4-12 STATIC AND DYNAMIC ANALYSIS

The twelve global equilibrium equations in x-y-z reference system are now given
by:

R = Ku + R L (4.26)

The frame element stiffness matrix is:

K = T T k I J T (4.27)

It can be shown that the six fixed-end forces r J caused by member loads, which
are defined in the local 1-2-3 system, can be transformed to twelve global loads
by:

RL = T T b T r J (4.28)

It should be pointed out that within most efficient computer programs, formal
matrix multiplication
multiplication is not used to form the matrices. Programming methods are
used to skip most multiplication
multiplication by zero terms.

4.5 MEMBER END-RELEASES

Including member loading in Equation (4.23), the twelve equilibrium equations

in the local IJ reference system can be written as

fIJ = k IJ u IJ + rIJ or, without subscripts f = ku + r (4.29)

If one end of the member has a hinge, or other type of release that causes the
corresponding
corresponding force to be equal to zero, Equation (4.29) requires modification.
modification. A
typical equation is of the following form:

12
 f n = ∑ k  u
 j =1
nj  j + rn (4.30)

If we know a specific value of   f n is zero because of a release, the corresponding
corresponding
displacement un can be written as:
ONE DIMENSIONAL ELEMENTS 4-13

n −1 k nj 12 k nj
un = ∑ k 
 j =1 nn
u j + ∑ k 
 j = n+1 nn
u j + rn (4.31)

Therefore, by substitution of equation (4.31) into the other eleven equilibrium

equations, the unknown un can be eliminated and the corresponding row and
column set to zero. Or:

fIJ = k IJ uIJ + rIJ (4.32)

The terms  f n = r n = 0 and the new stiffness and load terms are equal to:

k nj
k ij = k ij − k in
k nn
(4.33)
k 
ri = ri − rn ni
k nn

This procedure can be repeatedly applied to the element equilibrium equations

for all releases. After the other displacements associated with the element have
been found from a solution of the global equilibrium equations, the
displacements associated with the releases can be calculated from Equation
(4.31) in reverse order from the order in which the displacements were
eliminated. The repeated application of these simple numerical equations is
defined in Appendix C as static condensation or partial Gauss elimination.

4.6 SUMMARY

The force method should be used to develop the stiffness matrices for one-
dimensional elements where the internal section stress-resultants can be
expressed, by satisfying equilibrium, in terms of the forces acting on the ends of 
the element. First, the flexibility matrix, with respect to a stable support system,
is developed in the element local reference system. Second, this flexibility matrix
is inverted to form the element stiffness matrix. Third, the local stiffness matrix
is expanded to include the rigid-body displacements and is modified because of 
end releases. Finally, the stiffness and load matrices are transformed into the
global reference system.
 5.

ISOPARAMETRIC ELEMENTS

 Bruce Irons, in 196 8, Revolutionized the Finite Element 

 Method by Intro ducing a Natural Coord inate
 Refere nce System

5.1 INTRODUCTION

Before development of the Finite Element Method, researchers in the field of 
structural engineering and structural mechanics found “closed form” solutions in
terms of known mathematical functions of many problems in continuum
mechanics. However, practical structures of arbitrary geometry, non-
homogeneous materials or structures made of several different materials are
difficult to solve by this classical approach.

Professor Ray Clough coined the terminology “Finite Element Method” in a

paper presented in 1960 [1]. This paper proposed to use the method as an
alternative to the finite difference method for the numerical solution of stress
concentration problems in continuum mechanics. The major purpose of the
earlier work at the Boeing Airplane Company published in 1956 [2] was to
include the skin stiffness in the analysis of wing structures and was not intended
to accurately calculate stresses in continuous structures. The first, fully
automated, finite element computer program was developed during the period of 
1961 - 1962 [3].

It is the author’s opinion that the introduction of the isoparametric element

formulation in 1968 by Bruce Irons [4] was the single most significant
5-2 STATIC AND DYNAMIC ANALYSIS

contribution to the field of finite element analysis during the past 40 years. It
allowed very accurate, higher-order elements of arbitrary shape to be developed
and programmed with a minimum of effort. The addition of incompatible
displacement modes to isoparametric elements in 1971 was an important, but
minor, extension to the formulation [5].

5.2 A SIMPLE ONE-DIMENSIONAL EXAMPLE

To illustrate the fundamentals of the isoparametric approach, the one-

dimensional, three-node element shown in Figure 5.1 is formulated in a natural
coordinate reference system.

 x 
u1  R1  A ( x ) = 6 − u3  R3 u2  R2
10  2
1 3
 A1 = 10  A2 = 2
50 30
-40
 x
0
A. GLOBAL REFERENCE SYSTEM “X ”

s=-1.0 s=0 s=1.0

-s +s
1.0
 N 1 = −s(1− s)/ 2

1.0
 N 2 = s( 1+ s)/ 2

2 1.0
 N 3 =1− s

B. ISOPARAMETRIC REFERENCE SYSTEM “s”

 Figure 5.1 A Simple Example of an Isoparametric Element

ISOPARAMETRIC FORMULATION 5-3

The shape functions N i are written in terms of the element isoparametric
reference system. The "natural" coordinate s has a range of  s = ±1.0 . The
isoparametric and global reference systems are related by the following
elementary equation:

x( s) ≡ N 1 (s) x1 + N 2 ( s) x 2 + N 3 ( s) x 3 = N( s) x (5.1)

The validity of this equation can be verified at values of  s = −1 , s = 0 and

s = 1 . No additional mathematical references are required to understand
Equation (5.1).

The global displacement can now be expressed in terms of the fundamental

isoparametric shape functions. Or:

u( s) = N 1 (s) u1 + N 2 ( s) u 2 + N 3 ( s) u 3 = N( s) u (5.2)

Note that the sum of the shape functions is equal to 1.0 for all values of  s ;
therefore, rigid-body displacement of the element is possible. This is a
fundamental requirement of all displacement approximations for all types of 
finite elements.

The strain-displacement equation for this one-dimensional element is:

∂u(s) du(s) du(s) ds

εx = = = (5.3)
∂x dx ds dx

You may recall from sophomore calculus that this is a form of the chain rule. For
any value of  s the following equations can be written:

du( s)
= N(s),s u (5.4a)
ds

dx
= N(s),s x =  J (s) (5.4b)
ds

Therefore:

du(s) ds 1
ε x = = N(s),s u = B(s) u (5.5)
ds dx  J ( s)
5-4 STATIC AND DYNAMIC ANALYSIS

From Equation (5.1), the derivatives with respect to the global and isoparametric
reference systems are related by:

dx = N(s),s x ds =  J ( s) ds (5.6)

The 3 by 3 element stiffness can now be expressed in terms of the natural system:

+1

∫ 
K = B(s) T E B(s) J (s) ds
−1
(5.7)

In general, Equation (5.7) cannot be evaluated in closed form. However, it can be

accurately evaluated by numerical integration.

5.3 ONE-DIMENSIONAL INTEGRATION FORMULAS

Most engineers have used Simpson’s rule or the trapezoidal rule to integrate a
function evaluated at equal intervals. However, those traditional methods are not
as accurate, for the same computational effort, as the Gauss numerical integration
method presented in Appendix G. The Gauss integration formulas are of the
following form:
+1 n
 I  = ∫  f (s ) ds = ∑W   f (s )
−1 i =1
i i (5.8)

The Gauss points and weight factors for three different formulas are summarized
in Table 5.1.

Table 5.1 Gauss Points and Weight Factors for Numerical Integration

n s1 W 1 s2 W 2 s3 W 3

1 0 2

2 −1 3 1 1 3 1

3 − 0.6 59 0 89 0.6 59