DYNAMIC

STIFFNESS FORMULATION FOR STRUCTURAL

ELEMENTS: A GENERAL APPROACH
J. R. M

Nt

of Me~hankal En@cerin~ and Aeronautics, City University, Northampton Square, London

EClV OHB. U.K.
(Received 28 Seprembrr 1995)

Abrac+A
peacral theory to develop the dynamic stilfness matrix of a structural ekment is outlined.
Substantisl saving in computer time can be achieved if explicit analytical exprcasions for the elements of
the dynamic stiffnc~ matrix are used instead of numerical methods. Such exprarions can be de&d with
the help of symbolic computation. The application of the dynamic stifkss matrix to calculate tbc natural
freqwnaks of a stnbctprc is discus4 with particular reference to the Wittrick-Willkma a@ithm. The
method presented is fairly general. Q 1997 Elsevier Science Ltd. All rights reserved.

1. INTRODUCIION
The dynamic stiffness matrix method to solve free or
forced vibration problems of structures (or structural

elements) has received wide attention in recent years.

The method is often referred to as an exact method
as it is based 00 exact shape functions obtained from
the exact solution of the element differential
equations. The method provides the analyst with
much better model accuracy when compared to finite
element or other approximate methods. This is
because the dynamic stifiness matrix method is
indepemknt of the number of elements used in the
analysis. The usefulness of the method becomes
apparent when higher frequencies and better accuracies of results are required.
The basic building block used in the dynamic
stifiness matrix method is the single frequency
dependent element dynamic stiffness matrix which
aazotmts for both mass and stitTness properties.
Starting from the basic governing dXerential
equations of motion, the author has developed
dynamic stiffness matrices of different types of beam
elements [l-6] and has applied them to free vibration
analysis of space structures [7-f& coupled flexuraltorsional vibration of composite beams [9] as well as
flutter analysis of aimraft wings [10-l 1). The work on
dynamic st&ess formulation for beam elements has
led to developments of some well-known computer
programs such as BUNVIS [12], BUNVIS-RG [ 131
and CALFUN (141. This note gives an overview of
the general procedure the author adopted in the
derivation of the dynamic stiffness matrix of a
structural element.
2. DYNAMICm

FORMULATION

governing differential quations

of motion in free
vibration. This can be accomplished using various
techniques such as by applying Newtons laws,
DAlemberts principle, principk of virtual work,
Lagranges equations or Hamiltons principle which
have all been extensively covered in the literature.
The governing dilIerentia1 equation of motion of a
structural element undergoing free undamped vibration can be symbolically written as
L (II) = 0

(1)

where L is a differential operator and l is the

corresponding displacement vector.
The next step is to solve the above difIerent.ial
equation analytically for harmonically varying n.
(The dynamic stiffness matrix relates harmonically
varying forces to harmonically varying dispIacements
at the nodes of a structural element.) So, the
displacement P is expressed as
tl=IJP

(2)

where U represents amplitudes of displacements, a, is

frequency @d/s), r is time and i = J - 1.
Substitution of qn (2) into qn (1) eliminaoeS the
time dependent terms in the diITeremial equation to
give
LI (u.0) - 0
where LI is a diIferential operator.
The general solution of the di!Terential qn
sought in the fotm

The tirst rtep towards developing the dynamic

stitfness matrix of a strWuml element is to derive its

U=AC
101

(3

(3) is

(4)

102

J. R. Banerjee

where C is a constant vector and A is a frequency

dependent square matrix.
Now, the boundary (end) conditions for displace-

ments and forces are applied to eliminate the constant

vector C in eqn (4) in order to obtain the
force-displacement
relationship via the dynamic
stiffness matrix. First, the boundary conditions for
the displacements (i.e. displacements at nodes) are
applied to eqn (4) to give
6=BC

(5)

where 6 is the displacement vector which corresponds

to the nodal displacements and B is a square matrix
obtained from matrix A when boundary conditions
for displacements are substituted.
Next, the boundary conditions for forces are
applied. The forces at the nodes of the structural
element can be related to the constant vector C in a
similar manner to that of eqn (5) to give
F=DC

(6)

where F is the force vector which corresponds to the

nodal forces and D is a frequency dependent square
matrix.
The constant vector C can now be eliminated from
eqns (5) and (6) to give
F=DB-S=K6

(7)

where
K = D B-

(8)

is the required dynamic stiffness matrix.

In eqn (8), the two steps involved to obtain the
dynamic stiffness matrix are: (i) to invert the B matrix
to give B-i; and then (ii) to premultiply the inverted
matrix (i.e. B-i) by the D matrix to give K. Computer
implementation of these steps can be accomplished
either numerically or algebraically. The authors
investigations [5, 6, 91 have shown that there are
substantial savings in computer time if explicit
analytical expressions of the elements of the dynamic
stiffness matrix are used to compute K, as opposed to
the use of numerical method to obtain K (i.e. the
matrix inversion and matrix multiplication steps of
eqn (8)). However, the task of inverting the B matrix
algebraically and then premultiplying by the D
matrix, again algebraically,
to obtain explicit
expressions for the elements of the dynamic stiffness
matrix K, can be quite formidable [6,9]. This has only
become possible due to recent advances in symbolic
computing [ 151. For the type of problems the author
has investigated [46, 91, generating the explicit
expressions for the dynamic stiffness elements has
about six-fold advantage in computational time over
the numerical method. Explicit expressions are

particularly useful when some but not all of the

stiffness coefficients are needed.
3. APPLICATION
The application of the dynamic stiffness matrix to
solve free vibration problems of structures (or
structural elements) is quite simple. First, the
dynamic stiffness matrices of all the individual
elements in a structure are assembled in the usual way
as it is done in the finite element method (except that
there is only one matrix for each element to assemble,
i.e. there are no separate mass and stiffness matrices).
Then a suitable eigen-solution procedure can be
adopted to obtain the natural frequencies. It should
be noted that the solution procedure using the
dynamic stiffness method leads to a nonlinear
eigen-value problem, and quite often the elements of
the dynamic stiffness matrix are highly irregular
(transcendental) functions of the frequency (going
through zeros and infinities). A safe and reliable way
to solve the problem is to use the Wittrick-Williams
algorithm [16] which gives all required natural
frequencies with certainty and to any desired
accuracy. The main features of the algorithm are
discussed in Refs [l-4]. Basically the algorithm needs
the overall dynamic stiffness matrix of the final
structure and also information about the clampedclamped natural frequencies of the constituent
members of the structure. The clamped+lamped
natural frequencies of an individual member in the
structure are easily identified [ 17, 181 by the zeros of
A = 0 where A usually appears in the denominator of
each element of the dynamic stiffness matrix of
individual members. The use of the algorithm has
been frequently discussed in the literature but for a
detailed insight, readers are referred to Ref. [ 161.The
Wittrick-Williams algorithm [16] essentially gives the
number of natural frequencies of a structure that
exists below an arbitrarily chosen trial frequency.
This simple feature of the algorithm is exploited to
advantage to calculate any natural frequency of the
structure to any desirable accuracy.
4.

CONCLUSIONS

A systematic procedure is given on the derivation

of the dynamic stiffness matrix of a structural
element. Generating explicit analytical expressions of
the elements of the dynamic stiffness matrix gives
significant savings in computer time. Such expressions can be derived by using symbolic computing. The application of the dynamic stiffness matrix
is discussed with particular
reference to the
Wittrick-Williams algorithm. The method presented
is fairly general.
Acknowledgement-The author is grateful to Professor F.
W. Williams for sustained encouragement.

Dynamic stiffness formulation for structural elements

REFERENCES

I. J. R. Banerjee and F. W. Williams, Exact Bemoulli-Eukr dynamic stiffness matrix for a range of tapered
beams. ht. J. Numer. Meth. Engng 21, 2289-2302
(1985).
2. J. R. Banejee, Coupled bending-torsional dynamic
stiffness matrix for beam elements. Int. J. Numer. Meth.
Engng 28, 1283-1298 (1989).
3. J. R. Banerjee and S. A. Fisher, Coupled bendingtorsional dynamic stiffness matrix for axially loaded
beam elements. Int. J. Numer. Meth. Engng X3,739-751
(1992).
4. J. R. Banejee and F. W. Williams, Coupled bendingtorsional dynamic stiffness matrix for Timoshenko
beam elements. Comput. Struct. 42, 301-310 (1992).
5. J. R. Banerjee and F. W. Williams, An exact dynamic
stiffness matrix for coupled extensional-torsional
vibration of structural members. Compur. Struct. 50,
161-166 (1994).
6. J. R. Banerjee and F. W. Williams, Coupled bendingtorsional dynamic stiffness matrix of an axially loaded
Timoshenko beam element. Int. J. Solids Struct. 31,
749-762 (1994).
7. J. R. Banerjee and F. W. Williams, Vibration
characteristics of self-expanding stayed columns for use
in space. J. Sound Vibr. 90, 245-261 (1983).
8. J. R. Banerjee and F. W. Williams, Evaluation of
efliciently computed exact vibration characteristics of
space platforms assembled from stayed columns.
J. Sound Vibr. 95, 405-414 (1984).
9. J. R. Banerjee and F. W. Williams, Vibration of
composite beams-an
exact method using symbolic
computation. J. Aircraft 32, 636-642 (1995).

103

IO. J. R. Banerjee, Flutter characteristics of tailless aircraft.

J. Aircraft 21, 733-735 (1984).
11. J. R. Bane+, Flutter modes of high aspect ratio tailless
aircraft. J. Aircraft 25, 473-476 (1988).
12. J. R. Bane+ and F. W. Williams, Users guide to the
computer program BUNVIS (Buckling or Natural
Vibration of Space frames). Department of Civil
Engineering and Building Technoiogy, University of
Wales Institute of Science and Technoloav,
__ Reoort
_ no.
5 (1982).
13. M. !I. Anderson, F. W. Williams, J. R. Banerjee, B. J.
Durling, C. L. Herstrom, D. Kennedy and D. B.
Wamaar, User manual for BUNVISRG: An exact
buckling and vibration program for lattice structures,
with repetitive geometry and substructuring option.
NASA Technical Memo 87669 (1986).
14. J. R. Banerjee, Use and capability of CALFUN-a
program for CALculation of Flutter speed Using
Normal modes. Proc. Int. 84 Athens Summer Con/. on
Modelling and Simulation, Athens, Greece (June 21-29,
1984).
15. J. Fitch, Solving algebraic problems with REDUCE.
J. Symbol. Comput. 1, 21 l-227 (1985).
16. W. H. Wittrick and F. W. Williams, A general
algorithm for computing natural frequencies of elastic
structures. Quart. J. Mech. Appl. Math. 24, 263284
(1971).
17. J. R. Banerjee and F. W. Williams, Clampedclamped
natural frequencies of a bending-torsion coupled beam.
J. Sound Vibr. 176, 301-306 (1994).
18. J. R. Banerjee, S. Guo and W. P. Howson,
Exact dynamic stiffness matrix of a bending-torsion
coupled beam including warping. Comput. Struct. 59,
613-621 (1996).