Nonlinear Beam Kinematics by

D. A. Danielson
Associate Professor,
Decomposition of the Rotation
Department of Mathematics,
Naval Postgraduate School,
Monterey, CA 93943
Tensor
Mem. ASME A simple matrix expression is obtained for the strain components of a beam in which
the displacements and rotations are large. The only restrictions are on the
magnitudes of the strain and of the local rotation, a newly-identified kinematical
D. H. Hodges1 quantity. The local rotation is defined as the change of orientation of material
Research Scientist and1 elements relative to the change of orientation of the beam reference triad. The vec-
Theoretical Group Leader, tors and tensors in the theory are resolved along orthogonal triads of base vectors
Rotorcraft Dynamics Division, centered along the undeformed and deformed beam reference axes, so Cartesian ten-
Aeroflightdynamics Directorate, sor notation is used. Although a curvilinear coordinate system is natural to the beam
US Army Aviation Research & Technology problem, the complications usually associated with its use are circumvented. Local
Activity (AVSCOM),
rotations appear explicitly in the resulting strain expressions, facilitating the treat-
Ames Research Center,
Moffett Field, CA 94035
ment of beams with both open and closed cross sections in applications of the
theory. The theory is used to obtain the kinematical relations for coupled bending,
torsion, extension, shear deformation, and warping of an initially curved and
twisted beam.

1 Introduction
Beam theory has a long history (see Timoshenko, 1983). An ried out entirely in terms of matrices, however, and conse-
understanding of fundamental aspects of beam theory may be quently it was rather tedious, imprecise, and difficult to
obtained from the book by Wempner (1981). A summary of understand.
recent literature concerned chiefly with nonlinear beam theory Interaction between the authors led to introduction of dyads
is given by Hodges (1987b). In this paper our purpose is to ob- into the analysis and attendant simplifications and im-
tain, by means of the polar decomposition theorem, an ac- provements. The present paper presents a development of the
curate but simple expression for the strain in a beam or rod theory based on Cartesian tensors, together with a comprehen-
undergoing large deflections. By this theorem, the change of sive example. Hodges (1987b) embodies a rectified matrix
configuration for any material element in the beam is decom- derivation in order to accommodate engineers who are un-
posed into a pure strain and a pure rotation. A similar decom- familiar with tensor notation. The present derivation,
position was obtained for thin shells by Simmonds and however, offers far greater insight into the nature of the
Danielson (1970, 1972). kinematical assumptions. It is believed that decomposition of
The concept of decomposition was applied to beams in un- the rotation tensor is new to beam literature and that the
published work by the second author in which the novel idea simplified kinematical relations obtained go beyond others in
of separating the rotation into two parts was introduced—an rigor and generality.
arbitrarily large global rotation associated with the beam
reference triad, and a moderately small local rotation 2 Beam Geometry and the Global Rotation Tensor
associated with warp, shear, and other deformations. The
physically reasonable assumptions of small strain and Let xx denote length along a reference line r within an
moderate local rotation led to a rather simple kinematical ex- undeformed beam. Let xa denote lengths along lines or-
pression that was valid for arbitrarily large deflections and thogonal to the reference line r. (Here and throughout the
rotations of the reference triad. The original analysis was car- paper Greek indices assume values 2 and 3 while Latin indices
assume values 1, 2, and 3.) A particle of the beam is located
from a fixed point in space by the position vector r(x 1 , x2, x3).
The covariant base vectors g,- are tangent to the coordinate
Presently Professor, School of Aerospace Engineering, Georgia Institute of curves:
Technology, Atlanta, GA 30332.
Contributed by the Applied Mechanics Division for presentation at the 1987 dr
Applied Mechanics, Biomechanics, and Fluids Engineering Conference, Cincin- g/(*l,*2.*3)=' (1)
dX;
nati, OH, June 14-17, 1987, of the American Society of Mechanical Engineers.
Discussion of this paper should be addressed to the Editorial Department, Contravariant base vectors can be obtained by standard means
ASME, United Engineering Center, 345 East 47th Street, New York, N.Y. (see Budiansky, 1983, or Simmonds, 1982):
10017, and will be accepted until two months after final publication of the paper
itself in the JOURNAL OF APPLIED MECHANICS. Manuscript received by ASME 1
Applied Mechanics Division, May 3, 1986; final revision, September 2, 1986. gi(xux2,xi) = eyfcg/Xg* (2)
Paper No. 87-APM-17. 2Vg

2 5 8 / V o l . 54, JUNE 1987 Transactions of the ASME

Copyright © 1987 by ASME
Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Rotation from bf to bf is accomplished by pre-dot multiplica-
tion with the rotation tensor CrR.
bf = C * . b f (7)
r
The components Cfj are the direction cosines
cf=bf-b; (8)
Rr rR
The tensors C and C are the transpose and inverse of each
other, so that
CRr.CrR=I (9)

where I is the identity tensor. Note that I can be represented by

a dyadic b,b, in which b, is any orthonormal base vector.

3 Strain and the Local Rotation Tensor

Consider the deformation gradient tensor A defined by
A = G,g' (10)
The polar decomposition theorem states that A can be unique-
ly decomposed into an orthogonal rotation tensor C dotted in-
to a symmetric right stretch tensor U (see Ogden, 1984):
A = C«U (11)
Note that
A.g, = C.U.g, = G, (12)
This implies that g, is transformed by first undergoing a pure
strain to an intermediate state U • g, and then undergoing a
rotation to coincide with G ; . The rotation may be decomposed
into two parts: a large rotation equal to the global rotation
UNDEFORMED C*' and an additional rotation which we call local rotation.
Fig. 1 Beam of thin rectangular cross section Note that the local rotation may not vanish at the reference
line.2
For the purpose of eventually allowing only small local rota-
Here tion, we use the following representation for the local rotation
tensor (see Sedov, 1966):
g = det(g,-.gj) (3)
and eijk are the components of the permutation tensor in a exp(« = I + « + - y + - ^ - + . . . (13)
Cartesian coordinate system. (Repeated indices are always
summed over their range.) The reference cross section at the (Here and hereafter a tilde over a tensor or matrix denotes that
point x* is the surface whose position vector is r(x*, x2, x3). it is skew-symmetric) The antisymmetric tensor <f> is related to
In a similar manner, consider the deformed state configura- the local rotation vector <j> by
tion. The locus of material points along r has now assumed a 0 = <4xl (14)
different curved line denoted by R. The particle which had
position vector rCx^, x2, x3) in the undeformed beam now has The direction of <j> is along the axis of local rotation and the
position vector R(x{, x2, x}), relative to the same fixed point. amplitude of <j> is the angle of local rotation. Various other
The covariant base vectors G,- are tangent to the coordinate finite rotation vectors have been used in the literature, all dif-
curves of the deformed beam: fering from 0 in amplitude only (for example, see Simmonds
and Danielson, 1970, 1972; Reissner, 1973; or Kane et al.,
Gi(xux2,xi) = —— (4) 1983).
oXj We thus have the decomposition of the total rotation into
We will eventually resolve the position vectors along triads local rotation and global rotation:
of unit base vectors on the undeformed and deformed C = exp(£)»C*r (15)
reference lines. At each point along r define an orthogonal
A physical feeling for this may be gained by studying Fig. 1
reference triad bj(*[) tangent to the coordinate curves at r
and by experimenting with a rectangular rubber eraser. We
with bj tangent to r (see Fig. 1). At each point along R in-
picture an initially straight beam of thin rectangular cross sec-
troduce an orthogonal reference triad bf (xx) such that bf (xf)
tion with an arrow embedded in its side. The beam is then
= G*, where G* denotes the part of G„ (x*, 0, 0) arising from
bent, twisted, and stretched. The final orientation of the ar-
the rigid body rotation of the cross section at x*. Note that bf
row may be obtained by two rotations. First the cross section,
= bf x bf is not necessarily tangent to R unless the Euler-
with arrow attached, undergoes a large rigid body rotation to
Bernoulli hypothesis (that the reference cross section remains
bring the base vectors b, into coincidence with bf. Then the
normal to R when the beam is deformed) is adopted.
cross section, with the arrow remaining embedded in its side,
Rotation from b; to bf is accomplished by pre-dot undergoes a small warping to bring it into its final orientation.
multiplication with an orthogonal tensor which we call the The Jaumann-Biot-Cauchy strain tensor r is defined by
global rotation tensor CRr.
bf =C* r .bf = C f b ; (5)
The global rotation tensor can be represented as a linear com- It is possible to formulate the theory so that the local rotation always
bination of the dyads formed from the base vectors: vanishes at the reference line. We could choose bf = C(jf[,0,0) • bf. Hodges
(1987b) defines the global rotation at the reference axis to be equal to the total
C^'-bfbf (6) rotation there. The present analysis results in the simplest algebra.

Journal of Applied Mechanics JUNE 1987, Vol. 54/259

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 r=u-i (16)
A=4,+ *- + -L(y4 + h)+0(4>\4>2() (29)
Other strain tensors based on U could be chosen. With this o l
definition when the strains are small, as in subsequent sec-
tions, the components of r are simply the relative elongations Substituting equations (29) and (24) into (27), we obtain
and shears of material elements lying along the coordinate
curves. Use of equations (9), (11), (15), and the above leads us E = y + ^- + -^-($y-y$)+0(<t>\<t>2e) (30)
to
Finally, we solve equation (30) for the strain
C* r .r.e*=exp(--0).A.C r R -I (17)
y=E-^- + -L(E4>-4>E) + 0(^,^e) (31)
The ensuing formulas will look simpler if we now write the
tensors in component form. Since T is a Lagrangean strain The question now is when can the higher order terms be
tensor, it is appropriate to resolve it along the undeformed neglected? We assume that e can be neglected in comparison
beam reference triad b, yielding with unity. We also assume that 4> = 0(e r ). Two cases are of
r = bj 7( ,b; (18) interest:
(1) Small local rotation: r > 1. A theory based on this
Since the left side of equation (17) then becomes bf yybf, it is assumption would be suitable for solid beams without thin
appropriate to resolve the tensor 4> along the deformed beam cross sections or thick-walled beams with open or closed cross
triad bf yielding sections, where effects of local rotations would be expected to
* = bf*„bf (19) be negligible. The strain for this case reduces to
Comparison of equations (14) and the above leads to resolu- y=E (32)
tion of the local rotation vector <j> along the triad bf yielding (2) Moderate local rotation: 1/2 < r < 1. A theory based
on this assumption would be suitable for thin cross sections
4> = 4>lb?=—±-e(lk4Jkbf (20) such as thin-walled open cross sections, thin strips, rotor
blades, etc., where local rotations could be appreciable. This
It follows from equation (17) that the deformation gradient case yields the rotation from equation (29) and strain from
tensor is resolved along the mixed bases equation (31) as
A = bM v b; (21) <j>=A
Our formula (17) thus becomes in matrix form 4>2 1 - - (33)
y=E-—+—(E<t>-<t>E)
y = exp(-<j>)A — I (22)
The matrices E and A are related to A by the definitions (26).
The components of A follow from equations (10) and (21):
4 Simplifications for Small Strain and Local Rotation ^•=(bf.GJ(g*.b;) (34)
Our expression (22) for the strain is exact but very complex. These components are rather easily obtained because in a
Let e and 4> denote the maximum absolute values of the com- beam theory it is convenient to resolve the base vectors of the
ponents of the matrices y and 4>, respectively: undeformed state in the directions of the bf, and to resolve the
base vectors of the deformed state in the directions of the bf.
max 17,y (x, ,x2 ,x3) I = e < < 1 This will be illustrated by a comprehensive example in the next
max 14>ij (xx ,x2 ,x3) I = 4>< 1 section.
We will retain only terms of the lowest order in e and </>. The
Taylor expansion of exp( - <j>) is easily obtained from equation
(13):

exp(-<£) =7-0+4—l_+0(* 4 ) (24) 5 Application to an Initially Curved and Twisted

Beam
To expand A we first break it up into symmetric and antisym-
metric components: The position vector to points in any undeformed beam may
be written as
A=I+E+A (25)
T(.xl,x2,xi) = f(xi) + xabra (35)
Here we have defined
where f(x,) = r(xlt 0, 0) is the position vector to points on
the reference line r. With this choice of coordinates the
E-^L-I reference cross section is planar. The covariant base vectors
2
(26) are obtained from equation (1) by differentiation of equation
A-AT (35). This may be accomplished using the formulas
A=-
f' = =bf
T (36)
where A denotes the transpose of A. Substituting the above
into equation (22) and solving for E, we obtain (bf)'= = kxbf = k«bf
E=exp$)(I+y)-I-A (27) where primes denote differentiation with respect to xy. Here k
= fc,b, is the curvature vector of the undeformed beam (fcj is
Noting that the left side of equation (27) is symmetric, we can the pretwist of the beam while ka are components of the cur-
obtain an equation for A by equating the right side of equa- vature of the reference line (see Love, 1944)) and k is the cur-
tion (27) with its transpose: vature tensor of the undeformed beam defined by
k = k x I = brikijbrj = - brieu,k,brj (37)
A =-^-[exp(0) (I+y) - (7+7)exp(-</>)] (28)
The contravariant base vectors are obtained from equation
Using equation (24), we now expand the above in powers of <j>: (2). The final result is

260/Vol. 54, JUNE 1987 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 (46)
vg Components of the symmetric matrix E are obtained from
equations (26) and (45):
2
g = - ^ 7 ^ + b2 (38) „ _ in +X3K2- X2K3 +\\l/' +^kl{x3\2- x2\3)
Vg
vg
--y2^ibi
+ b; 2-Yi2-x3Ki+X2'/,Vg:+^'/'(^ + fc
3)
vg En —E-,
where we have obtained from equation (3) the relation 2Vg
(47)
Vg=l-x2k3+x3k2 (39) 2yl3+x2K1+\3\P\[g-\\P(K2 + k2)
E,% —E-,1 —
We assume that the reference cross section does not distort 2Vg
in its plane. 3 Thus, the position vector to points in the deform- £22 -E33- E23 - Ei2 - 0
ed beam can be represented by
R (*, ,x2 ,x3) = r + u + xabR + \\pbR (40) Components of the antisymmetric matrix 0 are obtained from
Here u = u,b£ is the displacement vector of points on the equations (26), (33), and (45):
reference line r, \(x2, x3) is the Saint-Venant warping func- - 2 7 , 2 + X3 Ki + \2Hg ~ M ( «3 + k3)
tion for the local cross section, and ^(*j) is the warp '12=
~ - 0V21
21 = -03=-
amplitude. The b* are chosen tangent to the xa coordinate 2Vg
curves at R. The reference axis is chosen so that X(0, 0) = X 1 A . -2yu-x2Kl+\3\pJg + M(K2 + k2)
Xa (0, 0) = 0 where Xa = d\/dxa. The covariant base vectors 013= -<P31=02= —p (48)
of the deformed beam are obtained from equation (4) by dif- 2Vg
ferentiation of equation (40). This is accomplished using for- 011=022=033=°; 0 2 3 = 0 3 2 = 0 1 = °
mulas analogous to equations (36):
From equation (32) the strain components for the small
R ' = d + 7 n ) b ? + 2 7 l a bS local rotation theory are the elements of the matrix E. The
(41)
( b f ) ' = K x b f = K-bf strain components for the moderate local rotation theory can
now be obtained directly by substitution of E and 0 into the
Here K = Kjbf is the curvature vector of the deformed beam
second of equations (33). Note that the matrix 4> contains
(the components of K are (1 + yn) times the twist and cur-
some terms of the order of strains (i.e., yXa). Care should be
vatures of the deformed beam), and the components Ky =
taken to discard these terms when squared and when
— eijlKl of the curvature tensor K of the deformed beam are in
multiplied by any of the terms of matrix E. Note that it is not
matrix form
necessary, however, to introduce any ad hoc arguments to
K = - (C R r )' CrR + CRrkCrR (42) remove terms of the order of squares and products of the
The strains -y,,- ( x j = 7,,- (xt, 0, 0) at the reference line are ob- strain components. It may be desirable to make some simplify-
tained by evaluation of equations (33)-(34) at the reference ing assumptions about the magnitude of the initial curvatures,
line which affect the order of magnitude of the quantity Vg - 1.
The elements of E have clear physical significance. We see
yn=CR[^\ + CR[(u'i + kijuJ) that En is the extensional strain for the beam; yn is the strain
(43) of the reference line, terms involving K„ are the bending
2yla = CR[ + CR[(u;+kuUj)
strains, and the remaining terms are extensional strains related
We thus have to warping. The off-diagonal terms Ela are shear strains; yla
G t = ( l + 7 H - * 2 * 3 +X3K2 + X ^ ' ) b f is the transverse shear of the reference line and the remaining
terms concern shear strain due to torsion and warp. The direc-
+ (2yl2-x3Kl+\4,K,)bR
(44) tion cosine matrix CRr (and thus K,- ) are taken as given in the
+ {2yx3+x2Kl-\^K2)bR present analysis. They may be expressed in a variety of ways as
discussed by Hodges (1987a).
G„=b*+\,*bf The present strain expressions for the small local rotation
The components Atj of the deformation gradient matrix theory are very similar to those obtained by Wempner (1981).
may now be calculated from equations (34), (38), and (44). The last term in En is missing from Wempner's equations
The result is (8-29). This term was shown by Hodges (1980) to be impor-
tant in correctly predicting the untwist of pretwisted beams
1 + 7ll + * 3 ^ 2 -X2K3 + W + i/k\{X3\2 ~X2\3)
A under an axial tension force. The terms in Ela involving XipKg
Vg are missing from Wempner's final expressions but are prob-
ably not very important anyway. It is unknown if there are any
2yn-x3{Kl-ki) + MK3
-"21 "
further differences in predictive capability between Wemp-
Vg ner's theory and our small local rotation theory. Moreover,
most of the nonlinear terms in equation (33) of the moderate
+ X 2 ( A T , - * , ) - -MK2
A - 2713 (45) local rotation theory are not included in Wempner's book.
•"31 "
Vg~ Perhaps as significant as any differences in the results is the
= \2\p A, 13 = x 3 iA number of approximations invoked in Wempner's arguments
-^12 "
in order to obtain his final result. The quadratic terms not
^22 :=
^33 = =1 ^23 = A
-yi =0 shown in his equations (8-25) are neglected. The gradual
Now let change of the cross section strains is neglected (the underlined
terms of equations (8-25)). Further approximations are made
based on the thinness of the beam. Although the present ap-
This assumption is made strictly for illustrative purposes. The general theory proach involves fewer approximations, the results are simpler.
does not require such an assumption. Nonclassical effects such as transverse
normal strains and distortion shear can be treated by assumption of a more The simplicity of the present analysis, despite the generality
general displacement field. of the example problem, is noteworthy. Indeed, the present

Journal of Applied Mechanics JUNE 1987, Vol. 54/261

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 expressions for strain were obtained with effort comparable to Moffett Field, California (April-June, 1986). The authors
that expended by Hodges (1980) in an analysis involving only thank Howard E. Hinnant of the Rotorcarft Dynamics Divi-
initial twist, extension, and torsion. Like Wempner, Hodges sion, Aeroflightdynamics Directorate, who helped design the
had to invoke several ad hoc approximations in order to figure.
simplify his result.
References
6 Conclusion
Budiansky, B., 1983, "Tensor Analysis," Handbook of Applied
The analysis was based solely on the two assumptions that Mathematics, Pearson, C. E., ed., Van Nostrand Reinhold, Chapter 4.
Hodges, D. H., 1980, "Torsion of Pretwisted Beams Due to Axial Loading,"
the strain components can be neglected compared to unity and ASME JOURNAL OF APPLIED MECHANICS, Vol. 47, pp. 393-397.
that the local rotation components are no larger than the Hodges, D. H., 1987a, "Finite Rotations and Nonlinear Beam Kinematics,"
square root of the strain. If local rotation components are Vertica, Vol. 11, No. 1-2, p. 297-307.
allowed to be as small as the strains, which might be the case Hodges, D. H., 1987b, "Nonlinear Beam Kinematics for Small Strains and
Finite Rotations," Vertica, Vol. 11, No. 3, to be published.
for a beam with a closed or thick cross section, a very simple Kane, T. R., Likins, P. W., and Levinson, D. A., 1983, Spacecraft Dynamics,
theory results. Unlike previous analyses, the removal of McGraw-Hill, Chapter 1.
higher-order terms based on subjective criteria is unnecessary. Love, A. E. H., 1944, A Treatise on the Mathematical Theory of Elasticity,
The theory was applied to a slender, precurved, pretwisted Dover, Chapter 18.
Ogden, R. W., 1984, Non-Linear Elastic Deformations, Ellis Horwood,
beam undergoing bending, torsion, and extension, as well as Chichester, Section 2.2.
shear deformation and warping of arbitrary amplitude, which Reissner, E., 1973, "On One-Dimensional Large-Displacement Finite-Strain
could be important for composite beams. The strain com- Beam Theory," Studies in Applied Mathematics, Vol. 52, No. 2, pp. 87-95.
ponents are explicitly functions of xa and depend implicitly Sedov, L. I., 1966, Foundations of the Non-linear Mechanics of Continua,
upon seven functions of x{ alone: yu, K,, and \j/. These could Pergamon Press, Chapter 1.
Simmonds, J. G., and Danielson, D. A., 1970, "Nonlinear Shell Theory with
be used as the generalized strains of a complete engineering A Finite Rotation Vector," Proceedings, Koninkl. Nederl. Akademie Van Wet-
beam theory. Such a theory will be developed in a later paper. tenschappen, Series B, Vol. 73, No. 5, pp. 460-478.
Simmonds, J. G., and Danielson, D. A., 1972, "Nonlinear Shell Theory with
Acknowledgments Finite Rotation and Stress-Function Vectors," ASME JOURNAL OF APPLIED
MECHANICS, Vol. 39, pp. 1085-1090.
The first author was supported by the Naval Postgraduate Simmonds, J. G., 1982, A Brief on Tensor Analysis, Springer-Verlag,
Chapter 2.
School Foundation Research Program while on leave at the Timoshenko, S. P . , 1983, History of Strength of Materials, Dover.
Aeroflightdynamics Directorate, US Army Aviation Research Wempner, G. A., 1981, Mechanics of Solids with Application to Thin Bodies,
and Technology Activity (AVSCOM), Ames Research Center, Sijthoff and Noordhoff, The Netherlands, Chapters 6, 8, and 10.

IUTAM Symposium on Structural Optimization

Monash University, Melbourne, Australia

February 9-13, 1988

An international symposium sponsored by the International Union of Theoretical and Applied Mechanics
and Australian Academy of Science will be held at Monash University, Melbourne, Australia, from 9th to
13th February, 1988. A special feature of the symposium is layout and shape optimization using analytical
and CAD techniques. Papers will be presented by leading international experts who have been selected by an
international committee chaired by Professors Karihaloo (Australia) and Rozvany (W. Germany). However,
there is scope for inclusion of additional papers on all aspects of structural optimization.

For further information contact:

Professor G. Rozvany
Essen University
Postfach 103764
4300 Essen, W. Germany

Telephone : (49201)1832795/2695
Fax : (49201) 183 2151
Telex : 0857 9091 unied

262/Vol. 54, JUNE 1987 Transactions of the ASME

Downloaded From: http://appliedmechanics.asmedigitalcollection.asme.org/ on 02/18/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use