN+, . 5. NUMERICAL EXAMPLES
To solve the resulting nonlinear algebraic equa-
tions, we utilize the Newton-Raphson method. Then Several numerical examples are considered in this
the 0 and i + 1 iterative solutions are given, by means section to demonstrate the validity and applicability
of the converged solutions at time t = t,, as of the present study. The considered structures con-
sist of straight members. Therefork the origin of the
WW coordinates Y” in each element is so chosen that
Z, = 0, II2 = Z21= 0, J, = 0 and J,2 = J2] = 0.
(49b) All solutions presented in the following have been
obtained using /? = l/4 and y = l/2 in the Newmark
where a superscript in parentheses denotes the iter- algorithms. The tangent stiffness matrix and the
ation number. Substituting eqns (49) into the non- residual forces are integrated by using a uniformly
linear algebraic equations and linearizing them with reduced one-point Gauss quadrature to avoid the
respect to the incremental values, we have the follow- shear locking [20]. The matrices associated with the
ing linear equations with respect to the incremental inertia terms are integrated with two-point Gauss
values: quadrature.
The iteration at each time step is terminated if the
[DM(r$.)+ ,) + DC(rj()+ ,) + DK(djl)+, , r$+ 1) Euclidean norm of residual forces is less than the
prescribed value.
-Df(rW+,)]{Ad!$+,,Ar$$+I}T
5.1. Flexible beam in free Jlight, subject to constant
=f(rv+ ,) -M(&)+ , , Pf+, , r$j+ ,) force and constant moment
Vu-Quoc [lo] has first solved this in-plane problem
- C(,$$+, , rlt)+ ,) - KM?+ ,, rW+A (50) by using a linear shape function. The beam is subject
 Analysis of three-dimensional space-curved beams 883

Material Properties:
EA = CA,. 10,000
El,,= El,, - GJ - 500
Ap-1
J11=J22-‘0

F.E. Mesh: IO linear elements

Time history of F(t) and T(t):

Tltl 1

FItI= Tltl/lo.o
Fig. 3. Flexible beam in free flight, subject to constant force and constant moment. Problem data.

to a force and a torque simultaneously at one end, as of the beam. In such a case, the point may be
shown in Fig. 3. The direction and the magnitude of imagined to be attached to a fictitious wall, which is
the force and the torque are assumed to be constant fixed at the beam axis and moves rigidly with the
during the deformation. In this example we use the beam.
definition for a torque introduced by Vu-Quoc [lo]. As illustrated in Fig. 2, the external conservative
Note, however, that a constant torque is not gener- force causes a configuration dependent moment as
ated by conservative forces as long as the definition the beam deforms. In this example, the magnitude of
for moments introduced herein or by Argyris et al. [6] external moment at the beam axis decreases due to
is employed. Figure 4 shows the present numerical the observed deformation. Consequently, the distinct
results. Good agreement is obtained between the difference in overall motions between the present
present results and the results of Vu-Quoc [lo]. example and that in Sec. 5.1 is observed in Fig. 8.
5.2. Right angle cantilever beam 5.4. Right angle beam in free flight
This out-of-plane problem has been simulated first This out-of-plane problem is solved for the first
by Vu-Quoc [lo] using a quadratic shape function. time in this paper. The material properties and the
Figure 5 shows the material properties and the load load conditions used are shown in Fig. 9. The forces
condition. The dynamic responses are shown in Fig. F, and F2 are applied at the beam axis, while the force
6. Although the present results are obtained with the F3 is applied at a point away from the beam axis. As
use of a linear shape function, an excellent agreement
is obtained between the present results with eight
elements and the results of Vu-Quoc [lo] using 10 t-0.5 t 12.5
elements. The results, obtained using four elements,
are also shown in Fig. 6. The results with four
elements provide a good fit to those with eight
elements.

5.3. Flexible beam in free flight, subject to conservative

force
We consider, once again, the problem discussed in
Sec. 5.1, where the constant force and the constant
moment are applied at the end of beam. As described
earlier in this paper, the external moments generated
by conservative forces are generally dependent on the
deformation. Therefore, we investigate numerically - Present
the ‘configuration dependency’ of external moments.
----Vu - Quoc
In this example, as shown in Fig. 7, only the external
force is applied at one end so that the initial con-
ditions are the same as those of the example in Sec. Fig. 4. Flexible beam in free fight, subject to constant force
5.1. Cases arise in which the point where the conser- and constant moment. Comparison of the present results
vative force is aDDlied does not lie in the cross-section and those of Vu-Ouoc. Time stea Ar = 0.1.
884 M. IURA and S. N. ATLURI

Material Properties:

EA = GA,, = lo6
El,,- El,, - GJ - lo3
Ap-1
I,, - I,, - ‘0

lime history of loading:

Fltl

0 1.0

Fig. 5. Right angle cantilever beam. Problem data.

mentioned in Sec. 5.3,if the point of application of these moments, the projections on the Y’-Y3 plane,
F3does not lie in the cross-section of the beam, it may of deformations of beams, with lower rigidities, are
be imagined to lie on a rigid fictitious wall fixed at the shown in Fig. 11. Even after removal of the con-
beam axis. servative forces, remarkable bending deformations
We analyze the three examples in which the bend- due to the F,,especially for the lowest rigidity, are
ing and the torsional rigidities are altered so that the observed.
behavior of the beam changes from a rigid to a highly As seen in this example, the total angle w at each
flexible body. The overall deformations obtained node exceeds ILrad. Although we do not employ the
using 10 elements are shown in Figs lOa-10c. At four quaternion, the large deformations with finite
time t = 0.0, the transverse force and the torsional rotations can be simulated using the conformal
moment are applied at point A, as shown in Fig. 9. rotation vector or the three rotation parameters.
As the beam deforms, the bending moments, in Table 1 shows the Euclidean norm of residual
addition to the loads described above, are applied at forces in the case of the beam with the lowest
point A. To show the bending deformations due to rigidities at time t = 2.0. In the numerical results

I”.”

-- 4 Elements

--- vu - Quoc

Time
Fig. 6. Right angle cantilever beam. Comparison of the present results and those of Vu-Quot. Time step
Ar = 0.25.
 Analysis of three-dimensional space-curved beams 885

Fig. 7. Flexible beam in free flight, subject to conservative

force. Problem data.
- Example in sect. 5.3
presented in this paper the convergence rate is quad-
ratic. This is consistent with the basic characteristic ----- Example in sect. 5.1
of the Newton-Raphson method.

6. CONCLUSIONS Fig. 8. Flexible beam in free fight, subject to conservative

force. Comparison of the present problem and that in
In nonlinear dynamic analysis of beams, a number Sec. 5.1. Time step At =O.l.
of important problems remain to be resolved. In this
paper, attention has been paid to develop the non-
linear elastodynamic theory of beams and to derive The rotation parameters ai introduced herein lead
the consistent linearized forms of the discrete equa- to a symmetric tangent stiffness matrix and also to
tions. With an emphasis on the definition of the a symmetric load-stiffness matrix. The AMB con-
external moments, we have shown the configuration ditions associated with ui are different, but have
dependency of external moments. In most of the forms equivalent to those derived from the static
structures encountered, the external forces do not act method. In the finite element formulation, the ro-
on the beam axis itself, but on the surface of the tation parameters ai have been defined as Lagrangian
beam. This is the case in which *the present beam components of the conformal rotation vector. As a
theory is particularly applicable. result, as shown in the numerical results, only three
Material Properties-
L
Y’

-I
EA - GA. - lo5
AP-1 9 J,,-Jz2- 10
Case 1: El,, - El,,- GJ - 1000
Fl 10.0 Case 2: El,,- El,,- GJ = 200
Case 3: El,, - El,,- GJ - 100
16.0
F.E. Mesh: IO elements
16.0

14.0 Time history of loadingi

y*/f~;~o,,o.o,
’ 2.0 t F.

5;$kE F

F: = F; - F,/5

Fig. 9. Right angle beam in free flight. Problem data.

Table 1. Euclidean norm of residual forces

Iteration number 0 1 2 3
Euclidean norm of residual 0.21505 x lo4 0.46917 x 10’ 0.20434 x 10’ 0.33963 x 1O-3
 36.0 36.0

t-6.0 4-c

Y3 ,Y3

Y2/
Fig. IOa. Right angle beam in free Right. Sequence of motion in Case I. Fig. lob. Right angle beam in free flight. Sequence of motion in Case 2.
Time step At = 0.1. Time step Ar = 0.1.

.. 7-a _I.c.Y_ _ __r,

-- ,
 Analysis of three-dimensional space-curved beams 887

Y’

t
t - 12.0

t - 11.0 -

t - 7.0

t = 6.0 /

Fig. 1Oc. Right angle beam in free flight. Sequence of motion in Case 3. Time step At = 0.1
 888 M. IURA and S. N. ATLURI

- El=100
t = 6.0
---- El= 200

L- t-o.0
Fig. 11.Right angle beam in free flight. The projections on the Y’-Y’ plane, of deformations of beam,
with lower rigidities.

parameters are enough to describe the finite rotations 3. E. Reissner, On finite deformations of space-curved
with a simple manipulation. The numerical results beams. J. appl. Math. Phy. (ZAMP) 32,734744 (1981).
presented herein show the validity and the applica- 4. J. C. Simo and L. Vu-Quoc, A three-dimensional
finite-strain rod model--II. Compuational aspects.
bility of the present beam theory. Comput. Meth. Appl. Mech. Engng 58, 79-l 16 (1986).
5. K. Schweizerhof and E. Ramm. Disolacement de-
Acknowledgements-The work described herein has been pendent pressure loads in nonlinear *finite element
supported by AFOSR under contract F49620-87-C-0064. analysis. Comput. Struct. 18, 1099-l 114 (1984).
The encouragement of Dr. A. K. Amos is sincerely appre- 6, J. H. Argyris, P. C. Dunne and D. W. Scharpf, On large
ciated. Ms. Cindi Anderson is thanked for her assistance in displacement-small strain analysis of stuctures with
the preparation of this paper. rotational degrees of freedom. Comput. Meth. Appl.
Mech. Engng 14, 401-451 (1978).
REFERENCES 7. K. J. Bathe, E. Ramm and E. L. Wilson, Finite element
formulations for large deformation dynamic analysis.
1. M. Iura and S. N. Atluri, On a consistent theory, and Int. J. Numer. Meth. Engng 9, 353-386 (1975).
variational formulation, of finitely stretched and rotated 8. M. S. Gadala, G. A’E. Oravas and M. A. Dokainish,
3-D space-curved beams. Comput. Mech. (in press). A consistent Eulerian formulation of large deformation
2. S. S. Antman and K. B. Jordan, Qualitative aspects of problems in statics and dynamics. Int. J. Non-linear
the spatial deformation on non-linearly elastic rods. Mech. 18, 21-35 (1983).
Proc. R. Sot. Edinb. 73A(5), 85-105 (1975). 9. H. M. Koh and R. B. Haber, Elastodynamic formu-
 Analysis of three-dimensional space-curved beams 889

lation of the Eulerian-Lagrangian description. J. appl. deformed solids, with application to plates and shells-
Mech. 53, 839-845 (1986). 1. Theory. Comput. Struct. 18,93-116 (1984).
10. L. Vu-Quoc, Dynamics of flexible structures performing 15. G. R. Cowper, The shear coefficient in Timoshenko’s
large overall motions: a geometrically-nonlinear ap- beam theory. J. appl. Mech. 33, 335-340 (1966).
proach. Electronics Research Laboratory Memor- 16. K. Washizu. Varational Methods in Elasticitv and Plas-
andum UCB/ERL M86/36, University of California, ticity, 3rd edn. Pergamon press, New York (1982).
Berkeley (1986). 17. J. Stuelpnagel, Dn the par~et~~tion of the three-
11. M. Geradin and A. Cardona, Kinematics and dynamics dimensional rotation group. SIAM Rev. 6, 422-430
of ridid and flexible mechanisms using finite elements (1964).
and quaternion algebra. Comput. Me& (in press). 18. T. R. Kane, P. W. Likins and D. A. Levinson, Space-
12. A. Rosen, R. G. Loewy and M. B. Mathew, Nonlinear crafi Dvnamics. McGraw-Hill. New York (1983).
dynamics of slender rods. AIAA J. 25,611419 (1987). 19. K.” W.* Spring, Euler parameters and the use of
13. T. Belytschko and T. J. R. Hughes, Computational quatemion algebra in the manipulation of finite
Methodr for Transient Analysis. Elsevier, Amsterdam rotations-A review. Mechanism Machine Theory 21,
(1983). 365-373 (1986).
14. S. N. Atluri, Alternate stress and conjugate strain 20. 0. C. Zienkiewia, The Finite Eiement Method, 3rd edn.
measures, and mixed variational formulations involving McGraw-Hill, New York (1977).
rigid rotations, for computational analyses of finitely

C.A.S. 29+-K