Customized Content
Bundlethe more
books you buy,
the greater your
discount!
Manufacturing
Engineering
Mechanical
& Chemical
Engineering
Materials Science
& Engineering
Civil &
Environmental
Engineering
Electrical
Engineering
THE TERMS
Perpetual access for
a one time fee
No subscriptions or
access fees
Unlimited
concurrent usage
Downloadable PDFs
Free MARC records
For further information,
a free trial, or to order,
contact:
sales@momentumpress.net
THE CONTENT
OHARA RAMMING
EBOOKS
FOR THE
ENGINEERING
LIBRARY
SUSTAINABLE STRUCTURAL
SYSTEMS COLLECTION
Mohammad Noori, Editor
Numerical
Structural
Analysis
Steven OHara
Carisa H. Ramming
NUMERICAL
STRUCTURAL
ANALYSIS
NUMERICAL
STRUCTURAL
ANALYSIS
STEVEN E. OHARA
CARISA H. RAMMING
Abstract
As structural engineers move further into the age of digital computation and rely more heavily on computers to solve problems, it remains
paramount that they understand the basic mathematics and engineering
principles used to design and analyze building structures. The analysis of
complex structural systems involves the knowledge of science, technology, engineering, and math to design and develop efficient and economical buildings and other structures. The link between the basic concepts and
application to real world problems is one of the most challenging learning
endeavors that structural engineers face. A thorough understanding of the
analysis procedures should lead to successful structures.
The primary purpose of this book is to develop a structural engineering students ability to solve complex structural analysis problems that
they may or may not have ever encountered before. The book will cover
and review numerical techniques to solve mathematical formulations.
These are the theoretical math and science principles learned as prerequisites to engineering courses, but will be emphasized in numerical formulation. A basic understanding of elementary structural analysis is important
and many methods will be reviewed. These formulations are necessary
in developing the analysis procedures for structural engineering. Once
the numerical formulations are understood, engineers can then develop
structural analysis methods that use these techniques. This will be done
primarily with matrix structural stiffness procedures. Both of these will
supplement both numerical and computer solutions. Finally, advanced
stiffness topics will be developed and presented to solve unique structural
problems. These include member end releases, nonprismatic, shear, geometric, and torsional stiffness.
KEY WORDS
adjoint matrix, algebraic equations, area moment, beam deflection, carryover factor, castiglianos theorems, cofactor matrix, column matrix,
v
vi Abstract
Contents
List of Figures
xi
List of Tables
xv
Acknowledgments
xix
1.1 Equations
1.2 Polynomials
10
12
13
15
17
18
21
24
38
References
46
47
47
2.2 Matrices
48
52
56
57
vii
viiiContents
63
68
72
73
78
80
86
87
90
91
References
95
97
97
99
104
109
112
114
116
118
123
136
140
References
145
147
147
148
153
155
158
161
165
168
Contents ix
170
174
181
186
187
References
204
205
205
214
217
229
239
243
247
251
258
260
262
5.12 Torsion
265
5.13 Substructuring
266
References
268
271
Index
273
List of Figures
Figure 1.1. Incremental search method.
11
12
14
16
Figure 1.5.
Secant method.
18
Figure 1.6.
19
20
20
21
23
27
98
100
104
129
132
134
141
142
142
xi
148
Figure4.2.
Example4.1 Rotation, a.
149
150
151
153
153
154
156
156
157
157
159
159
160
161
162
162
163
164
164
Figure4.21.
Example4.10 Castiglianos second theorem.
166
166
167
169
172
175
176
176
178
181
183
184
190
191
197
206
209
210
211
218
230
235
236
237
239
241
243
245
249
250
251
252
255
258
260
List of Tables
Table 1.1. Synthetic division
10
10
11
13
15
17
18
21
24
26
39
58
xv
Table 2.2.
Example 2.5 Cofactor method
61
Table 2.3.
Example 2.6 Method of adjoints
62
65
67
74
77
78
78
78
81
82
83
84
85
85
86
88
92
94
98
99
102
102
102
102
107
107
108
108
110
110
111
111
113
115
143
143
144
145
173
174
199
200
200
202
202
203
203
203
214
217
Table5.3. Example5.3 M
ember stiffness, member 1
220
221
222
223
224
225
226
Table 5.10
Example 5.3 Member stiffness
227
228
228
228
229
229
251
Acknowledgments
Our sincere thanks go to Associate Professor Christopher M. Papadopoulos,
PhD, Department of Engineering Science and Materials University of
Puerto Rico, Mayagez.
We would also like to thank the ARCH 6243 Structures: AnalysisII, Spring 2014 class: Kendall Belcher, Conner Bowen, Harishma
Donthineni, Gaurang Malviya, Alejandro Marco Perea, Michael
Nachreiner, Sai Sankurubhuktha, Timothy Smith, Nuttapong Tanasap,
Ignatius Vasant, and Lawrence Wilson.
A special thanks to Nicholas Prather for his assistance with figures.
xix
CHAPTER 1
Roots of Algebraic
and Transcendental
Equations
In structural engineering, it is important to have a basic knowledge of how
computers and calculators solve equations for unknowns. Some equations
are solved simply by algebra while higher order equations will require
other methods to solve for the unknowns. In this chapter, methods of finding roots to various equations are explored. The roots of an equation are
defined as values of x where the solution of an equation is true. The most
common roots are where the value of the function is zero. This would
indicate where a function crosses an axis. Roots are sometimes complex
roots where they contain both a real number and an imaginary unit.
1.1EQUATIONS
Equations are generally grouped into two main categories, algebraic equations and transcendental equations. The first type, an algebraic equation,
is defined as an equation that involves only powers of x. The powers of x
can be any real number whether positive or negative. The following are
examples of algebraic equations:
8 x3 3x 2 + 5 x 6 = 0
1
+2 x =0
x
x1.25 3p = 0
The second type is transcendental equations. These are nonalgebraic
equations or functions that transcend, or cannot be expressed in terms of
algebra. Examples of such are exponential functions, trigonometric functions, and the inverses of each. The following are examples of transcendental equations:
cos ( x ) + sin ( x ) = 0
e x + 15 = 0
Transcendental functions may have an infinite number of roots or may not
have any roots at all. For example, the function sin(x) =0 has an infinite
number of roots x = kx and k=0, 1, 2.
The solution of algebraic or transcendental equations is rarely carried
out from the beginning to end by one method. The roots of the equation
can generally be determined by one method with some small accuracy,
and then made more accurate by other methods. For the intent and purpose of this text, only a handful of the available methods are discussed.
These methods include: Descartes Rule, Synthetic Division, Incremental
Search, Refined Incremental Search, Bisection, False Position, Secant,
NewtonRaphson, Newtons Second Order, Graeffes Root Squaring, and
Bairstows methods. Some of these methods are used to solve specific
types of equations, while others can be used for both equation types.
1.2POLYNOMIALS
A polynomial is defined as an algebraic equation involving only positive
integer (whole number) powers of x. Polynomials are generally expressed
in the following form:
a0 x n + a1 x n 1 + a2 x n 2 + a3 x n 3 + + an 1 x1 + an = 0
In most cases, the polynomial form is revised by dividing the entire equation by the coefficient of the highest power of a, a0, resulting in the following form:
x n + a1 x n 1 + a2 x n 2 + a3 x n 3 + + an 1 x1 + an = 0
For these polynomials, the following apply:
= 3 sign changes
Since there are three sign changes, there is a maximum of three positive
roots. Three positive real roots exist or one positive real root plus two
imaginary roots.
Find possible negative roots by rewriting the function for f(x)=0:
0 sign changes
Notice the signs of all the odd powers reverse while the signs of the even
powers remain unchanged. Count the number of sign changes, n. This
number is the maximum possible negative roots. Since there is no sign
change, zero negative roots exist.
Possible complex roots:
Complex roots appear in conjugate pairs. Therefore, either zero or two
complex roots exist. In this example the roots are x=1, 2, 3.
Example 1.2Descartes rule
Find the possible number of positive, negative, and complex roots for the
following polynomial:
x3 7 x 2 + 6 = 0
Find possible positive roots for f(x)=0:
x3 7 x 2 + 6 = 0
1
2 =
2 sign changes
Since there are two sign changes, there is a maximum of two positive
roots. Two or zero positive real roots exist.
( x )3 7 ( x ) 2 + 6 = x 3 7 x 2 + 6 = 0
x3 7 x 2 + 6 = 0
0
1 =
1 sign change
Again, the signs of all the odd powers reverse while the signs of the even
powers remain unchanged. Count the number of sign changes, n. This
number is the maximum possible negative roots. Since there is one sign
change, one negative root exists.
Possible complex roots:
Complex roots appear in conjugate pairs. Therefore, either zero or two
complex roots exist. In this example, the roots are x=1, 2, 3.
Example 1.3 Descartes rule
Find the possible number of positive, negative, and complex roots for the
following polynomial:
x3 3x 2 + 4 x 6 = 0
Find possible positive roots for f(x)=0:
x3 3x 2 + 4 x 6 = 0
1
3 = 3 sign changes
Since there are three sign changes, there is a maximum of three positive
roots. Three or one positive real roots exist.
Find possible negative roots by rewriting the function for f(x)=0:
( x )3 3 ( x ) 2 + 4 ( x ) 6 = x 3 3 x 2 4 x 6 = 0
x3 3x 2 4 x 6 = 0
0
0 =
0 sign changes
Count the number of sign changes. Since there is no sign change, zero
negative roots exist.
2 sign changes
Since there are two sign changes, there is a maximum of two positive
roots. Two or zero positive real roots exist.
Find possible negative roots by rewriting the function for f(x) = 0:
( x )3 ( x ) 2 + 2 ( x ) = x 3 x 2 2 x = 0
x3 x 2 2 x = 0
0
0 sign changes
Count the number of sign changes, n. Since there is no sign change, zero
negative roots exist.
Possible complex roots:
Complex roots appear in conjugate pairs. Therefore, either zero or two
complex roots exist. In this example, the roots are x = 0, 1+i, 1i.
The existence of zero as a root could have been discovered by noticing
that there was not a constant term in the equation. Therefore, dividing the
equation by x yields the same as x=0.
a0
a1
a2
an1
an
rb1
rb2
rbn1
rbn
b1
b2
b3
bn
6
11
6
Add the columns by starting at the left. Multiply each result by r=1 and
add this to the next column.
Table 1.3. Example 1.5 Synthetic division
1
1
0
1
6
11
6
5
5
6
0
Since the remainder, R, is zero, f(r)=0 and r=1 is a root. The polynomial can now be written as a linear equation, xr or x1, and
the resulting reduced polynomial with coefficient of the resultants as
follows:
x 3 6 x 2 + 11x 6 = ( x 1) x 2 5 x + 6 = 0
Use the quadratic equation to reduce the remaining polynomial as follows:
b b 2 4ac ( 5) ( 5) 4 (1) 6 5 1
x=
=
=
= 2, 3
2a
2 (1)
2
2
1
0
1
5
6
3
Since the remainder, R, is zero, f(2)=0 and r=2 is a root. The resulting
polynomial is x3, thus x=r is the third root. This can also be shown by
repeating division with x3.
Table 1.5. Example 1.5 Synthetic division
3
1
0
1
3
3
0
Since the remainder, R, is zero, f(3)=0 and r=3 is a root. The polynomial
is now written as:
x 3 6 x 2 + 11x 6 = ( x 1) x 2 5 x + 6 = ( x 1) ( x 2) ( x 3) = 0
The roots are x=1, 2, 3.
Example 1.6Synthetic division
Find f(1), f (1), and f (1) or perform three divisions of the following
polynomial by x +1:
x 3 6 x 2 + 11x 6 = 0
Set up the equation as shown in the following by writing the divisor, r, and
coefficient, a, in the first row.
Table 1.6. Example 1.6 Synthetic division
1
1
6
11
6
Add the columns by starting at the left. Multiply each result by r=1 and
add this to the next column.
Table 1.7. Example 1.6 Synthetic division
1
1
0
1
11
7
18
6
1
7
6
18
24
1
0
1
0
1
6
1
7
1
8
11
7
18
8
26
6
18
24
f ( x ) = 3 x 2 12 x + 11 = 0
f ( 1) = 3( 1) 12 ( 1) + 11 = 26
2
6
11
6
0
1
0
1
0
1
1
7
18
8
26
18
7
1
8
24
1
9
f ( 1) = 6 ( 1) 12 = 18
0.25
0.5
0.75
13.8240
9.2934
5.7190
3.0071
1.0640
1.25
0.2041
f(x)
f(x)
f(xi)
xi
xi+1
f(xi+1)
(x)
Since the sign of f(x) changed between x=1 and x=1.25, it is assumed
that a root was passed between those values. The actual root occurs at
x=1.2. In this example, five digits of precision were used, but in most
cases it is a good rule to carry one more digit in the calculations than in the
desired accuracy of the answer.
Once the roots have been bounded by the incremental search method,
other methods can be utilized in finding more accurate roots: The following sections will cover the refined incremental search, bisection, false
position, secant, NewtonRaphson, and Newtons second order methods to determine more accurate roots of algebraic and transcendental
equations.
f(x)
x/10
f(x)
f(xi)
xi
f(xi+1)
xi+1
(x)
1.1
0.4810
1.125
0.3514
1.15
0.2281
1.175
0.1110
1.2
0.0000
Since the sign of f(x) changed between x=1.175 and x=1.2, it is assumed
that a root was passed between those values. The actual root occurs at
x=1.2.
1.7BISECTION METHOD
After a sign change has occurred in a search method, another way to rapidly converge (become closer and closer to the same number) on a root is
the bisection method, also known as the halfinterval method or the Bolzano method developed in 1817 by Bernard Bolzano. This method takes
the bounded increment between two points xi and xi+1 where f(xi)f(xi+1)0
and divides it in two equal halves or bisects the increment. The two subintervals have the first interval from xi to xi+ and the second interval from
xi+ to xi+1 as seen in Figure 1.3.
Next, the subinterval containing the root can be found by the following algorithm:
f(xi)f(xi+ ) < 0, first interval contains the root
f(xi)f(xi+ ) > 0, second interval contains the root
f(xi)f(xi+ ) = 0, xi+ is the root
f(x)
f(xi)
xi
f(xi+1/2)
xi+1/2
xi+1
(x)
f(xi+1)
x
1.25
1.125
1.1875
1&2
2&3
1.0640
0.2041
0.3514
0.0548
x
interval
f(x)
5
1.2188
1.2031
1.1953
1.1992
interval
2&4
4&5
4&6
6&7
f(x)
0.0793
0.0135
0.0204
0.0034
0 f ( x1 )
x3 x1
x3 = x1 f ( x1 )
x2 x1
f ( x2 ) f ( x1 )
x1 f ( x2 ) x2 f ( x1 )
f ( x2 ) f ( x1 )
The relationship between x1, x2, and x3 can be seen in Figure 1.4.
f(x1)f(x3) < 0, first interval contains the root
f(x1)f(x3) > 0, second interval contains the root
f(x1)f(x3) = 0, x3 is the root
(1.1)
f(x)
f(x1)
x1
x3
x2
(x)
f(x2)
If the first interval contains the root, the values for the next cycle for x1 and
x2 and the corresponding functions f(x1) and f(x2) are as follows:
x1 and f(x1) remain unchanged
x2= x3
f(x2)= f(x3)
If the second interval contains the root, then the values are used for x1 and
x2 and the corresponding functions f(x1) and f(x2) are as follows:
x2 and f(x2) remain unchanged
x1=x3
f(x1)=f(x3)
The process is continued until the desired accuracy is obtained.
Example 1.10 Method of false position
Refine the search of the function from Example 1.7 between 1.0 and 1.25
using the false position method to increase the accuracy of the approximate root. For the accuracy test use =0.01 that is f(x) < :
Using Equation 1.1 to solve for a closer point between x1=1 and x2=1.25.
x3 =
x1 f ( x2 ) x2 f ( x1 )
f ( x2 ) f ( x1 )
Repeat this process until the desired accuracy of the root is achieved as
shown in Table 1.13.
Table 1.13. Example 1.10 Method of false position
x
1.25
1.2098
1.2018
1&2
1&3
0.0417
0.0080
interval
f(x)
1.064
0.2041
f(x3)
x3
x1
x2
(x)
f(x1)
f(x2)
The process of finding the new value is the same as linear interpolation
using Equation 1.1 to solve for a closer point between x1=1 and x2=1.25.
x3 =
x1 f ( x2 ) x2 f ( x1 )
f ( x2 ) f ( x1 )
The reassignment of the values simply uses the last two values and their
corresponding functions as shown in Table 1.14.
Table 1.14. Example 1.11 Secant method
x
f(x)
1.25
1.2098
1.1994
1&2
2&3
1.0640
0.2041
0.0417
0.0025
This happens to be similar to the false position Example 1.10 as only interpolations occur, but with different subintervals.
1.10NEWTONRAPHSON METHOD OR
NEWTONS TANGENT
The NewtonRaphson method uses more information about the function
to speed up convergence. It was originally developed by Issac Newton in
f(x)
f(xn)
xn
xn+1
(x)
f(xn)
f ( xn )
f ( xn )
xn+1 = xn
xn xn+1
f ( xn )
(1.2)
Repeat the process using a new value until convergence occurs. Convergence may not occur in the following two cases:
xn+3
xn
xn+1
(x)
xn+2
xn
xn+2
xn+4
xn+1 xn+3
(x)
The derivative of the function must be obtained to find the slope at any
given value.
Beginning with xn=1.25, use Equation 1.2 to determine the next value.
f ( xn )
f ( xn )
= 1.25
0.2041
= 1.1969
3.8475
Repeat the process until the desired accuracy is obtained in Table 1.15.
Table 1.15. Example 1.12 NewtonRaphson method
x
1.25
1.1969
1.19999
f(x)
f(x)
0.2041
0.0132
0.00004
3.8475
4.3493
4.32010
f ( xn ) ( x )
2!
f ( xn ) ( x )
3!
For a means of determining a value of x that will make the Taylor series
expansion go to zero, the first three terms of the right hand side of the
equation are set equal to zero to obtain an approximate value.
f(x)
f(x)
xn
xn+1
(x)
f ( xn ) ( x )
f ( xn ) + ( x ) f ( xn ) +
=0
2
The exact value of x cannot be determined from this equation since only
the first three terms of the infinite series were used in the calculation.
However, a close approximation of the root is a result. When using this
equation to calculate x, a quadratic must be solved yielding two possible
roots. In order to avoid this problem, x = f (xn)/f (xn) from Newtons
tangent may be substituted into the bracketed term only.
f ( xn ) f ( xn )
f ( xn ) + ( x ) f ( xn )
=0
2 f ( xn )
f ( xn )
x =
f ( xn ) f ( xn )
f ( xn )
2 f ( xn )
Observing Figure 1.9 we see that x = xn+1  xn. Substituting into the previous equation, Equation 1.3 is obtained as follows:
f ( xn )
xn+1 = xn
f ( xn ) f ( xn )
f ( xn )
2 f ( xn )
(1.3)
If the first derivative is small, the slope is close to zero near the value
and the next approximation may be inaccurate. Therefore, use the second
derivative term as follows:
f ( xn ) = 0
f ( xn ) ( x )
f ( xn ) + ( x )
=0
2
f ( xn ) ( x )2
f ( xn ) = ( x )
f ( xn )
x2 +
=0
f ( xn )
2
b b 2 4ac
=
2a
x =
2 (1)
f ( xn )
f ( xn )
2
f ( xn )
f ( xn )
2
2
=
2
f ( xn )
f ( xn )
2
f ( xn )
(1.4)
f ( xn )
2
This process is a good tool for finding two roots that are near each other.
This will happen when the slope is close to zero near a root. Double roots
occur when the first derivative is zero, triple roots occur when the first
and second derivatives are zero, and so on. These are shown graphically
in Figure1.10.
Example 1.13 Newtons second order method
Refine the search from Example 1.7 with a starting value of 1.25 using the
Newtons second order method to increase the accuracy of the approximate root. For the accuracy test use e = 0.01.
f(x)
f(x)
f(x)
(x)
Single root
(x)
Double root
(x)
Triple root
Beginning with xn=1.25, use Equation 1.3 to determine the next value.
xn +1
f
x
( n)
0.2401
= 1.25
= xn
9.3) 0.2041
f ( xn ) f ( xn )
(
f ( xn )
3.8475
2 (3.8475)
2 f ( xn )
= 1.2001
Repeat the process until the desired accuracy is obtained in Table 1.16.
Table 1.16. Example 1.13 Newtons second order method
x
1.25
1.200143
1.2
f(x)
0.2041
0.00062
0.000000
f(x)
3.8475
4.31863
4.320000
f(x)
9.3000
9.59914
9.600000
f ( x ) = ( 1) ( x a1 ) ( x a2 ) ( x a3 )( x an )
)(
)(
) (
For example, use a third degree polynomial with roots x1, x2, and x3 as
follows:
f ( x ) = 0 = x 3 + a1 x 2 + a2 x + a3
A polynomial with roots x1, x2, and x3 follows:
f ( x ) = 0 = x 3 + a1 x 2 a2 x + a3
Multiplying the two equations together yields the following:
a2
a1
(c1/c0)1/m
c1
(c2/c1)1/m
c2
2b0b6
c0
c3
(c3/c2)1/m
(cn1/cn2)1/m
cn1
2b2b4
+2b1b5
2b1b3
bn1
bn12
+2b0b4
b3
b32
2b0b2
b2
b22
b1
2a0a6
b12
2a2a4
+2a1a5
2a1a3
+2a0a4
an1
2
an1
an1x
b0
2a0a2
a
2
3
a3
a3xn3
b02
a
2
2
0
a2
a1
a0
a2xn2
a1xn1
a0xn
(cn/cn1)1/m
cn
bn2
bn
a
2
n
an
anx0
cycle
a n 1
n 2 2a2 a4 n 3
0 = yn +
y
+
2
a
a
y
+
y
1 3
2a2
+2a1a5
+2a
4
2a6
a42
2a3 a5
+ +2a2 a6 y n 4 + + an2
2a a
1 7
2
a
+
2
1
f(x)
f(x)
f(x)
(x)
(x)
(x)
f ( x ) = 0 = ( x x1 ) ( x x2 ) ( x x3 )
If the previous equation is multiplied out, the following is the result:
f ( x ) = 0 = x 3 ( x1 + x2 + x3 ) x 2 + ( x1 x2 + x1 x3 + x2 x3 ) x x1 x2 x3
Therefore, for the polynomial, the original coefficients are as follows:
a1 = ( x1 + x2 + x3 )
a2 = x1 x2 + x1 x3 + x2 x3
a3 = x1 x2 x3
The Enke roots of x1 = r1, x2 = r2, and x3 = r3 are substituted in. The
sign has been lost so the Enke roots are used as the basis (x1 = r1, x2 = r2,
etc.), then the following is true:
a1 = r1 + r2 + r3
a2 = r1r2 + r1r3 + r2 r3
a3 = r1r2 r3
As the cycles (m) continue, the derived polynomial becomes the following:
f ( x ) = 0 = y 3 + b1 y 2 + b2 y + b3
b1 r1m r1 b1 m
b
b
b
b2 r r r m2 2 r2 = 2
b1
r1
b1
m m
1 2
m
2
b
b
b
b3 r r r r m 3 m 3 r3 = 3
b2
r1 r2
b2
m m m
1 2 3
m
3
The Enke roots only lack the proper sign and either positive or negative
may be correct, so a check is necessary.
Example 1.14
Graeffes root squaring methodreal and distinct
roots.
Find the root of the following polynomial using Graeffes root squaring
method.
f ( x) = 0 = x 4 10 x 3 + 35 x 2 50 x + 24
x4
a1x3
a2x2
a3x
a4x0
cycle
10
35
50
24
100
1225
2500
576
70
1000
1680
48
2
1
1
30
273
820
576
331776
900
74529
672400
546
49200
314496
1152
4
354
26481
357904
331776
1.253E+05
7.012E+08
1.281E+11
1.101E+11
5.296E+04
2.534E+08
1.757E+10
6.636E+05
8
7.235E+04
4.485E+08
1.105E+11
1.101E+11
5.235E+09
2.012E+17
1.222E+22
1.212E+22
8.970E+08
1.599E+16
9.874E+19
4.338E+09
1.852E+17
1.212E+22
1.212E+22
1.882E+19
3.429E+34
1.468E+44
1.468E+44
3.703E+17
1.051E+32
4.487E+39
2.202E+11
16
2.423E+22
32
1.845E+19
3.418E+34
1.468E+44
1.468E+44
3.404E+38
1.168E+69
2.155E+88
2.155E+88
6.836E+34
5.417E+63
1.004E+79
2.936E+44
64
3.403E+38
1.168E+69
2.155E+88
2.155E+88
1.158E+77
1.365E+138
4.646E+176
4.646E+176
2.337E+69
1.467E+127
5.037E+157
4.311E+88
128
r
1.158E+77
1.365E+138
4.646E+176
4.646E+176
Refer to Table 1.17 for the basic procedure for the root squaring. Table1.18
shows the process for this polynomial.
To determine the proper sign of the roots from the Enke roots, a check
is required.
1
77
r1 b1 m = 1.158 (10)
b
r2 = 2
b
b
r3 = 3
b
b
r4 = 4
b
3
1
128
= 4.000
1.365 (10)138
=
77
1.158 (10)
1
128
4.646 (10)176
=
138
1.365 (10)
4.646 (10)176
=
176
4.646 (10)
= 3.000
1
128
= 2.000
1
128
= 1.000
Substituting the Enke roots into the original equations yields x1 = 4.000,
x2= 3.000, x3 = 2.000, and x4 = 1.000.
1.12.2 REAL AND EQUAL ROOTS
After many cycles of squaring the polynomial, the second possible solution type will occur when the coefficients of the derived polynomial are
the squares of the terms in the preceding cycle with the exception of one
term that is the square of the term in the preceding cycle. This indicates
that two of the roots are equal, the one with the squared term and the
next one to the right. Furthermore, if one term is 1/3 the square of the term
in the proceeding cycle, three of the roots are equalthe term with the
1
/3 squared term and the next two to the right. A similar relationship will
occur if four or more roots are equal. The roots (Enke roots) will have a
relationship similar to the following assuming r1 = r2 and considering only
the dominant terms in each expression:
b1 = r1m + r2m + r3m = r1m + r1m b1 2r1m
b2 = r1m r2m + r1m r3m + r2m r3m = r1m r1m b2 r12 m
b3 = r1m r2m r3m = r1m r1m r3m b3 r12 m r3m
m
1
b2 r12 m r1 = (b2 )
2m
= r2
b
b3 r r = b r r3 = 3
b
2m m
1
3
m
2 3
After the multiple roots have been passed, the rest of the terms have the
regular solution relationship and will appear as follows:
b
b
r n rn = n
bn 1
bn 1
m
n
If the second term was the square of the term in the previous cycle, then
the solution would appear as follows, assuming r2 = r3 and considering
only the dominant terms in each expression:
b1 = r1m + r2m + r3m = r1m b1 r1m
b2 = r1m r2m + r1m r3m + r2m r3m = r1m r2m + r1m r2m b2 2r1m r2m
b3 = r1m r2m r3m = r1m r2m r2m b3 r1m r22 m
These become the following:
b1 r1m r1 (b1 )
b
b2 2r r = 2b r r2 = 2
2b1
m m
1 2
m 2m
1 2
b3 r r
m
1 2
2m
1 2
=br
b
r2 = 3
b
2m
= r3
= r3
Similar to the previous case where r1 = r2, after the multiple roots have
been passed, the rest of the terms have the regular solution relationship
and will appear as follows:
b
b
r n rn = n
bn 1
bn 1
m
n
Note the pattern of the powers and the relationship for any other variations
of two roots may be found. Derive one more case for a triple root. If the
first term was 1/3 the square of the term in the previous cycle, it would indicate a triple root or r1 = r2 = r3. If we consider only the dominant terms in
each expression, the following relationships occur:
b1 = r1m + r2m + r3m + r4m = r1m + r1m + r1m b1 3r1m
b2 = r1m r2m + r1m r3m + r1m r4m + r2m r3m + r2m r4m + r3m r4m = r1m r1m
+ r1m r1m + r1m r1m b2 3r12 m
b3 = r1m r2m r3m + r1m r2m r4m + r1m r3m r4m + r2m r3m r4m = r1m r1m r1m b3 r13m
These become the following:
b
b1 3r1m r1 1
3
b
b2 3r12 m r1 = 2
2
2m
= r2 = r3
3m
= r2 = r3
After the multiple roots have been passed, the rest of the terms have the
regular solution relationship and will appear as follows:
b
b
r n rn = n
bn 1
bn 1
m
n
Just like with the regular solution for real and distinct roots, the Enke roots
only lack the proper sign and either + or must be checked.
Example 1.15
Graeffes root squaring methodreal and equal
roots
Find the root of the following polynomial using Graeffes root squaring
method.
f ( x) = 0 = x3 + 3x 2 4
Refer to Table 1.17 for the basic procedure for the root squaring. Table 1.19
shows the process for this polynomial.
x3
a1x2
a2x1
a3x0
cycle
1
1
1
1
16
256
1
1
256
65536
1
1
65536
4.295E+09
16
1
1
4.295E+09
1.845E+19
32
1
1
1.845E+19
3.403E+38
64
1
1
3.403E+38
1.158E+77
128
1
1
1.158E+77
1.341E+154
256
r
0
0
24
24
576
288
288
82944
16896
66048
4.362E+09
6.724E+07
4.295E+09
1.845E+19
1.126E+15
1.845E+19
3.403E+38
3.169E+29
3.403E+38
1.158E+77
2.511E+58
1.158E+77
1.341E+154
1.576E+116
1.341E+154
2
4
16
3
9
0
9
81
48
33
1089
576
513
2.632E+05
1.321E+05
1.311E+05
1.718E+10
8.590E+09
8.590E+09
7.379E+19
3.689E+19
3.689E+19
1.361E+39
6.806E+38
6.806E+38
4.632E+77
2.316E+77
2.316E+77
2
1.341E+154
1
Notice that the first term in the table for cycle 9 is the square of the term
in the previous cycles and the following solution applies:
b
r1 1
2
r2 (b2 )
b
r3 3
b2
2.316(10)77
=
2m
= 1.341(10)154
1.341(10)154
=
154
1.341(10)
256
= 2.000
2 ( 256)
256
= 2.000 = r1
= 1.000
Substituting the Enke roots into the original equations yields x1=2.000,
x2=2.000, and x3=1.000.
1.12.3 REAL AND COMPLEX ROOTS
The third possible solution type will occur after many cycles of squaring
the polynomial; the coefficients of the derived polynomial are the squares
of the terms in the preceding cycle, except if one or more terms have a
sign fluctuation, then two of the roots are complexthe one with the sign
fluctuation term and the next one to the right constitutes the complex conjugate pair of roots. The roots (Enke roots) will have a relationship similar
to the following assuming r3 and r4 are the complex conjugate pair of roots
and considering only the dominant terms in each expression:
x3 = Reiq = ( cosq + isinq ) = u + iv
x4 = Re iq = ( cosq isinq ) = u iv
The values i and R for the complex form in polar or Cartesian simple
form are:
i = 1
and R = u 2 + v 2
)
= r r + ( r Re q ) + ( r Re q ) + ( r Re q ) + ( r Re q )
= ( r r Re q ) + ( r r Re q ) + ( r R ) + ( r R )
= (r r R )
m m
1 2
1 2
1 2
1 2
2 m
2 m
2 m
b4 = r1r2 R
2 m
2m
(r
m
1
+ r2m
+ R2m
b
b2 = r r = b r r2 = 2
b
m m
1 2
m
1 2
b3 = 2 ( r1r2 R ) cos mq
m
b4 = r1r2 R 2
2m
a1 = ( x1 + x2 + (u + vi ) + (u vi ))
a1 = ( x1 + x2 + 2u )
Use b1 and b2 to find r1 and r2, then x1 and x2. Use b4 and b2 to find R then
use a1 to find u and R to find v. The Enke roots, once again, only lack the
proper sign and either + or may be correct.
Example 1.16Graeffes root squaring methodreal and
complex roots
Find the root of the following polynomial using Graeffes root squaring
method.
f ( x) = 0 = x 4 + x 3 6 x 2 14 x 12
Refer to Table 1.17 for the basic procedure for root squaring. Table 1.20
shows the process for this polynomial.
The third term has a sign fluctuation thus the previously derived relationships apply and the following is the solution:
x4
a1x3
a2x2
a3x
a3x0
cycle
6
14
12
36
196
144
12
28
144
24
2
13
40
52
144
169
1600
2704
20736
80
1352
11520
288
4
1
1
89
536
8816
20736
429981696
7921
287296
77721856
1072
1569248
22228992
6.849E+03
1.898E+06
5.549E+07
4.300E+08
4.691E+07
3.602E+12
3.079E+15
1.849E+17
41472
8
4.311E+07
2.843E+12
1.447E+15
1.849E+17
1.859E+15
8.084E+24
2.094E+30
3.418E+34
1.853E+15
7.959E+24
1.043E+30
3.418E+34
3.434E+30
6.334E+49
1.088E+60
1.168E+69
1
1
3.434E+30
6.334E+49
5.441E+59
1.168E+69
1.179E+61
4.012E+99
2.960E+119
1.365E+138
1.179E+61
4.012E+99
1.480E+119
1.365E+138
1.421892602
1.40657599
r1 = (b1 )
b
r2 = 2
b
1
= 1.179 (10)
m
61
1
128
4.012 (10)99
=
61
1.179 (10)
= 3.000
1
128
= 2.000
Substituting the Enke roots into the original equations yields x1 = 3.000
and x2 = 2.000.
b
R = 4
b
2
2m
1.3652 (10)138
=
99
4.012 (10)
256
= 1.414 = 2
Using the fact that R2 = u2 + v2, the following is used to find u and v:
1 = a1 = ( x1 + x2 + 2u ) = (3 2 + 2u ) u = 1
R = u 2 + v2 v = R2 u 2 = 2 1 = 1
Thus results are x3 = 1 + i and x4 = 1  i.
1.13BAIRSTOWS METHOD
Bairstows method was first published by Leonard Bairstow in 1920
(Bairstow 1920). If we divided a polynomial of nth degree by a quadratic
equation, the result will be a polynomial of n2 degree plus some remainder. This remainder can be used to give a closer approximation of the root
quadratic equation. When the remainder is zero, the quadratic is a root
equation. Bairstows method involves using the remainders from double
synthetic division to approximate the error in an assumed quadratic root
equation of a polynomial. The derivation is omitted from this text, but
may be found in Applied Numerical Methods for Digital Computations,
by James, Smith and Wolford (1977). Look at the process of factoring a
polynomial into a quadratic equation times a polynomial of two degrees
less than the original polynomial as follows:
(x
x n + a1 x n 1 + a2 x n 2 + a3 x n 3 + + an 1 x1 + an = 0
2
)(
+ ux + v x n 2 + b1 x n 3 + b2 x n 4 + + bn 3 x1 + bn 2 + remainder = 0
The derived polynomial follows with the terms in the brackets being the
remainder:
(x
n2
+ b1 x n 3 + b2 x n 4 + + bn 3 x1 + bn 2 + [bn 1 + bn ] = 0
(x
n4
+ c1 x n 5 + c2 x n 6 + + cn 5 x1 + cn 4 + [ cn 3 + cn 2 + cn 1 ] = 0
The solution may be set up in synthetic division form shown in Table 1.21:
Table 1.21. Bairstows method
a0
a1
a2
ub0
ub1
v
vb0
u
b0
b1
b2
uc0
uc1
vc0
c0
c1
c2
u
v
an3
an2
an1
an
ubn4
ubn3
ubn2
ubn1
vbn5
vbn4
vbn3
vbn2
bn3
bn2
bn1
bn
ucn4
ucn3
ucn2
vcn5
vcn4
vcn3
cn3
cn2
cn1
Using the preceding values, the approximations for the change in u and v
values denoted u and v are as follows:
bn cn 2
bn 1 cn 3
u =
cn 1 cn 2
cn 2 cn 3
cn 1 bn
cn 2 bn 1
and v =
cn 1 cn 2
cn 2
cn 3
u2 = u + u and v2 = v + v
Continue the process until u and v are equal to zero. The two roots are
as follows by the quadratic equation:
(x
+ ux + v
with x1, 2 =
u u 2 4v
2
1.5
3
10
10
44
48
1.5
6.75
7.125
35.813
22.969
1.5
6.75
7.125
35.813
4.5
4.75
23.875
15.313
10.781
1.5
4.125
43.125
1.5
4.125
2.75
28.75
31.938
1.5
1
1.5
1.5
6
10.781 28.750
15.313 2.750
( 10.781)( 2.750) ( 28.750)(15.313)
u =
=
31.938 28.750 ( 31.935)( 2.750) ( 28.750) ( 28.750)
28.750 2.750
469.88
914.39
31.938
28.750
v =
31.938
28.750
=
= 0.5139
10.781
15.313
( 31.9938)(15.313) ( 10.781)( 28.750)
=
28.750
( 31.935)( 2.750) ( 28.750) ( 28.750)
2.750
179.08
= 0.1958
914.39
u2 = u + u = 1.5 + 0.5139 = 2.0139
v2 = v + v = 1.5 + 0.1958 = 1.6958
Now, repeat the process using the revised values for u and v shown in
Table 1.23.
Table 1.23. Example 1.17 Bairstows method
2.0139
3
10
10
44
48
2.014
10.097
3.219
43.745
5.972
1.696
8.503
2.711
36.837
5.014
1.599
21.722
2.965
5.191
2.014
14.153
21.868
23.707
1.696
11.918
18.415
10.859
11.772
39.157
1.6958
2.0139
1.6958
1
7.028
5.191
2.965
u =
39.157
11.772
=
11.772
10.859
(5.191) (10.859) (11.772)( 2.965)
=
11.772 ( 39.157)(10.859) (11.772)(11.772)
10.859
21.45
= 0.0381
563.77
39.157
11.772
v =
39.157
11.772
=
5.191
2.965
( 39.157)( 2.965) (5.191)(11.772)
=
11.772 ( 39.157)(10.859) (11.772) (11.772)
10.859
177.22
= 0.3145
563.77
u2 = u + u = 2.0139 + 0.0381 = 1.9758
v2 = v + v = 1.6958 + 0.3145 = 2.0102
Now, repeat the process using the revised values for u and v shown in
Table 1.24.
Table 1.24. Example 1.17 Bairstows method
1.9758
3
10
10
44
48
1.976
9.831
4.305
48.027
0.698
2.010
10.002
4.380
48.863
4.976
2.179
24.308
0.353
1.561
1.976
13.735
18.861
38.373
2.010
13.974
19.189
9.546
19.421
57.208
2.0102
1.9758
2.0102
1
6.952
1.561 19.421
0.353 9.546
( 1.561) (9.546) (19.421)(0.353)
u =
=
57.208 19.421 ( 57.208)(9.546) (19.421)(19.421)
19.421 9.546
=
21.74
= 0.0236
923.27
57.208 1.561
19.421 0.353
( 57.208)(0.353) ( 1.561)(19.421)
v =
=
57.208 19.421 ( 57.208)(9.546) (19.421) (19.421)
19.421 9.546
=
10.11
= 0.0110
923.27
u2 = u + u = 1.9758 + 0.0236 = 1.999
v2 = v + v = 2.0102 0.0110 = 1.999
It appears the values are u = 2 and v = 2. Repeat the process using the
revised values for u and v shown in Table 1.25.
Table 1.25. Example 1.17 Bairstows method
2
2
3
2
10
4
10
24
10
10
2
2
5
44
48
0
48
0
48
4
0
Since the remainders of the first division bn1 and bn are both zero, u =
2 and v = 2 are the coefficients of the root quadratic. Substitute into the
quadratic equation to find the roots.
(x
+ 2 x + 2 with x1, 2 =
2 22 4 ( 2)
2
= 1 1 = 1 i
)(
f ( x) = x 2 + 2 x + 2 x 3 5 x 2 2 x + 24
The remaining polynomial may be solved using the same method. This
time begin with u = 0 and v = 0 in Table 1.26.
Table 1.26. Example 1.17 Bairstows method
0
0
0
5
0
5
0
5
2
0
0
2
0
0
2
24
0
0
24
24 5
2 1
( 24)(1) ( 5) ( 2) = 14 = 0.5185
=
u =
2 5 ( 2)(1) ( 5) ( 5) 27
5 1
2 24
5 2 ( 2) ( 2) ( 24)(5) 124
=
v =
=
= 4.5626
2 5 ( 2)(1) ( 5) ( 5) 27
5 1
u2 = u + u = 0 0.0236 = 0.5185
v2 = v + v = 0 4.5626 = 4.5626
Repeat the process using the revised values for u and v shown in
Table1.27.
Table 1.27. Example 1.17 Bairstows method
0.5185
2
24
2.324
0.139
4.593
20.582
4.481
0.269
3.558
0.519
2.055
5
0.519
4.5926
0.5185
4.5926
4.593
1
3.963
2.807
3.558 3.963
0.269
1
(3.558)(1) ( 3.963) (0.269)
u =
=
2.807 3.963 ( 2.807)(1) ( 3.963) ( 3.963)
1
3.963
=
4.623
= 0.3584
12.900
2.807 3.558
3.963 0.269
( 2.807)(0.269) (3.558) ( 3.963)
v =
=
2.807 3.963
( 2.807)(1) ( 3.963) ( 3.963)
1
3.963
=
14.844
= 1.1516
12.9007
5
2
24
0.877
3.616
0.113
5.744
23.684
4.123
0.129
0.429
0.877
2.847
5.7442
0.8769
5.7442
5.744
1
3.246
3.026
0.429 3.246
0.129
1
(0.429)(1) ( 3.246) (0.129)
u =
=
3.026 3.246 (3.026)(1) ( 3.246) ( 3.246)
1
3.246
=
0.848
= 0.1129
7.511
3.026 0.429
3.246 0.129
(3.026)(0.129) (0.429) ( 3.246)
v =
=
3.026 3.246
(3.026)(1) ( 3.246) ( 3.246)
1
3.246
=
1.783
= 0.2374
7.511
5
2
24
0.990
3.969
0.012
5.9816
0.9898
5.982
23.987
4.010
0.012
0.025
0.990
2.990
3.020
3.004
5.9816
5.982
0.025 3.020
0.012
1
(0.025)(1) ( 3.020) (0.012)
u =
=
3.004 3.020 (3.004)(1) ( 3.020) ( 3.020)
1
3.020
0.061
= 0.0100
6.116
3.004 0.025
3.020 0.012
(3.004)(0.012) (0.025) ( 3.020)
v =
=
3.004 3.020
(3.004)(1) ( 3.020) ( 3.020)
1
3.020
=
0.111
= 0.0181
6.116
u2 = u + u = 0.9898 0.0100 = 0.9998
v2 = v + v = 5.9816 0.0181 = 5.9997
It appears the values are u = 1 and v = 6. Repeat the process using the
revised values for u and v as shown in Table 1.30.
Table 1.30. Example 1.17 Bairstows method
1
5
2
4
4
6
0
6
1
24
0
24
Since the remainders of the first division bn1 and bn are both zero, u=1
and v = 6 are the coefficients of the root quadratic. Substitute them into
the quadratic equation to find the roots.
(x
1x 6 with x1, 2 =
( 1)2 4 ( 6)
2
= 0.5 2.5 = 2, 3
The first two roots are x3 = 2 and x4 = 3. The remaining values are the
coefficients of the factored polynomial.
)(
f ( x) = x 2 + 2 x + 2 x 2 1x 6 ( x 4)
The last root is x5 = 4, which is the value in the remaining polynomial of
degree one x r = x 4. The final factored form of the original quadratic
can be written and the five roots are x1 = 1+i, x2 = 1i, x3 = 2, x4 = 3,
and x5 = 4.
f ( x) = x 2 + 2 x + 2 ( x + 2) ( x 3) ( x 4)
REFERENCES
Bairstow, L. 1920. Applied Aerodynamics. London, England: Longmans, Green
and Co.
Descartes, R. 1637. Discours de la mthode pour bien conduire sa raison, et
chercher la vrit dans les sciences. Lieden, Netherlands.
James, M.L., G.M. Smith, and J.C. Wolford. 1977. Applied Numerical Methods
for Digital Computations. 2nd ed. Boston, Massachusetts: AddisonWesley
Longman Publishing Co., Inc.
Newton, I. 1669. De analysi per aequationes numero terminorum infinitas,
London, England.
Taylor, B. 1715. Methodus Incrementorum Directa et Inversa. London, England.
CHAPTER 2
Solutions of Simultaneous
Linear Algebraic Equations
Using Matrix Algebra
Matrix algebra is commonly utilized in structural analysis as a method of
solving simultaneous equations. Each motion, translation or rotation, at
each discrete location in a structure is normally the desired variable. This
chapter explores matrix terminology, matrix algebra, and various methods
of linear algebra to determine solutions to simultaneous equations.
an1 x1 + an 2 x2 + an 3 x3 + + ann xn = Cn
These equations can be written in matrix form, [A][x]=[C] as follows:
a11
a
21
a31
an1
a12
a22
a32
an 2
a13
a23
a33
an 3
a1n x1 C1
a2 n x2 C2
a3n x3 = C3
ann xn Cn
2.2MATRICES
A matrix can be defined as a rectangular array of symbols or numerical
quantities arranged in rows and columns. This array is enclosed in brackets and if there are n rows and m columns, the general form of this matrix
is expressed by the following:
a11
a
21
A
=
[ ] a31
an1
a12
a22
a32
an 2
a13
a23
a33
an 3
a1m
a2 m
a3m
anm
[ A] = [ a11
a12
a13 a1m ]
an1
a12
a22
a32
an 2
a13
a23
a33
an 3
a1n
a2 n
a3n
ann
Square matrices are unique because they are the only matrix that has a
reciprocal or invert as described later in this section. Several types of
square matrices exist such as the diagonal matrix, the identity matrix, the
triangular matrix, and the invert matrix.
2.2.3 DIAGONAL MATRIX
A diagonal matrix is defined as a matrix where all elements outside of the
principal diagonal are equal to zero. The diagonal running from the upper
left corner of the array to the lower right corner is considered the principal
diagonal.
a11
0
[ A] = 0
0
0
a22
0
0
0
0
a33
0
0
0
0
ann
[ I ] = 0
0
0 0
1 0
0 1
0 0
0
0
0
1
[U ] = 0
0
a11
a
21
[ L ] = a31
an1
a12
a22
0
0
0
a22
a32
an 2
a13
a23
a33
0
0
0
a33
an 3
a1n
a2 n
a3n
ann
0
0
0
ann
[ I ] = [ A][ A]1
2.2.7 MATRIX MINOR
The matrix minor, [Aij], is found by omitting the ith row and the jth column
of a matrix and writing the remaining terms in a matrix of one size smaller
in rows and columns. It is used in the computation of the determinant. For
example, the minor, [A22] is shown in the following. Note that i = 2 and j = 2:
a11
a
21
[ A22 ] = a31
an1
a12
a22
a32
a13
a23
a33
an 2
an 3
a1m
a
a2 m 11
a
a3m = 31
a
anm n1
a13
a33
an 3
a1m
a3m
anm
an1
[ A]
a11
a
12
= a13
a1m
a12
a22
a32
an 2
a21
a22
a23
a2 m
a13
a23
a33
an 3
a31
a32
a33
a3m
a1m
a2 m
a3m
anm
an1
an 2
an 3
amn
[ A ] + [ B ] = [C ]
a11 + b11 = c11
a12 + b12 = c12
aij + bij = cij
Matrix subtraction follows the same form as addition where the matrices
are of the same size.
[ A ] [ B ] = [C ]
a11 b11 = c11
a12 b12 = c12
aij bij = cij
[ A] = 7 9 3
6 5 1
and
1 2 4
[ B ] = 6 5 2
4 3 0
Addition:
2 4 6 1 2 4 3 6 10
[ A] + [ B ] = 7 9 3 + 6 5 2 = 13 14 5
6 5 1 4 3 0 100 8 1
Subtraction:
2 4 6 1 2 4 1 2 2
A
B
=
[ ] [ ] 7 9 3 6 5 2 = 1 4 1
6 5 1 4 3 0 2 2 1
an1
a12
a22
a32
an 2
ca12
ca22
ca32
can 2
ca13 ca1m
ca23 ca2 m
ca33 ca3m
can 3 canm
[ A]n m [ B ]m o = [C ]n o
Each term of the product matrix (row i and column j) is obtained by
multiplying each term in row i of the first matrix by the term in row j of
the second matrix and then summing these products.
cij = ai1 b1 j + ai 2 b2 j + + ain bnj
4
2
0
and
1 4 2 3 2 4
2 3 2 4 1 1
[ B] =
[ A]4 2 [ B ]2 6 = [C ]4 6
2
3
[ A][ B ] = 1
4
10 20 12 22 8 12
7 18 10 17 8 14
2 1 4 2 3 2 4
=
C
=
[
]
1 4 2 3 2 4
0 2 3 2 4 1 1
4
11 24 14 25 10 16
c12 = 2 ( 4) + 4 (3) = 8 + 12 = 20
2.3.4 MATRIX DETERMINANTS
A determinant is only defined for square matrices and can be easily
achieved through expansion by minors of a row or a column when dealing
with small matrices. A row expansion is as follows:
A = a11 A11 a12 A12 + (1)1+ j a1 j A1 j + (1) n +1 a1n A1n
a11
a21
an1
a12
a22
an 2
a1n
a22
a2 n
a32
= a11
anm
an 2
+ + (1)1+ j a1 j
a23 a2 n
a33 a3n
a12
an 3 anm
a21
a31
A1 j + (1) n +1 a1n
an1
a21
a31
an1
a22 a2, n 1
a32 a3, n 1
an 2 an , n 1
a23 a2 n
a33 a3n
an 3 anm
If A is the determinant of the matrix [A], then the following equations are
valid when row k and column k are expanded:
n
and
j =1
a12
a22
a32
a13
a11
a23 = a21
a33
a31
a12
a22
a32
a13
a23
a33
a11
a21
a31
a12
a22
a32
a11
a21
a31
a12
a22
a32
a13
a11
a23 = a21
a31
a33
a12
a22
a32
a13
a23
a33
a11
a21
a31
a12
a22
a32
(a11a22 a33 + a12 a23 a31 + a13 a21a32 ) (a13 a22 a31 + a11a23 a32 + a12 a21a33 )
Also, a determinant may be found by the product of the diagonal of any
triangular matrix.
Example 2.3 Matrix determinants
Find the determinant of the following matrix, [A], by expansion of minors
and by the basket weave method.
3 4 6
[ A] = 1 2 3
2 1 2
Expansion of row 1 yields:
A =3
2 3
1 3
1 2
4
+6
1 2
2 2
2 1
A = 3 ( 4 3) 4 ( 2 6) + 6 (1 4)
A = 3 + 16 18 = 1
2.4CRAMERS RULE
Two common methods to solve simultaneous equations exist. One is the
elimination of unknowns by elementary row operations and the second
involves the use of determinates. One of the methods involving determinates is known as Cramers rule. This method was published in 1750 by
Gabriel Cramer (1750). The procedure for Cramers rule in the solution to
n linear equations with n unknowns is as follows:
x1 =
A1
A
c1
c2
A1 = c3
a12
a22
a32
a13
a23
a33
cn
an 2
an 3
a11
a21
A3 = a31
an1
a12
a22
a32
an 2
, x2 =
A2
A
, x3 =
A3
A
, , xn =
a1n
a11
a2 n
a21
a3n , A2 = a31
ann
an1
c1
c2
c3
cn
c1 a1n
a11
c2 a2 n
a21
c3 a3n , , An = a31
cn ann
an1
a13
a23
a33
an 3
a12
a22
a32
an 2
An
A
a13
a23
a33
an 3
a1n
a2 n
a3n ,
ann
c1
c2
c3
cn
As you might see, A1 is the original coefficient matrix, [A], with columnone replaced with the constant column matrix, [c]. The solution to
n simultaneous equations by Cramers rule requires (n1)*(n+1)! multiplications. In other words, the solution of ten simultaneous equations by
determinants would require (9)*(11!) =359,251,200 multiplications.
A1
A
A2
A
A3
A
180
=5
36
36
=1
36
72
= 2
36
2
1
8
6
1
2
1
2

1
1
2
22
32
14
13
8
6
A1
1
A1
14
1
2
180
14
1
2
1
2
42
36
13
72
14
13
13
1
13
1
13
2
1
2
A2
62
56
A3
2
1
2
8
6
1
14
13
5
86
13
2
1
1
2
86
A3
36
A2
248
26
154
A3
1
2
1
2
1
A1
A2
13
1
168
182
14
1
of a matrix, Cij, is given in the following equation where Aij is the determinant of the minor as defined in Section 2.2.
Cij = ( 1)
i+ j
Aij
Therefore, given the matrix [A], the cofactor matrix is shown as follows.
The matrix of cofactors should not be confused with the constant matrix
of the original linear algebraic equation, although they have the same variable, [C]:
a11
a
21
A
=
[ ] a31
an1
a12
a13
a23
a33
an 3
a22
a32
an 2
[C ] = [ A]cofactor
C11 C12
C
21 C22
= C31 C32
Cn1 Cn 2
a1m
a2 m
a3m
anm
C13
C23
C33
Cn 3
C1m
C2 m
C3m
Cnm
Once the cofactor matrix is known, the invert can be easily calculated:
T
C]
[
=
A
1
The solutions to the simultaneous equations are now found from matrix
multiplication.
[ A][ x] = [C ] Or [ A1 ][C ] = [ x]
Adj [ A]
A
The adjoint matrix, Adj[A], is simply the transpose of the cofactor matrix.
This can be expressed in a few ways.
Adj[ A] = [C ]T = [ Acofactor ]T
c11
c
21
Adj [ A] = c31
cn1
c12
c22
c32
cn 2
Or ci j = (1)( i + j ) [ Aj i ]
T
c13 c1m
c11
c
c23 c2 m
12
c33 c3m = c13
c1m
cn 3 cnm
c21
c31
c22
c23
c2 m
c32
c33
c3m
cn1
cn 2
cn 3
cmn
It is noted that the subscripts of the adjoint matrix are the reverse of the
cofactor matrix. The main difference is that the transpose is performed
during the operation of taking the adjoint, while in the cofactor method is
done at the end.
Example 2.5 Cofactor method
Find the solution set to the following nonhomogeneous linear algebraic
equations using the cofactor method.
x1 + x2 + x3 + x4 = 10
8 x1 + 4 x2 + 2 x3 + x4 = 26
x1 + x2 x3 + x4 = 2
3 x1 + 2 x2 + 1x3 + 0 x4 = 10
[C ]
A =
1
And Cij = ( 1)
i+ j
Aij
The determinants are shown by row expansion for the 4 4 matrix and by
the basket weave for all the 3 3 matrices in Table 2.2.
The last step in Table 2.2 is to multiply the invert of A, [A]1, times the
constant vector, [C], to get the final solution vector, [x].
Example 2.6 Method of adjoints
Determine the solution to the following set of equations using the adjoint
method. Use the basket weave method for determinants.
A1
c41=
c31=
c21=
2
1
1
4
1
2
c11=
c12=
c31 = 1
0.2500
0.0000
0.7500
0.5000
3
4
1
6
1
2
1
0
4
4
12
c41 = 6
1
4
1
0.3333
0.3333
0.3333
0.3333
9
4
5
6
1
1
1
c32=
c42=
6
4
2
12
0.0833
0.3333
0.4167
0.1667
c42 = 12
(1)4+1
c32 = 4
(1)
c22 = 4
3+1
1
1
0
1
4
2
1
2
1
c21 = 4
c22=
c12 = 0
(1)
9
2+1
1
1
0
1+1
(1)
c11 = 3
1
1
0
1
1
0
1
1
1
0
1
1
2
1
1
1
2
1
1
4
1
2
1
2
1
1
1
4
1
2
1
8
1
3
CT
1
8
1
1
8
3
1
1
0
2
1
1
10
26
2
10
1
1
1
1
1
0
1
1
0
0.5000
1.0000
0.5000
1.0000
3
0
9
6
c43 = 6
1
2
1
c33 = 5
1
2
1
c23 = 4
1
1
1
c13 = 9
2
1
1
8
1
3
1
1
3
6
8
1
3
4
4
4
4
(1)4+2
(1)
3+2
(1)
1+2
2+2
(1)
1
1
0
8
1
3
12
1
1
3
1
8
3
1
8
1
1
4
5
2
6
12
6
12
c44 = 12
c43=
c34 = 2
c33=
c24 = 4
c23=
c14 = 6
c13=
1
2
3
4
1
4
1
1
4
2
1
1
2
4
1
2
8
1
3
1
1
1
1
1
0
1
1
0
1
1
0
4
1
2
(1)4+3
(1)3+3
(1)2+3
(1)1+3
1
1
0
c44=
c34=
c24=
c14=
1
8
1
1
8
3
1
1
3
8
1
3
1
4
1
1
4
2
1
1
2
4
1
2
8
1
3
1
2
1
1
2
1
1
1
1
2
1
1
4
1
2
(1)4+4
(1)3+4
(1)2+4
(1)1+4
2
1
1
2
1
0
4
1
0
c11=
A1
c14=
c13=
c12=
6
0.5000
0.5000
1.0000
0.0000
6
6
12
0
c14 = 6
1
4
1
1
2
0
2
0
2
0
1
2
1
c13 = 2
1
4
0
c12 = 2
1
1
0
c11 = 6
1
1
0
0.1667
0.0000
0.1667
0.0000
2
6
4
0
1
1
1
1
1
1
1
1
1
1
1
1
2
1
0
4
1
0
1
2
1
0
1
4
1
0
1
8
1
0
c21=
c22=
c23=
c24=
6
12
6
12
0.1667
0.5000
0.3333
0.0000
c24 = 12
(1)4+1
c23 = 6
(1)3+1
c22 = 0
(1)2+1
12
c21 = 6
(1)1+1
1
1
1
1
1
1
1
1
8
1
1
8
0
1
1
0
8
1
0
1
1
1
2
1
0
10
26
2
4
1
1
1
1
1
1
1
1
1
0.5000
1.0000
0.5000
1.0000
c34 = 6
1
2
1
c33 = 4
1
2
0
c32 = 2
1
1
0
c31 = 12
2
1
0
8
1
0
(1)4+2
(1)3+2
(1)2+2
(1)1+2
1
1
1
8
1
0
1
8
1
c44 = 12
c34=
c43 = 0
c33=
1
8
0
1
1
0
c42 = 0
c32=
1
12
c41 = 0
c31=
1
2
3
4
1
4
1
1
4
0
1
1
0
4
1
0
8
1
0
1
1
1
1
1
1
1
1
1
1
1
1
4
1
0
(1)4+3
(1)3+3
(1)2+3
(1)1+3
1
1
1
c44=
c43=
c42=
c41=
1
8
1
1
8
0
1
1
0
8
1
0
1
4
1
1
4
0
1
1
0
4
1
0
8
1
0
1
2
1
1
2
0
1
1
0
2
1
0
4
1
0
(1)4+4
(1)3+4
(1)2+3
(1)1+4
2
1
0
x1 + x2 + x3 + x4 = 10
8 x1 + 4 x2 + 2 x3 + x4 = 26
x1 + x2 x3 + x4 = 2
0 x1 + 0 x2 + 0 x3 + x4 = 4
[C ]T
[A ] =
The determinants are shown by row expansion for the 4 4 matrix and by
the basket weave for all the 3 3 matrices in Table 2.3.
The last step in Table 2.3 is to multiply the invert of A, [A]1, by the
constant vector, [C], to get the final solution vector, [x].
akjk1
a
k 1
kk
(a )
k 1
ik
k + 1 j m
where
k +1 i n
xn =
anm
ann
aim j = i +1aij x j
n
xi =
aii
where i = n 1, n 2, ,1
AC
AC
AC
1
8
1
0
1
4
1
0
1
2
1
0
1
1
1
1
10
26
2
4
1
0
0
0
1
4
2
0
1
6
0
0
1
7
2
1
10
54
12
4
8
1
0
1
0
0
0
1
4
0
0
1
6
3
0
1
7
1.5
1
10
54
15
4
0.5
0
1
0
0
0
1
4
0
0
1
6
3
0
1
7
1.5
1
10
54
15
4
1
2
3
4
10
26
10
26
Swap
10
Pivot
10
26
0.5
0.75
0.88
6.75
1.13
5.25
0.125
0.25
0.375
1.5
0.5
0.8
0.25
2
0.4
0.125
1.5
0.5
0.75
0.5
0.25
26
1.5
0.8
1.13
5.25
0.5
0.333
0.5
0.8
1.5
0.333
26
1.5
0.8
1.13
5.25
0.5
0.500
12
* (1)2
0.8
26
1.13
5.25
Swap
0.88
6.75
Pivot
0.4
0.25
1
x
2
3
4
(2.1a)
2 x1 + 3 x2 + 4 x3 = 20
(2.1b)
3 x1 x2 + 3 x3 = 10
(2.1c)
The first step is to divide the first equation of the set by the coefficient of
the first unknown in that equation, 2. Equation 2.2a is then multiplied by
the corresponding coefficient of that unknown of Equations 2.1b and 2.1c
to give the following:
x1 x2 +
5
13
x3 =
2
2
2 x1 2 x2 + 5 x3 = 13
3 x1 3 x2 +
15
39
x3 =
2
2
(2.2a)
(2.2b)
(2.2c)
(2.3a)
19
9
x3 =
2
2
(2.3b)
5
13
x3 =
2
2
(2.3c)
x1 x2 +
(2.4a)
14
2
x3 =
5
5
(2.4b)
1
7
x2 + x3 =
5
5
(2.4c)
2 x2
Just as the previous cycle, Equation 2.4b is subtracted from 2.3b and
becomes 2.5a. Equation 2.4c is subtracted from 2.3c and becomes 2.5b.
Equation 2.4a now becomes 2.5c.
123
41
x3 =
10
10
(2.5a)
23
79
x3 =
10
10
(2.5b)
x1 +
1
7
x2 x3 =
5
5
(2.5c)
(2.6a)
23
69
x3 =
10
10
(2.6b)
3
1
x3 =
5
5
(2.6c)
Equation 2.6b is subtracted from 2.5b and becomes 2.7a. Equation 2.6c is
subtracted from 2.5c and becomes 2.7b. Equation 2.6a now becomes 2.7c.
All the unknowns are found with the following solution:
x1 = 1
(2.7a)
x2 = 2
(2.7b)
x3 = 3
(2.7c)
The same result may be obtained by working with just the coefficients
and constants of the equations. Given the same equations, the following
augmented matrix is valid:
2 x1 2 x2 + 5 x3 = 13
2 x1 + 3 x2 + 4 x3 = 20
3 x1 x2 + 3 x3 = 10
2 2 5 13
[ A] = 2 3 4 20
3 11 3 10
An augmented matrix, [B], is established from the following algorithm:
bi 1, j 1 = aij
a1 j ai1
a11
1< i n
where 1 < j m
a 0
11
bn , j 1 =
a1 j
a11
1 < j m
where
a11 0
where,
a
b
i
j
For example,
b2, 2 = a2, 2
a1, 2 a2,1
a11
= 3
2 ( 2)
=5
2
5 1
9
[ B ] = 2 2
5
1
19
2
13
23
79
[C ] = 10 10
7
1
5
5
Finally, one more cycle reduces column 3 and obtains the solution:
1
D
=
[ ] 2
3
It is known that applying these equations to a number of simultaneous
equations, n, will produce the same number of matrices, n, to achieve the
solution.
akj =
akk
k + 1 j n + 1
where,
a
a
i
j
k
original elements
new elements
row (n)
column (m)
pivotal row number
In other words, normalize the matrix then utilize partial pivoting to reduce
the matrix. However, there is no need to reduce the elements under the
pivot. Reducing up and down is easier without a need to reorder the rows.
This is how Example 2.10 is performed.
1
1
16 8
A C = 1 1
81 27
16 8
1 1
4 2
1 1
9 3
4 2
1 8
1 44
1 2
1 44
1 8
From partial pivoting, the first row can be swapped with the fourth row to
form the following matrix:
81 27
16 8
A C = 1 1
1
1
16 8
9 3 1 44
4 2 1 44
1 1 1 2
1 1 1 8
4 2 1 8
Now elimination may be performed as shown in Table 2.6. Note that for
each column reduction, elements are reduced to zero below and above
the pivot position. Once it reduces to a diagonal matrix, the solution is
foundby dividing each row by the pivot element. The determinant is
found as A = (81)(13.3333)(1)(1.3333) (2) = 2880.
a12
a22
a32
a13 x1 c1
a23 x2 = c2
a33 x3 c3
If [A] represents the coefficient matrix, [x] represents the column matrix of
the unknowns, and [C] represents the column matrix of the constants, the
previous can be expressed as the following equation:
81
16
1
1
16
27
8
1
1
8
9
4
1
1
4
3
2
1
1
2
1
1
1
1
1
44
44
2
8
8
81
0
0
0
0
27
13.3333
0.6667
1.3333
2.6667
9
2.2222
0.8889
0.8889
2.2222
3
2.5926
0.9630
1.0370
1.4074
1
0.8025
0.9877
0.9877
0.8025
44
35.3086
1.4568
7.4568
0.6914
81
0
0
0
0
0
13.5
13.3333 2.2222
0
1
0
0.6667
0
2.6667
2.25
2.5926
0.8333
0.7778
0.8889
2.625
0.8025
1.0278
0.9074
0.9630
115.5
35.3086
3.2222
3.9259
6.3704
0
13.3333
0
0
0
0
0
1
0
0
13.5
4.4444
0.8333
1.3333
1.3333
11.25
1.4815
1.0278
0.2222
1.7778
72
28.1481
3.2222
1.7778
2.2222
13.5000
2.2222
81
0
0
0
0
81
0
0
0
0
0
13.3333
0
0
0
0
0
1
0
0
0
0
0
1.33333
0
13.5
2.2222
1.1667
0.2222
2
54
22.2222
4.3333
1.7778
4
10.1250
3.3333
0.6250
81
0
0
0
0
0
13.3333
0
0
0
0
0
1
0
0
0
0
0
1.33333
0
0
0
0
0
2
1
2
2
1
2
2880.00
0.1975
0.0123
0.0123
0.1975
2.0250
0.0500
0.1000
0.2000
0.6667
2.6667
1.0000
81
6.7500
26.6667 1.1111
2
0.5833
1.33333 0.1111
4
[ A][ x ] = [C ]
or
[ A][ x ] [C ] = 0
0 0
u13 x1 d1
u23 x2 = d 2
1 x3 d3
[U ][ x ] [ D ] = 0
Also, a lower triangular matrix exists, such that, when the first set is
premultiplied by [L], the result is the second set as follows:
l11 0
l
21 l22
l31 l32
0
0
l33
[ L ]([U ][ x ] [ D ]) = [ A][ x ] [C ]
[ L ][U ] = [ A]
and
[ L ][ D ] = [C ]
0 1 u12
0 0 1
l33 0 0
u13
u23
1
d1 a11
d 2 = a21
d3 a31
a12
a22
a32
a13
a23
a33
c1
c2
c3
j =1
a1 j
j = 2, 3, 4, , n + 1
u1 j =
For
i =1
a11
j = 2, 3, 4, , n
For i = j , j + 1, j + 2, , n
( for each value of j )
j 1
ui , j =
lii
i = 2, 3, 4, , n
For j = i, i + 1, i + 2, , n + 1
( for each value of i )
xn = un , n +1
xi = ui , n +1
j = i +1
ij
xj
For {i = n 1, n 2, n 3, , 1}
a1 j
a11
For
{ j = 2, 3, 4, , n + 1}
j 1
j = 2, 3, 4, , n
For i = j , j + 1, j + 2, , n
( for each value of j )
i = 2, 3, 4, , n
For j = i, i + 1, i + 2, , n + 1
( for each value of i )
ai , j =
aii
l11 0
l
21 l22
l31 l32
0 1 u12
0 0 1
l33 0 0
u13
u23
1
d1 a11
d 2 = a21
d3 a31
a12
a22
a32
a13
a23
a33
c1
c2
c3
16
1
44
1
81
27
3
44
16
8
2
16
8
2
81
108
90
60
16
24
24
12
2
The upper triangular matrix is found at the same time and is shown in
Table 2.9.
Table 2.9. Example 2.11 Cholesky decomposition method
UD
1.5
1.75
1.875
10.5
0.5
1.25
0.167
1.333
Finally, the solution is calculated from the [UD] matrix using back substitution and shown in Table 2.10.
Table 2.10. Example 2.11 Cholesky decomposition method
1
2
x
2
1
2
an1 x1 + an 2 x2 + an 3 x3 + + ann xn = Cn Cn = e
This shows that the corrections, xs, can be obtained by replacing the
constant vector of the solution with the difference of the constant vectors,
(CCs)s, and applying reduction to find the error. These are then added
to the approximate solution and the process is repeated until accuracy is
achieved.
Example 2.12 Error equations
Determine the solution to the following set of equations using any
GaussJordan elimination, but only carry two decimals of accuracy
(i.e., x.xx) then apply error equations to increase accuracy.
2.11x1 + 2.11x2 3.04 x3 + 1.11x4 = 1.65
0.02 x1 + 1.23 x2 + 2.22 x3 + 1.02 x4 = 13.18
0.14 x1 0.06 x2 + 1.21x3 1.08 x4 = 0.67
1.32 x1 + 0.20 x2 + 0.000 x3 + 3.90 x4 = 17.32
The process of three complete cycles is shown in Tables 2.112.13.
an1
a12
a22
a32
an 2
a13 a1n
a23 a2 n
a33 a3n
an 3 ann
1 0 0
0 1 0
0 0 1
0 0 0
0
0
0
1
From here, the coefficient matrix is reduced until the identity matrix is
on the left and the original identity on the right becomes the invert of A.
AC
AC
AC
AC
0.02
3.04
2.22
1.21
0.00
1.11
1.02
1.65
13.18
1.08
0.67
3.90
17.32
1.65
13.20
0.78
16.29
0.14
1.32
0.06
2.11
0.00
0.00
0.00
2.11
1.25
0.20
1.12
3.04
2.19
1.41
1.90
1.11
1.03
1.15
3.21
2.11
0.00
0.00
1.25
6.74
0.63
20.63
2.19
1.03
13.20
0.00
0.00
0.00
0.00
1.76
3.86
0.99
4.13
1.33
28.12
2.11
0.00
0.00
0.00
0.00
1.25
0.00
0.00
0.00
0.00
1.76
0.00
4.42
15.54
2.26
6.30
11.55
1.33
25.20
2.11
0.00
0.00
0.00
0.00
1.25
0.00
0.00
0.00
0.00
1.76
0.00
0.00
0.00
0.00
6.30
2.14
2.51
5.29
25.20
1.01
2.01
3.01
4.00
0.01
e1
2.11
1.23
0.03
0.01
0.02
0.20
C1
0.99
1.66
13.21
0.66
17.34
AC
AC
AC
AC
2.11
2.11
3.04
1.11
0.01
0.02
1.23
2.22
1.02
0.03
0.14
0.06
1.21
1.08
0.01
1.32
0.20
0.00
3.90
0.02
2.11
2.11
3.04
1.11
0.01
0.00
1.25
2.19
1.03
0.03
0.00
0.20
1.41
1.15
0.01
0.00
1.12
1.90
3.21
0.01
2.11
0.00
6.74
0.63
0.04
0.00
1.25
2.19
1.03
0.03
0.00
0.00
1.76
0.99
0.01
0.00
0.00
3.86
4.13
0.04
2.11
0.00
0.00
4.42
0.00
0.00
1.25
0.00
2.26
0.02
0.00
0.00
1.76
0.99
0.01
0.00
0.00
0.00
6.30
0.02
2.11
0.00
0.00
0.00
0.01
0.00
1.25
0.00
0.00
0.01
0.00
0.00
1.76
0.00
0.01
0.00
0.00
0.00
6.30
0.02
0.00
x
0.01
1.01
Dx1
0.01
3.00
0.00
4.00
1.67
C2
2.00
13.18
0.02
e2
0.00
0.67
0.00
17.33
0.01
AC
AC
AC
AC
2.11
2.11
3.04
1.11
0.02
0.02
1.23
2.22
1.02
0.00
0.14
0.06
1.21
1.08
0.00
1.32
0.20
0.00
3.90
0.01
2.11
2.11
3.04
1.11
0.02
0.00
1.25
2.19
1.03
0.00
0.00
0.20
1.41
1.15
0.00
0.00
1.12
1.90
3.21
0.00
2.11
0.00
6.74
0.63
0.02
0.00
1.25
2.19
1.03
0.00
0.00
0.00
1.76
0.99
0.00
0.00
0.00
3.86
4.13
0.00
2.11
0.00
0.00
4.42
0.02
0.00
1.25
0.00
2.26
0.00
0.00
0.00
1.76
0.99
0.00
0.00
0.00
0.00
6.30
0.00
2.11
0.00
0.00
0.00
0.02
0.00
1.25
0.00
0.00
0.00
0.00
0.00
1.76
0.00
0.00
0.00
0.00
0.00
6.30
0.00
0.01
x
0.00
1.00
Dx2
0.00
3.00
0.00
4.00
0.00
1.67
C3
2.00
13.18
e3
0.00
0.67
0.00
17.32
0.00
0
1
I  A = 0
0
0
1
0
0
0
0
1
0
0 a111
0 a211
1
0 a31
1 an11
a121
1
a22
1
a32
an21
a131
1
a23
1
a33
an31
a1n1
a2n1
a3n1
1
ann
2.11
1.23
AC = 0.02
0.14
1.32
0.06
0.20
2.11
2.11
AI = 0.02
0.14
1.32
0.06
0.20
1.23
3.04
2.22
1.21
0.00
3.04
2.22
1.21
0.00
1.11
1.02
1.65
13.18
1.08
3.90
0.67
17.32
1.11
1.00
0.00
0.00
0.00
1.02
0.00
0.00
0.00
1.00
0.00
0.00
0.00
1.00
0.00
0.00
0.00
1.00
1.08
3.90
(Continued)
Table 2.14.(Continued)
AI
AI
2.110
0.000
0.000
2.110
1.250
0.200
3.040
2.191
1.412
0.000
1.120
1.902
2.1100
0.0000
0.0000
0.0000
1.2500
0.0000
0.0000
0.0000
1.110
1.031
1.000
0.009
0.066
0.000
1.000
0.000
0.000
0.000
1.000
0.000
0.000
0.000
1.154
3.206
0.626
0.000
0.000
1.000
0.0000
0.0000
0.0000
1.0000
Note that partial pivoting should be performed and row three is now
swapped with row four. Columns 3 and 4 are then eliminated as shown in
Table 2.15.
Table 2.15. Example 2.13 Matrix inversion method
AI
AI
AI
0.9840 1.6880
0.0000
0.0000
0.0000 1.2500
2.1912
1.0305
0.0095
1.0000
0.0000
0.0000
0.0000 0.0000
3.8651
4.1289 0.6171
0.8960
0.0000
1.0000
0.0000 0.0000
0.1600
1.0000
0.0000
2.1100 0.0000
0.0000
0.0000
1.7435
0.0000 1.2500
0.0000 1.3102
0.0000 0.0000
3.8651
0.0000 0.0000
0.0000 2.8714
0.2165 0.2485
1.0000 0.4560
2.1100 0.0000
0.0000
0.0000
0.4035 0.6944
2.2878
0.0000 1.2500
0.0000
0.0000
0.2605
0.0000 0.0000
3.8651
0.0000 0.3057
0.0000 0.0000
0.0000 2.8714
0.3593
0.4920
0.0000 0.5669
4.1289 0.6171
0.8960
0.0000
1.0000
0.7003
0.2165 0.2485
1.4380
0.3444
1.0000 0.4560
Now the third and fourth columns are swapped back for the inverse
matrix shown in Table 2.16.
Table 2.16. Example 2.13 Matrix inversion method
IA1
1.0843
0.3319
1.0000
0.0000
0.0000
0.0000
0.1912 0.3291
0.0000
1.0000
0.0000
0.0000
0.2084
0.0000
0.0000
1.0000
0.0000
0.0791
0.1394
0.3720
0.0891
0.0000
0.0000
0.0000
1.0000
0.0754
0.0866 0.3483
0.1588
(Continued)
Table 2.16.(Continued)
0.1912 0.3291
A
1
0.2084
0.3319
1.0843
0.0791
0.1394
0.0891
0.0754
0.0866
0.1588 0.3483
0.3720
Now the third and fourth rows are swapped in coefficient matrix and
the solution is found as [x]=[A]1[C] in Table 2.17:
Table 2.17. Example 2.13 Matrix inversion method
1.65
C
13.18
1.65
C
13.18
1
x
0.67
17.32
17.32
0.67
10 x1 + x2 + 2 x3 + x4 = 50
2 x1 + 10 x2 + x3 + 2 x4 = 63
x1 + 2 x2 + 10 x3 + x4 = 67
2 x1 + x2 + x3 + 10 x4 = 75
Note that this is a diagonal system with 10s on the diagonal and all other
coefficients are much less. Begin the iteration by setting x1 = x2 = x3 = x4 = 0
and solving for each of the unknowns using the corresponding equation in
a top down order.
10 x1 + 0 + 2 ( 0) + 0 = 50 x1 = 5.000
2 (5.000) + 10 x2 + 0 + 2 ( 0) = 63 x2 = 5.300
After completing the first cycle, start with the first equation using the new
values and find a closer approximation for each unknown. Also, check the
difference between the new values and the previous values to determine if
the desired accuracy is achieved.
10 x1 + 5.300 + 2 ( 5.140 ) + 5.456 = 50 x1 = 2.896 and x1 = 2.104
None of the values of Dx are less than e = 0.01, so the process is repeated.
Table 2.18 shows the entire process to convergence. The process can be
stopped when each value has changed less than e or when a cycle results
in each value changing less than e.
an1 x1 + an 2 x2 + an 3 x3 + + ann xn = 0
This can be written in matrix form, [A][x]=[0], as follows:
5.000
5.300
5.140
5.456
x1=
x2=
x3=
x4=
5.000
5.300
5.140
5.456
x1=
x2=
x3=
x4=
2.896
4.116
5.042
6.005
x1=
x2=
x3=
x4=
2.104
1.184
0.098
0.549
x1=
x2=
x3=
x4=
2.980
3.999
5.002
6.004
x1=
x2=
x3=
x4=
0.083
0.117
0.040
0.001
<epsilon
x1=
x2=
x3=
x4=
2.999
3.999
5.000
6.000
x1=
x2=
x3=
x4=
0.020
0.000
0.002
0.004
<epsilon
<epsilon
<epsilon
x1=
x2=
x3=
x4=
3.000
4.000
5.000
6.000
x1=
x2=
x3=
x4=
0.001
0.001
0.000
0.000
<epsilon
<epsilon
<epsilon
<epsilon
a11
a
21
a31
an1
a12
a22
a32
an 2
a13
a23
a33
an 3
a1n x1 0
a2 n x2 0
a3n x3 = 0
0
ann xn 0
a21
a31
a
n1
a12
(a22 l )
a32
an 2
a13
a23
(a33 l )
an 3
a1n x1 0
a2 n x2 0
a3n x3 = 0
0
(ann l ) xn 0
(0 l )
10
2
2
( 1 l )
4
3
2
=0
(7 l )
(0 l )
2
( 1 l )
4
10
2
3
2
(7 l )
(0 l )
10
2
2
( 1 l )
4
l 6l 2 + 11l 6 = 0
3
The solution to the cubic equation can be found by many of the methods
from Chapter 1 and represent the eigenvalues l = 1, 2, and 3.
2.14FADDEEVLEVERRIER METHOD
The FaddeevLeverrier method is a polynomial method used to find the
eigenvalues. The method is named for Dmitrii Konstantinovich Faddeev
and published by Urbain Jean Joseph Le Verrier in 1840 (Le Verrier
1839). From linear algebra, the trace of a matrix is the sum of the diagonal terms. The process for determining the characteristic polynomial is
as follows:
[ B ] = [ A]([ B ] p [ I ])
2
p2 =
[ B ] = [ A]([ B ] p [ I ])
3
pk =
1
tr [ Bk ]
k
pn =
1
tr [ Bn ]
n
[ B ] = [ A]([ B ] p [ I ])
k
k 1
k 1
[ B ] = [ A]([ B ] p [ I ])
n
n 1
n 1
Example 2.16 shows the process of determining the characteristic polynomial and the eigenvalues by the FaddeevLeverrier method.
Example 2.16 FaddeevLeverrier method
Determine the eigenvalues for the following set of equations using the
FaddeevLeverrier method.
x1 + 2 x2 + 3 x3 = 0
10 x1 + 0 x2 + 2 x3 = 0
2 x1 + 4 x2 + 8 x3 = 0
The matrix operations are shown in Table 2.19.
The characteristic polynomial is found from the trace values.
1
10
2
2
0
4
3
2
8
[B2]=([A]([B1]p1I)
1
10
2
2
0
4
3
2
8
1
10
2
2
0
4
3
2
8
[B3]=([A]([B2]p2I)
1
10
2
2
0
4
3
2
8
1
10
2
2
0
4
3
2
8
1
10
2
2
0
4
3
2
8
p1=tr[B1] = 9
1
10
2
2
0
4
3
2
8
8
10
2
2
9
4
3
2
1
9
9
9
34 4
4
76 12 32
40 8 6
p2=(1/2)tr[B2] = 26
34 4
4
76 12 32
40 8 6
4
32
20
8
76
40
4
14
8
26
26
26
24
0
0
0
24
0
0
0
24
p3=(1/3)tr[B3] = 24
[ B ][ M ][ x ] = l [ x ]
The values of the matrices are [K] for stiffness, [M] for mass, and [B] for
flexibility.
Example 2.17 Power method
Determine the first mode shape (lowest eigenvector) for the following set
of equations using the power method with e = 0.001 and assuming [x] =
[1] as an initial value. The [B] and [M] matrices are for a fourstory single
mass structural model.
106
1 / 12
1 / 12
1 / 12 1 / 12
1 / 12 11 / 60 11 / 60 11 / 60
1 / 12 11 / 60 37 / 120 37 / 120
1 / 12 11 / 60 37 / 120 57 / 120
6
0
0
5
0
0
0
0
4
0
103
0
0
0
0
0
=
0
0
0.000500
0.000500
0.000417
0.000917
0.000917
0.000917
0.000333
0.000733
0.001233
0.001233
0.000250 0
0.0000550 0
=
0.000925 0
0.001425 0
The iteration process and the final solution are shown in Table 2.20. The
xvector was normalized to the top value.
0.0004167
0.0003333
0.00025
0.0005
0.0009167
0.0007333
0.00055
0.0005
0.0009167
0.0012333
0.000925
0.0005
0.0009167
0.0012333
0.001425
1
0.0015
0.0027
1.8
2.3833333
0.004075
2.7166667
1
1.9797042
0.007602
2.79128
0.008961
3.2900051
1
2.0050559
0.008801
2.8593634
0.010446
3.3938361
4
=
0.006301
2.0087357
0.009004
2.8701807
0.010701
3.4111138
5
0.003146
emax= 0.5733384
0.0030778
emax= 0.103831
0.003137
emax = 0.0172777
0.0031465
emax = 0.002883
5
=
0.006322
2.0093106
0.009037
2.8719442
0.010742
3.4139968
6
0.003148
0.0027236
3
=
0.006171
0.003137
0.0015
2
=
0.005392
0.003078
1
=
0.003575
0.002724
6
=
0.003148
emax = 0.000484
0.006326
2.0094044
e = 0.001
0.009042
2.87223733
l1= 317.6593
0.010749
3.4144812
REFERENCES
Anon 179. The Nine Chapters on the Mathematical Art. China.
Clasen, B.I. 1888. Sur une nouvelle method de resolution de equations lineaires et
sul lapplication de cette method au calcil des determinates. 3rd ed. Brussels,
Belgium.
Commandant Benoit. 1924. Note sur une mthode de rsolution des quations
normales provenant de lapplication de la mthode des moindres carrs
un systme dquations linaires en nombre infrieur celui des inconnues,
Bulletin Godsique. Heidelberg, Germany.
Cramer, G. 1750. Introduction lanalyse des lignes courbes algbriques. Geneva,
Switzerland.
Gauss, C.F. 1903. Werke 9, Gttingen, Germany.
Le Verrier, U. 1839. Thesis on the secular variations of the orbits of the planets.
Paris, France: Academy of Sciences.
Newton, I. 1707. Arithmetica Universalis. London, England.
CHAPTER 3
3.1TRAPEZOIDAL RULE
Consider a function f(x) graphed between points a and b along the xaxis
as shown in Figure3.1. One approximation of the area under the curve is
to apply the trapezoidal rule by dividing the area into n strips of width Dx.
Then, approximate the area of each strip as a trapezoid.
Calling the ordinates f(xi) = yi (i = 1, 2, 3,, n, n+1), the areas of each
strip are as follows:
y + y3
y + y4
y + y2
A1 = x 1
, A = x 2
, A = x 3
, An
2 2
2 3
2
y + y n +1
= x n
2
b
A = f ( x ) dx = A1 + A2 + A3 + + An
a
A = f ( x ) dx =
a
A=
y +
x
( y1 + 2 y2 + 2 y3 + + 2 yn + yn +1 )
2
2y + y
yn1
y3
yn
yn+1
y2
y1
A1
A2
A3
x=a
An1
An
xi
x=b
y + y3
y + y4
y + y2
A1 = x 1
, A2 = x 2
, A3 = x 3
, An
2 x x
2
x
x x 2
y + yn +1rule.
Figure3.1.
x n
= Trapezoidal
A = f ( x ) dx = A1 + A2 + A3 + + An
a
A = f ( x ) dx =
a
A=
x
( y1 + 2 y2 + 2 y3 + + 2 yn + yn +1 )
2
x
+
y
2 yi + yn +1
2
i=2
x
2 yi + yn +1
y
+
2
i=2
p/2
1.5708
3.1416
y=sin(x)
1.23E16
A=
p
p
0 + 2 (1) + 0) = = 1.5708
(
4
2
p/4
p/2
3p/4
0.7854
1.5708
2.3562
3.1416
y=sin(x)
0.7071
0.7071
1.23E16
A=
p
(0 + 2 (0.7071) + 2 (1) + 2 (0.7071) + 0) = p8 ( 4.8284) = 1.8961
8
sin ( x) dx = cos ( x) 
p
0
= cos ( p ) cos ( 0) = 1 + 1 = 2
yih 2 yih3
+
+ Higher order terms
2!
3!
The change in y between points iand i+1 is equal to the area under the
ycurve between those two points, therefore the exact area in the strip is
as follows:
yih 2 yih3
+
+ Higher order terms
2!
3!
f h 2 f h3
yi +1 yi = f i h + i + i
+ Higher order terms (3.1)
2!
3!
yi +1 yi = yih +
y = f(x)
fi
fi+1
x
x
y
yi+1 yi
yi+1
yi
xi
xi+1
f h 2
f i +1 = f i + f i h + i
+ Higher order terms
2!
f i h = f i +1 f i
f i h 2
Higher order terms
2!
(3.2)
Substituting Equation 3.2 into Equation 3.1, the following can be derived:
f h 2
f i +1 f i i
Higher order terms h
2!
yi +1 yi = f i h +
2!
+
f i h3
+ Higher order terms
3!
yi +1 yi = exact area =
( fi + fi +1 ) h
2
f i h3
+ Higher order terms (3.3)
12
The first term on the right side in Equation 3.3 is the area or the trapezoid
and the rest is the error as follows:
ET =
f i h3
+ Higher order terms
12
The exact integral, I,can be derived using this error relation from two separate approximate integrals. The derivation is omitted from this text, but
may be found in Applied Numerical Methods for Digital Computations,
by James, Smith, and Wolford (1977). The improved integral is based on
two approximate integrals with a strip where h2 < h1 as follows:
I I
I I
I I h 2 + h12 h22 h22 = I h 2 + h1 2h 2
h1 1
h2 h1
h2
Ih2
h1
h2 I h1
2
h1 1
h2
If the second integration uses a strip onehalf that of the first with h1/h2=2,
the equation becomes the following:
I h 2 ( 2) I h1
2
( 2) 2 1
This is a defined as a firstorder extrapolation. If two firstorder extrapolations are performed, then their results can be combined into a secondorder relationship with the following:
I h 2 ( 2) I h1
4
( 2) 4 1
The general nth order extrapolation would take the following form with n
being the order of extrapolation:
I
I h 2 4n I h1
4n 1
x
y1 + 2 yi + yn +1
2
i=2
0
1.0000
A=
1.000
10.0000
1
(1 + 10) = 5.5
2
0
1.0000
0.500
3.1623
1.000
10.0000
1
(1 + 2 (3.1623) + 10) = 4.33114
4
0
1.0000
A=
0.250
1.7783
0.500
3.1623
0.750
5.6234
1.000
10.0000
1
(1 + 2 (1.7783) + 2 (3.1623) + 2 (5.6234) + 10) = 4.01599
8
0
1.0000
A=
1
(1 + 2 (1.3335) + 2 (1.7783) + 2 ( 2.3724) + 2 (3.1623)
16
+ 2 ( 4.2170) + 2 (5.6234) + 2 ( 7.4989) + 10) = 3.93560
I h 2 4n I h1
4n 1
First order:
4.33114 ( 4) 5.5
1
( 4)1 1
= 3.941518
4.01599 ( 4) 4.33114
1
( 4)1 1
3.93560 ( 4) 4.01599
= 3.910944
( 4)1 1
= 3.908798
Second order:
3.910944 ( 4) 3.941518
2
( 4) 2 1
2
3.908798 ( 4) 3.910943
I
( 4) 2 1
= 3.9089055
= 3.9808648
Third order:
3.9086548 ( 4) 3.9089055
3
5.5
( 4 )3 1
4.33114
3.941518
= 3.98086509
4.01599
3.910944
3.9089055
3.908798
3.6086548
3.98086509
3.9560
10
0
dx =
10 x 1
0 = 4.342944819 0.434294482 = 3.9808650337
ln 10
A=
xx
ax 3 bx 2
+
+ cx
ax 2 + bx + c dx =
3
2
x
2
3
a ( x ) + 2c ( x )
3
(3.4)
The constants a, b, and c are found using the three points (Dx, yi), (0, yi+1),
and (Dx, yi+2) as follows:
y=f(x)
(xi, yi)
(xi+1, yi+1)
(xi+2, yi+2)
y=ax2+bx+c
x
x
Figure3.3. Simpsons rule.
yi = a ( x ) + b ( x ) + c
2
yi +1 = c
yi + 2 = f ( x) = a (x) 2 + b(x) + c
Solving the three equations with three unknowns we obtain the following:
a=
b=
yi 2 yi +1 + yi + 2
2 ( x)
yi + yi + 2
2 ( x )
c = yi +1
Substituting the expressions for a and c into Equation 3.4, the following
is achieved:
A=
x
( yi + 4 yi +1 + yi + 2 )
3
n 1
n
x
y1 + 4 yi + 2 yi + yn +1
i = 3, 5 , 7
i = 2, 4,6
I h 2 16n I h1
16n 1
3 x
2
3 x
3 x
2
A=
2
ax 4 bx 3 cx 2
+ dx
+
+
ax + bx + cx + d dx =
3
2
3x
4
3
9
3
b ( x ) + 3d ( x )
4
(3.5)
The constants a, b, c, and d are found using the four points (3Dx/2, yi),
(Dx/2, yi+1), (Dx/2, yi+2), and (3Dx/2, yi+3) as follows:
3
3 x
3 x
3 x
yi = a
+ b
+ c
+d
2
2
2
3
x
x
x
yi +1 = a
+ b
+ c
+d
2
2
2
2
x
x
x
yi + 2 = a + b + c + d
2
2
2
3
yi + 3
3 x
3 x
3 x
+d
= a
+ b
+ c
2
2
2
Solve the four equations with four unknowns and then substitute these
back into Equation 3.5 to achieve the following:
A=
3 x
( yi + 3 yi +1 + 3 yi + 2 + yi + 3 )
8
n 1, n
n2
3 x
3
2
y
+
y
+
yi + yn +1
1
i
i = 2 , 3, 5 , 6
i = 4, 7
For an odd number of strips, both the onethird and threeeighths rules
must be used. The threeeighths rule is used to obtain the area contained
in three strips under the curve and then the onethird rule is used for the
remaining n3 strips.
A=
n 1
n
x
4
2
y
+
y
+
yi + yn +1
i
1
i = 3, 5 , 7
i = 2, 4,6
x
y
p/4
0.78540
0.43618
0
0.00000
0.00000
p/4
p/2
0.78540
0.43618
1.57080
0.00000
p
p
0 + 4 ( 0.43618) + 2 ( 0) + 4 ( 0.43618) + 0) = (3.48944)
(
12
12
= 0.913533
A=
3p/8
p/4
p/8
p/8
p/4
3p/8
p/2
A=
0.53113
p
(0 + 4 (0.53113) + 2 (0.43618) + 4 (0.14247) + 2 (0) + 4 (0.14247)
24
p
+ 2 ( 0.43618) + 4 ( 0.53113) + 0) =
(7.13352) = 0.933776
24
x cos ( x) dx = 2xcos ( x) + ( x
2
2 sin ( x ) p
3 x
( yi + 3 yi +1 + 3 yi + 2 + yi + 3 )
8
x
y
A=
3p
3p
1 + 3 ( 0.88975) + 3 ( 0.73258) + 0.73258) =
(6.59957) = 0.777494
(
80
80
A=
p/10
p/5
3p/10
0.31416
0.88975
0.62832
0.73258
0.94248
0.73258
2p/5
1.25664
0.88975
p/2
1.57080
1.00000
p
p
0.73258 + 4 ( 0.88975) + 1) = (5.29158) = 0.554133
(
30
30
2 n 1
x + cos 2 n 1 x dx
2 ( 4) 6 ( 2n)
2 ( 4) 6 ( 2n)
2 2 4
+
= + =
3 (5) 7 ( 2n + 1) 3 (5) 7 ( 2n + 1) 3 3 3
3.4GAUSSIAN QUADRATURE
The main difference of Gaussian quadrature from the previous methods
is that the interval to be integrated is not divided into strips. Instead, a
central point is used to determine the best places to evaluate the function.
The Gauss points indicate how far from the central point to go and then
each point is weighted. The derivation of this method is not included
here, but can be found in many advanced mathematics textbooks. The
method is named for Carl Friedrich Gauss (1801). The following is a general equation that shows the process for n Gauss points for integration:
b
i =1
A = f ( x ) dx = s f ( s1 sxi ) wi
ba
2
b+a
s1 =
2
s=
The number of points used should closely match the degree of the equation
to integrate. Table3.11 shows some of the Gaussian quadrature points, xi,
and their weights, wi.
Example 3.5. Gaussian quadrature
Determine the area under the curve from 1 to 10 for y = log10x using
Gaussian quadrature with 2, 3, and 4 points.
b
i =1
A = f ( x ) dx = s f ( s1 sxi ) wi
b a 10 1
s=
=
= 4.5
2
2
b + a 10 + 1
s1 =
=
= 5.5
2
2
One point
Two points
Three points
Four points
Five points
Six points
wi
2
1
1
0.555555556
0.888888889
0.555555556
0.347854845
0.652145155
0.652145155
0.347854845
0.236926885
0.478628670
0.568888889
0.478628670
0.236926885
0.171324492
0.360761573
0.467913935
0.467913935
0.360761573
0.171324492
0.577350269
2points
0.577350269
wi
s1+sxi
2.90192379
f(s1+sxi)
0.462686003
f(s1+sxi)wi
0.462686003
1
8.098076211
0.90838186
0.90838186
S
1.371067862
S*s
6.169805381
0.774596669
0
0.774596669
0.555555556
0.888888889
0.555555556
2.01431499
5.5
8.985685011
0.304127385
0.740362689
0.953551191
0.168959658
0.658100169
0.529750662
1.356810489
S*s
6.105647198
wi
s1+sxi
f(s1+sxi)
f(s1+sxi)wi
0.861136312
4points
0.339981044
0.339981044
0.861136312
0.347854845
0.652145155
0.652145155
0.347854845
1.624886596
3.970085302
7.029914698
9.375113404
0.210823056
0.598799838
0.846950055
0.97197653
0.073335822
0.390504413
0.552334375
0.338106745
1.354281355
S*s
6.094266097
wi
s1+sxi
f(s1+sxi)
f(s1+sxi)wi
10
1
x=0 to 8 Dx=2=hxand
y=0 to 8 Dy=2=hy
hxhy
= f ( x, y ) = V
9
Weighting operator:
1 4 2 4 1
4 16 8 16 4
2 8 4 8 2
4 16 8 16 4
1 4 2 4 1
Solving for z:
z=
xy
16
y
0
2
4
6
8
0
2
4
6
weight
1
4
2
4
1
4
16
8
16
f(x,y)
0
0
0
0
0
0
0.25
0.5
0.75
weight*f(x,y)
0
0
0
0
0
0
4
4
12
(Continued)
Table3.15.(Continued)
x
2
4
4
4
4
4
6
6
6
6
6
8
8
8
8
8
y
8
0
2
4
6
8
0
2
4
6
8
0
2
4
6
8
weight
4
2
8
4
8
2
4
16
8
16
4
1
4
2
4
1
f(x,y)
1
0
0.5
1
1.5
2
0
0.75
1.5
2.25
3
0
1
2
3
4
S
weight*f(x,y)
4
0
4
4
12
4
0
12
12
36
12
0
4
4
12
4
144
bx , by
ax , bx
bx ax
b + ax
and sx1 = x
2
2
by a y
by + a y
sy =
and s y1 =
2
2
sx =
The number of points used should closely match the degree of the equation to integrate. The same Gauss points and weights from Section 3.4 are
used in each direction.
Table3.16 shows the set up and summation as follows with xi, yi given
values and wx, wy corresponding weights when using three points:
Table3.16. Example 3.7 Double integration by Gaussian quadrature
xi
yi
x=
y=
sx1 + sxxi
sy1 + svyi
wx
wy
f(x,y)
wx*wy*f(x,y)
2.065827
0.774597
3.642289
2.065827
3.098387
0.774597
0.774597
0.888889 0.888889
3.642289
6.320988
3.642289
2.065827
0.774597
3.642289
0.774597
3.098387
V = sx s y = 4 ( 4) 29.153 = 466.45
2.065827
29.153453
f ( 0) = 1b1
f ( 0) = 1( 2) b2
f ( 0) = 1( 2) 3b3 = 3!b3
f
f ( 0) = 1b1 = cos ( 0) = 1 b1 = 1
f ( 0) = 2b2 = sinn ( 0) = 0 b2 = 0
f ( 0) = 3!b3 = cos ( 0) = 1 b3 =
f ( 0) = 4 !b4 = sin ( 0) = 0 b4 = 0
1
3!
1
5!
f ( 0) = 6 !b6 = sin ( 0) = 0 b6 = 0
f ( 0) = 5!b5 = cos ( 0) = 1 b5 =
f ( 0) = 7 !b7 = cos ( 0) = 1 b7 =
f ( 0) = 8!b8 = sin ( 0) = 0 b8 = 0
f ( 0) = 9 !b9 = cos ( 0) = 1 b9 =
1
7!
1
9!
1 3 1 5 1 7 1 9
x + x x + x
3!
5!
7!
9!
( )
( )
( )
( )
( )
3
5
1
1
1 p 7 1 p 9
y = sin p 4 = p 4 p 4 + p 4
+
3!
5!
7! 4
9! 4
y = 0.7853981634 0.0808455122 + 0.0024903946 0.0000365762
+ 0.0000003134
y = 0.7071067829
1
2
f (0) = 3!b3 = e 0 = 1 b3 =
f (0) = 4!b4 = e 0 = 1 b4 =
1
4!
1
3!
f (0) = 5!b5 = e 0 = 1 b5 =
1
5!
1 2 1 3 1 4 1 5
x x + x x
2!
3!
4!
5!
yih 2 yih3
+
+
2!
3!
(3.6)
Next, expanding the Taylor series for a function y=f(x) at x=(xih) gives
the following equation:
y ( xi h ) = yi yi h +
yi' h 2
2!
yih3
+
3!
(3.7)
If Equation 3.7 is subtracted from Equation 3.6, a first derivative relationship is as follows:
yi =
y ( xi + h ) y ( xi h ) yih 2
+
2h
6
(3.8)
This may be written as follows with successive values of y and the higher
order terms omitted:
yi =
yi +1 yi 1
2h
y ( xi + h ) 2 yi + y ( xi h ) yih 2
+ (3.9)
2
h
12
This may be written as follows with successive values of y and the higher
order terms omitted:
yi=
yi +1 2 yi + yi 1
h2
yi ( 2h)
y ( 2h) yi( 2h)
+ i
+
+ (3.10)
2!
3!
4!
2
yi ( 2h)
y ( 2h)
y ( 2h)
i
+ i
(3.11)
2!
3!
4!
2
If Equation 3.11 is subtracted from Equation 3.10 and the equation for the
first derivative is substituted into the result, a third derivative relationship
is as follows, with an order of error h2:
yi=
yi + 2 2 yi +1 + 2 yi 1 yi 2
2h 3
If Equation 3.11 is added to Equation 3.10 and the equation for the second
derivative is substituted into the result, a fourth derivative relationship is
as follows with an order of error h2:
yi =
yi + 2 4 yi +1 + 6 yi 4 yi 1 + yi 2
h4
These are the central difference expressions with error order h2. Higher
order expressions can be derived if we include more terms in each expansion. Forward difference expressions can be derived by using Taylor series
expansion of x=(xi+h), x=(xi+2h), x=(xi+3h), and so forth. Backward
difference expressions may also be derived by Taylor series expansion of
x=(xih), x=(xi2h), x=(xi3h), and so forth. The following are the central, forward, and backward difference expressions of varying error order.
These were compiled from Applied Numerical Methods for Digital Computations, by James, Smith, and Wolford (1977). They can also be written
in a reverse graphical form that is sometimes used and compiled from
Numerical Methods in Engineering, by Salvadori and Baron (1961).
Central difference expressions with error order h2:
yi +1 yi 1
2h
yi +1 2 yi + yi 1
yi=
h2
y 2 yi +1 + 2 yi 1 yi 2
yi = i + 2
2h 3
y 4 yi +1 + 6 yi 4 yi 1 + yi 2
yi= i + 2
h4
yi =
2hD
h2 D 2
2h3 D 3
h4 D 4
i 2 i 1 i i +1 i + 2
1 0
1
1 2 1
1
2
0 2
1
1
4 6 4
1
12hD
12h 2 D 2
8h3 D 3
6h 4 D 4
i 3 i 2 i 1 i
i +1 i + 2 i + 3
1
8
0
8
1
1 16 30 16
1
1
8 13
0 13
8
1
1 12 39 56 39 12
1
hD
h2 D 2
h3 D 3
h4 D 4
i i +1 i + 2 i + 3 i + 4
1 1
1 2
1
1 3
3
1
1 4
6
4
1
2hD
h2 D 2
2h 3 D 3
h4 D 4
i i +1 i + 2 i + 3 i + 4 i + 5
3 4
1
2 5
4
1
5 18 24 14
3
3 14 26 24 11
2
hD
h2 D 2
h3 D 3
h4 D 4
i 4 i 3 i 2 i 1
1
1
2
1
3
3
1
4
6
4
i
1
1
1
1
3 yi 4 yi 1 + yi 2
2h
2Numerical
yi 5 yi 1 + 4Integration
yi 2 yi 3
and Differentiation 123
yi =
h2
5 y 18 yi 1 + 24 yi 2 14 yi 3 + 3 yi 4
yi = i
2h 3
3 y 14 yi 1 + 26 yi 2 24 yi 3 + 11 yi 4 2 yi 5
yi= i
h4
yi =
2hD
h2 D 2
2h 3 D 3
h4 D 4
i 5 i 4 i 3 i 2 i 1
1
4
1
4
5
3
14 24 18
2
11 24 26 14
i
3
2
5
3
4 k/ft
A
B
250
Y3
Y2 Y1
Y0
Y1
Y2
Y3
Y2
Y1
Y0
Y1 Y2 Y3
Placing the central difference operator on y0, the first equation can be written from Figure3.5:
Y3
Y2
Y1
Y0
Y1
Y2
Y3
Y2
Y1
Y0
Y1
Y2
Y3
q0 =
EI
EI
y2 ) 4 ( y1 ) + 6 y0 4 y1 + y2 ) = 4 (6 y0 )
4 ((
h
h
For the second equation, the central difference operator is placed on y1 and
is shown in Figure3.6.
q1 =
EI
(4 y0 + 5 y1 4 y2 + y3 )
(( y1 ) 4 y0 + 6 y1 4 y2 + y3 ) = EI
h4
h4
The third and fourth equations can be written by placing the central difference operator on y2 and y3.
EI
EI
y 4 y1 + 6 y2 4 y3 + y2 ) = 4 ( y0 4 y1 + 7 y2 4 y3 )
4 ( 0
h
h
EI
EI
q3 = 4 ( y1 4 y2 + 6 y3 4 y2 + y1 ) = 4 ( 2 y1 8 y2 + 6 y3 )
h
h
q2 =
0 2 8 6 y3 q3
From the conditions of the beam, two simplifications can be made. First,
the load is uniform and all the values of q are the same. Second, the deflection at point 0 is known to be zero, so the first row and column can be
eliminated since they correspond to those values.
5 4 1 y1 q
EI
4 7 4 y2 = q
h4
2 8 6 y3 q
y3
0.89855
Once the deflections are found, the other three desired values, q, M, and
V, can each be found from the corresponding central difference operators
using the same order of error. The procedure is the same as the operator
is laid upon each of the values that are unknown then the corresponding
equation may be written. The relationship for rotation is as follows:
yi +1 yi 1
2h
0
2
0 0 y0 q 0
1 0 1 0 y q
1
1 = 1
2h 0 1 0 1 y2 q 2
0 0 0 0 y3 q 3
0 2 0 0 y0 q 0
3 1 0 1 0 y1 q1
=
L 0 1 0 1 y2 q 2
0 0 0 0 y3 q 3
2 y1 q 0
3 y0 + y 2 q 1
=
L y1 + y3 q 2
0
q 3
yi = q =
q0
0.00912
q
0.00781
radians = 1
q2
0.00443
0.0000
q3
The exact solution at the end is q0=0.00936 radians, which is a 2.51% error,
while the value in the center is exact. The relationship for moment is as follows:
y 2 yi yi 1
M
= i +1
EI
h2
2 0 0 0 y0 M 0
EI 1 2 1 0 y1 M 1
=
2
h 0 1 2 1 y2 M 2
0 0 2 2 y3 M 3
yi =
2 0 0 0 y0 M 0
36 EI 1 2 1 0 y1 M 1
2
=
L 0 1 2 1 y2 M 2
0 0 2 2 y3 M 3
2 y0
M 0
y 2y + y M
36 EI 0
1
1
2
2
=
+
y
y
y
M
2
L
2
1
2
3
2 y2 2 y3 M 3
M 0
0.00
M
2083.4
kipinches = 1
M 2
3333.3
3750.0
M3
The exact solution at the center is M3=3750 kin which is exact, while the
values at the ends is also exact. The relationship for shear is as follows:
y 2 yi +1 + 2 yi 1 yi 2
V
= i+2
EI
2h 3
0 4 2 0 y0 V0
EI 2 1 2 1 y1 V1
=
3
2h 1 2 1 2 y2 V2
0 0 0 0 y3 V3
yi =
0 4 2 0 y0 V0
108 EI 2 1 2 1 y1 V1
=
L3 1 2 1 2 y2 V2
y 2 yi +1 + 2 yi 1 yi 2
V
= i+2
EI
2h3
0 4 2 0 y0 V0
EI 2 1 2 1 y1 V1
=
3
1 2 y2 V2
h 1 2 ANALYSIS
128NUMERICAL 2STRUCTURAL
0 0 0 0 y3 V3
yi =
0 4 2 0 y0 V0
108 EI 2 1 2 1 y1 V1
=
3
L 1 2 1 2 y2 V2
0 0 0 0 y3 V3
4 y1 + 2 y2
V0
2 y + y 2 y + y V
108 EI 0
1
1
2
3
=
L3 y0 + 2 y1 + y2 2 y3 V2
0
V3
V0
35.27
V
34.92
kips = 1
V2
19.90
0
V3
The exact solution at the end is V0 = 50.00 k, which is a 16.67% error,
while the value in the center is exact. The large error at the end is due to
the fact that the shear drops there, which creates a discontinuity in the
equation. Using more segments would reduce the error, but it would still
be more inaccurate than the other values.
The next example is similar to the previous one, but is included to
show differences in modeling and accuracy. It also uses the higher order
(smaller error) of error equations for more accuracy.
5 k/ft
A
B
300
Y3
Y2
Y1
Y0
Y1
Y2
Y3
Y2
Y1
Y0
Y1
Y2
Y3
yi=
q=
y + 12 yi + 2 39 yi +1 + 56 yi 39 yi 1 + 12 yi 2 yi 3
q
= i+3
EI
6h 4
EI
( yi + 3 + 12 yi + 2 39 yi +1 + 56 yi 39 yi 1 + 12 yi 2 yi 3 )
6h 4
Placing the central difference operator on all four unknown points in the
model, the linear nonhomogeneous solution set can be obtained and
solved.
56 78 24 2 y0 q0
EI 39 68 41 12 y1 q1
4
=
6h 12 41 68 39 y2 q2
2 24 78 56 y3 q3
68
EI
4 41
6h
24
766
6qh 4
1360
EI
1566
41 12 y1 q
68 39 y2 = q
78 56 y3 q
1360 783
1 y1
1
3520 2160
1 = y2
15120
1 y3
4320 2943
2909 y1
q L4
7040 = y
2
3265920 EI
8829 y3
y1
0.2150
0.5202 inches = y
2
y3
0.6524
The exact solution at the center is y3=0.6284 inches, which is a 3.81%
error. This is more error than the deflection found in Example 3.10,
even though a more accurate operator was used. This is due to the
fact that this physical model has more variation in deflection than
that of the simply supported beam. The relationship for rotation is as
follows:
yi + 2 + 8 yi +1 8 yi 1 + yi 2
12h
0
0
0
0 y0 q 0
8 1 8 1 y q
1
1 = 1
12h 1 8 1 8 y2 q 2
0 0 0 0 y3 q 3
0 0 0 0 y0 q 0
1 8 1 8 1 y1 q1
=
2 L 1 8 1 8 y2 q 2
0 0 0 0 y3 q 3
0
q 0
8 y + y + 8 y y q
1
1
0
1
2
3
=
2 L y0 8 y1 y2 + 8 y3 q 2
0
q 3
yi = q =
q 0
0.0000
q
0.0052
radians = 1
q 2
0.0041
0.0000
q 3
The solution at the end is q0=0.00000 radians, which is exact, while
the value in the center is also exact. The relationship for moment is as
follows:
M yi + 2 + 16 yi +1 30 yi + 16 yi 1 yi 2
=
EI
12h 2
0 y0 M 0
30 32 2
16 31 16
1 y1 M 1
EI
=
12h 2 1 16 31 16 y2 M 2
2 32 30 y3 M 3
0
yi =
+ 16 yi +1 30 and
yi + 16
yi 1 yi 2
M yi + 2Integration
Numerical
Differentiation
131
yi =
=
EI
12h 2
0 y0 M 0
30 32 2
16 31 16
1 y1 M 1
EI
=
12h 2 1 16 31 16 y2 M 2
2 32 30 y3 M 3
0
0 y0 M 0
30 32 2
16 31 16
1 y1 M 1
3EI
=
2
L 1 16 31 16 y2 M 2
2 32 30 y3 M 3
0
30 y0 + 32 y1 2 y2 M 0
3EI 16 y0 31y1 + 16 y2 y3 M 1
=
2
L y0 + 16 y1 31y2 + 16 y3 M 2
2 y1 + 32 y2 30 y3 M 3
M 0
3920
M
676
kipinches = 1
M 2
1510
2252
M3
The exact solution at the end is M0=4500 kin, which is a 12.89% error,
while the value in the center is 2250 kin, which is a 0.11% error. The large
error at the end is due to the fact that the moment drops there, which creates a discontinuity in the equation. Using more segments would reduce
the error, but it would still be more inaccurate than the other values. The
relationship for shear is as follows:
y + 8 yi + 2 13 yi +1 + 13 yi 1 8 yi 2 + yi 3
V
= i+3
EI
8h3
0
0 y0 V0
0 0
13 8 13 8 y V
EI
1 = 1
3
8h 8 13 8 13 y2 V2
0
0 y3 V3
0 0
yi =
0
0 y0 V0
0 0
13 8 13 8 y V
27 EI
1 = 1
3
L 8 13 8 13 y2 V2
0
0 y3 V3
0 0
0
V0
13 y 8 y 13 y + 8 y V
27 EI
1
0
1
2
3
=
3
L 8 y0 + 13 y1 + 8 y2 13 y3 V2
0
V3
8h 8 13 8 13 y2 V2
0
0 y3 V3
0 0
0
0 y0 V0
0 0
13 8 13 8 y V
27 EI
1 = 1
3
8 13 8 13 y2 V2
L
132NUMERICAL STRUCTURAL ANALYSIS
0
0 y3 V3
0 0
0
V0
13 y 8 y 13 y + 8 y V
27 EI
1
0
1
2
3
=
3
L 8 y0 + 13 y1 + 8 y2 13 y3 V2
V3
V0
0
V
07
V0
54.76 i = V1
2
0 kips = V21
V2
25.60
0
V3
0
V3
The exact solution at the end is V0=75.00 k, which is a huge error, while
the value in the center is exact. The large error at the end is due to the fact
that the shear and moment drop dramatically at that point, which creates a
discontinuity in the equation.
The final example of using difference operators to solve differential equations is the critical buckling load of a column. The critical buckling load
of a pinned end column is sometimes included in strength of materials, but
will be derived here. The derivations start with the differential equation of
the elastic curve similar to a beam. The deflected column under the critical
load is shown in Figure3.8.
d2y
M
= y =
EI
dx 2
At any point x along the column, there is a deflection y that will produce an
eccentric moment in the column equal to Pcry. This is used in the equation
of the elastic curve as follows:
Z
Pcr
Pcr
B
A
X
L
P
d2y
= cr y
2
EI
dx
Pcr
y +
y=0
EI
This is a secondorder, linear, ordinary differential equation with the following general solution.
y = Acos ( kx ) + Bsin ( kx )
where k =
Pcr
EI
No deflection
No load
where n=1, 2, 3,
Therefore, the following can be found and is the critical buckling load:
kL =
Pcr
L = np
EI
Pcr 2
L = n 2p 2
EI
n 2p 2 EI
Pcr =
L2
p 2 EI
L2
Pcr p 2 EI p 2 EIr 2
p 2E
=
=
=
A
AL2
IL2
L2 2
r
( )
This is known as the Euler buckling stress and was derived by Leonhard
Euler in 1757.
Example 3.12 Column buckling with difference operator
Calculate the critical buckling load, Pcr, for a 25 foot long fixed end column using central difference operator of order of error h2 at 1/6th points.
The column has E=29,000 ksi and I=1000 in4.
This problem has the same model as Example 3.11 except an axial
load is applied instead of a uniform lateral load and is shown in Figure3.9.
The central difference expressions with error of order h2 will be used
to solve for the values. Since the load is known, we will use the fourth
derivative relationship between load and deflection. This can be found
Z
P
A
B
250
Y3
Y2
Y1
Y0
Y1
Y2
Y3
Y2
Y1
Y0
Y1
Y2
Y3
Pcr
y=0
EI
y +
Pcr
y = 0
EI
y 2 yi + yi 1
yi + 2 4 yi +1 + 6 yi 4 yi 1 + yi 2
+ Q i +1
h2
h4
Q
1
0 = 4 ( yi + 2 4 yi +1 + 6 yi 4 yi 1 + yi 2 ) + 2 ( yi +1 2 yi + yi 1 )
h
h
y + Qy = 0 =
The solution process is similar to Example 3.10 and 3.11 except that the
two operators are placed on the model at the location of the unknown
deflections. The value at point 0 of y0 is zero and can be eliminated from
the solutions. This becomes a homogeneous linear algebraic solution set.
7 4 1
1296 4 7 4 + 36 Q
L4
L2
2 8 6
2 1 0 y1 0
1 2 1 y = 0
2
0 2 2 y3 0
144
36
252
L2 2Q L2 + Q
L2
y1 0
144 + Q 252 2Q 144 + Q y = 0
L2
2
L2
L2
y3 0
288
216
72
Q
2
Q
L2
L2
L2
A nontrivial solution to a homogeneous linear algebraic set exists if the
determinant of the coefficient matrix is zero. Therefore, we can find Q by
setting the determinant of the coefficient matrix equal to zero. This will be
done using the basket weave method for a 33 matrix.
252
252
72
216
144
144
0 = 2 2Q 2 2Q 2 2Q + 2 + Q 2 + Q 2
L
L
L
L
L
L
36 252
36 144
72
288
+ 2 2 + Q 2 + 2Q 2 2 2Q 2
L L
L L
L
L
144
144
216
2 + Q 2 + Q 2 2Q
L
L
L
252
144
288
2 2Q 2 + Q 2 + 2Q
L
L
L
252
54432
36Q
41472
576Q
252
252
216
144
144
72
0 = 2 2Q 2 2Q 2 2Q + 2 + Q 2 + Q 2
L
L
L
L
L
L
36 252
36 144 STRUCTURAL
288 ANALYSIS
136NUMERICAL
72
+ 2 2 + Q 2 + 2Q 2 2 2Q 2
L L
L
L L
L
144
144
216
2 + Q 2 + Q 2 2Q
L
L
L
252
144
288
2 2Q 2 + Q 2 + 2Q
L
L
L
36QQ
41472
Q Q 2 2
252
54432 936
41472576576
252
54432
4Q42Q2 4 4 2 2 2Q
0 = 22 22QQ 4 4 2 2 + +
2Q
L
LL
LL
L L
L L
Multiplying the values for like terms yields the following:
504Q
10386 72Q
144
0 = 2 Q 31104 + 2 2Q 2 +
2
L
L
L4
L
18144 144Q
36 41472 576Q
2 + 2Q 2
2
+ 2
L L4
L
L4
L
Multiplying all the values and combining like terms:
0=
1119744 49248
576
Q + 2 Q 2 2Q 3
L6
L4
L
The solution to these cubic equations yields the following general value
of Q:
Q=
Pcr
= 0.0004
EI
equations you simply add the location of the operator at i, known as the
pivot point. To take the product of two differential equations, you must
multiply the value of the operator at iof one differential to each of the
terms in the other differential. Partial differentials also require both
the x and y direction, so they will be written horizontally and vertically.
The following are six examples of creating partial difference operators for
the equation used to solve plate bending. The partial differential equation
is listed first in each of the six examples followed by the basic operators
that represent the equation.
Example 3.13 Partial difference operator
d 4z
d 4z
d 4z 1
+
2
+
=
(1 4 6 4 1)
dx 4
dx 2 dy 2 dy 4 h 4
1
1 1
1
+ 2 * 2 (1 2 1) * 2 2 + 4
h
h h
1
1
= 4 (1 4 6 4 1) + 2 * 4
h
1
2
1
2 4 2 + 1
h4
1 2 1
1
4
6
4
1
1
4
6
4
1
2 8 2
1
= 4 1 8 20 8 1
h
2 8 2
1
1
2 8 2
q d 4z
d 4z
d 4z 1
= 4 + 2 2 2 + 4 = 4 1 8 20 8 1 (3.12)
D dx
dx dy
dy
h
2 8 2
The value D is the flexural rigidity of the plate and is equal to the following:
D=
Et 3
12 1 n 2
The value of Poissons ratio, u, relates the elastic modulus to the shear
modulus and is given as the following from strength of materials:
G=
E
2 (1 + n )
d 2z
d 2z 1
1
1
=
+
=
n
n
1
2
1
2
1
2
1
+
n
1
(
) 2 2
( )
dx 2
dy 2 h 2
h
h
n
d 2 z
d 2z 1
M x = D 2 + n 2 = 2 1 2 (1 + n ) 1 (3.13)
dy h
dx
n
d 2z d 2z 1
n 2 + 2 = 2 (n
dx
dy
h
2n
1
n ) + 2
h
2 = 1 n
h2
1
d 2 z d 2 z 1
M y = D 2 + 2 = 2 n
dy h
dx
2 (1 + n ) n
2 (1 + n ) n
(3.14)
M xy
1 0 1
d 2z
1
= D (1 n )
= 2 0 0 0
dxdy 4h 1 0 1
M yx
1 0 1
d 2z
1
= D (1 n )
= 2 0 0 0 (3.16)
dxdy
4
h
1 0 1
(3.15)
1 0 1
1
1
= 3 ( 1 2 0 2 1) + 3 2 0 2
2h
2h
1 0 1
1 0 1
1
= 3 1 4 0 4 1
2h
1 0 1
1 0 1
d 3 z
d 3z
1
(3.17)
=
Qx = D 3 +
1
4
0
4
1
2
3
2
dx
dxdy
h
1 0 1
2
1
3
3
d z d z 1
1
1
+ 3 = 2 (1 2 1) * 0 + 3 0
2
dx dy dy
2h 2h
h
1
2
1
1 4 1
2
1 2 1
1
1
1
2h 3 0 0 0 + 2h 3 0 = 2h 3 0 0 0
2
1 4 1
1 2 1
1 4 1
d 3z
d 3z
1
Qy = D 2 + 3 = 3 0 0 0 (3.18)
dx dy dy 2h 1 4 1
dy
My
My + y dy
Myx
dy
y
Qy
Qy + y dy
Myx +
dx
Mx + Mx
x dx
Mxy
dx
x
Qx
Qx +
dx
x
Mxy +
The plate will be divided into 10 segments in the horizontal and vertical
directions. Since the plate is square with a uniform load and fixed on all
edges, it has horizontal and vertical symmetry. It is also symmetric about
both diagonals of the plate. The symmetry will result in only 21 unique points
on the plate for the 121 possible locations. This is shown in Figure3.11.
The difference operator for Equation 3.12 is repeated here and can be
shown in graphical form in Figure3.12.
1
2 8 2
q d 4z
d 4z
d 4z 1
= 4 + 2 2 2 + 4 = 4 1 8 20 8 1
D dx
dx dy
dy
h
2 8 2
1
The operator is laid upon each of the 21 unique points on the plate resulting in 21 equations involving the deflections. Figure3.13 shows the operator laid upon point 1.
The result will be a nonhomogenous set of 21 equations involving
the unknown deflections. Since the plate is fixed at the boundaries, the
13
13
16
17
18
17 16
14
10
10 14
17
19
20
19 17
15
11
11 15
18
20
21
20 18
14
10
10 14
17
19
20
19 17
13
13
16
17
18
17 16
12
12
13
14
15
14 13
10
11
10 9
10
11 10
12
12
13
14
15 14
13
8
8
20
8
8
d 2 z
d 2z 1
M x = D 2 + n 2 = 2 1 2 (1 + n ) 1
dy h
dx
n
15
11
11
15
18
20
21
14
10
10
14
17
19
20
13
13
16
17
18
12
31
12
13
14
15
72
28
72
10
11
31
28
10
28
31
72
28
72
10
11
31
12
13
14
15
12
0.000291121
0.000749271
z9
0.001156404
z10
0.001421248
z11
0.001511939
z12
0.001945784
z13
0.003037557
z14
0.003762980
z15
0.004014203
z16
0.004780224
z17
0.005951021
z18
0.006358843
z19
0.007429735
z20
0.007946380
z21
0.008501580
inches
z7 = q7
h4
z8
q8 * D
z9
q9
z10
q10
z11
q11
z12
q12
z13
q13
z14
q14
z15
q15
z16
q16
z17
q17
z18
q18
z19
q19
z20
q20
z21
q21
2.5
2.5
1.0
0.0
0.0
0.0
1.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.5
1.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
0.0
1.0
2.5
1.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
0.0
1.0
2.5
1.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
0.0
2.0
2.5
1.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
0.0
0.0
2.5
1.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.0
2.5
0.0
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.3
0.0
0.0
1.0
2.5
1.3
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
1.0
2.5
1.0
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
2.0
2.5
1.0
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
0.0
2.5
1.0
0.0
0.0
0.0
0.5
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.0
2.5
0.0
0.0
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.3
0.0
1.0
2.5
1.3
0.0
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
2.0
2.5
1.0
0.0
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
0.0
2.5
1.0
0.0
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.0
2.5
0.0
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.3
2.0
2.5
1.3
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.3
0.0
2.5
1.0
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.0
2.5
0.0
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.5
2.5
1.3
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
2.5
0.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
z21
z20
z19
z18
z17
z16
z15
z14
z13
z12
z11
z10
z9
z8
z7
z6
z5
z4
z3
z2
z1
0.000000000 kin s =
x
0.00000
0.093805542
0.56283
0.241431854
1.44859
0.372619154
2.23571
0.457957773
2.74775
0.487180224
2.92308
0.134551926
0.80731
0.039177480
0.23506
0.025067632
0.15041
0.036067778
0.21641
0.042662612
0.25598
0.084373209
0.50624
0.258403727
1.55042
0.330357218
1.98214
0.349192954
2.09516
0.460673396
2.76404
0.605975261
3.63585
0.647615145
3.88569
0.774999710
4.65000
0.832219502
4.99332
0.894489044
5.36693
ksi
REFERENCES
Gauss, C.F. 1801. Disquisitiones Arithmeticae. Lipsia, Germany.
Romberg, W. 1955. Vereinfachte numerische Integration. Trondheim, Norway:
Det Kongelige Norske Videnskabers Selskab Forhandlinger.
Salvadori, M., and M. Baron. 1961. Numerical Methods in Engineering. 2nd ed.
Englewood Cliffs, New Jersey: PrenticeHall.
Simpson, T. 1750. Doctrine and Application of Fluxions. London, England.
Timoshenko, S., and S. WoinowskyKrieger. 1959. Theory of Plates and Shells.
2nd ed. New York, New York: McGrawHill.
CHAPTER 4
Matrix Structural
Stiffness
Building structures are made up of columns, beams, girders, joists, slabs,
walls, shells, and many other components that act together to resist the
loads placed on them. These members can be at various orientations, but
they must be represented in a common mathematical form. The structural
stiffness method is used to represent members, loads, support constraints,
and other components of a building structure in a consistent manner. This
chapter will focus on linearly elastic members subjected to axial, bending,
shear, and torsional forces.
y0
xl
yl
x0
a = x0 cosa
b = y0 sina
c = y0 cosa
d = x0 sina
xl = a + b = x0 cosa + y0 sina
yl = c d = y0 cosa x0 sina
zl = z 0
The equations for x, y, and z can be represented in matrix form as follows:
cosa
sina
sina
cosa
0
0 xo xl
0 yo = yl
1 zo zl
(4.1)
d
zl
c
z0
x0
xl
a = x0 cos
x b
b = z0 sin b
c = z cos b
0
Figure4.3. Example4.2 Rotation,
b.
d = x0 sin b
xl = a b = x0 cos b z0 sin b
zl = c + d = z0 cos b + x0 sin b
yl = y0
The equations for x, y, and z can be represented in matrix form as follows:
cos 0 sin xo xl
0
1
0 yo = yl
sin 0 cos zo zl
(4.2)
z0
yl
zl
a = y0 cosg
b = z0 sing
y0
= z0 cosgg.
Figure4.4. Example4.3cRotation,
d = y0 sing
yl = a + b = y0 cosg + z0 sing
zl = c d = z0 cosg y0 sing
xl = x0
The equations for x, y, and z can be represented in matrix form as follows:
0 xo xl
sing yo = yl
cosg zo zl
0
1
0 cosg
0 sing
(4.3)
The rotation matrix from the global to the local system, [R0l], is found
from these rotations about the global Z, Y, and X axes, in that order. The
order of the matrix multiplication goes from right to left. Alpha is first
multiplied by beta and the resultant is multiplied by gamma.
[ R ] = [g ][ b ][a ]
0l
Substituting Equations 4.1 through 4.3 for [g], [b], and [a], respectively,
results in the following:
0
0 cos b 0 sin b cosa sina 0
1
sinacos b
cosa
sina sin b
sin b
0
cos b
cosacos b
sinacos b
sin b
cos b 0
0
0 cosacos b sinacos b
1
cosa
[ R0l ] = 0 cosg sing sina
0 sing STRUCTURAL
cosg cosa sin
b sina sin b
152NUMERICAL
ANALYSIS
sin b
0
cos b
cosacos b
sinacos b
sin b
[g ] = [g ]
l0
0l
0
0
1
= 0 cosg sing
0 sing cosg
The rotation matrix from local to global system, [Rl0], can be found from
the individual rotations or directly from [R0l].
[ Rl 0 ] = [ R0l ]
cosa
[ Rl 0 ] = sina
0
= [a ] [ b ] [g ]
T
sina
cosa
0
cosa cosb
[ Rl 0 ] = sina cosb
sinb
[ Rl 0 ] = [ R0l ]
0 cosb
0 0
1 sinb
0 sinb 1
0
1
0 0 cosg
0 cosb 0 sing
cosa cosb
0
sing
cosg
sina cosb
sina sinb sing + cosa cosg
sina sinb cosg cosa sing
sinb
cosb sing
cosb cosg
These rotation transformations will be used later to relate the local member stiffness to the global member stiffness.
4.3TRANSMISSION MATRIX
In addition to knowing the local components of global forces, the effect a
force applied to one end of a member has on the other end is important. It
is often necessary to state the effect at one point in a structural system due
to a cause known to exist at some other point in the system. This is where
the transmission transformation is used. In Figure4.5, the cause of a force
at point 2 is transmitted to the effect at point 1. This is achieved by using
an equivalent static force at 1. In the study of rigid body equilibrium, this
is stated as SF1= SF2.
The six orthogonal forces at a point are shown in Figure4.6. These
are the forces and moments in each direction X, Y, and Z. The moments
are represented with double arrowheads.
The static equivalent force system is found where the moment arm
distances (x2 x1), (y2 y1), and (z2 z1) are measured from the effect point,
1, to the cause point, 2.
Z
Y
2 . (x2,y2,z2)
1 . (x1,y1,z1)
MZ
MY
PZ
1
PY
PX
MX
Fx1 = Fx 2 Px1 = Px 2
Fy1 = Fy 2 Py1 = Py 2
Fz1 = Fz 2 Pz1 = Pz 2
M x1 = M x 2 M x1 = M x 2 Py 2 ( z2 z1 ) + Pz 2 ( y2 y1 )
M y1 = M y 2 M y1 = M y 2 + Px 2 ( z2 z1 ) Pz 2 ( x2 x1 )
M z1 = M z 2 M z1 = M z 2 Px 2 ( y2 y1 ) + Py 2 ( x2 x1 )
0
0
1
0
1
0
0
0
1
0 z y
z
0 x
0
y x
0
0
0
1
0
0
0 Px 2 Px1
0 Py 2 Py1
0 Pz 2 Pz1
=
0 M x 2 M x1
0 M y 2 M y1
1 M z 2 M z1
0
0
0
0
1
0
(4.4)
Example4.4Transmission matrix
Determine the transmission matrix for the coplanar XY system from the
origin end, i, to other end, j.
The cause end is i as shown in Figure4.7 and the effect end is j at the
right end a distance L away.
Y
i = cause
j = effect
x = ( xc xe ) = xi x j = L
This can be substituted into Equation 4.4 and the unneeded distances and
forces are removed to get a 33 matrix involving only Px, Py, and Mz.
1 0 0 0 0 0 Px 2 Px1
0 1 0 0 0 0 P P
y 2 y1
0 0 1 0 0 0 Pz 2 Pz1
=
0 0 0 1 0 0 M x 2 M x1
0 0 L 0 1 0 M y 2 M y1
0 L 0 0 0 1 M z 2 M z1
1 0 0 Pxi Pxj
0 1 0 P = P
yi yj
0 L 1 M zi M zj
The axial force in the member, Px, is a direct transmission, whereas the
shear, Py, and bending, Mz, forces are linked. The transmission for just the
shear and bending will be used later in the chapter and can be written as
follows:
1 0 Pyi Pyj
L 1 M = M
zi zj
(4.5)
A similar transmission matrix can be derived for shear and bending in the
XZ system.
1 0 Pzi Pzj
L 1 M = M
yi yj
(4.6)
q12 =
1
M
dx
EI
The second theorem of area moment states that the tangential deviation
of a point, 1, from a tangent to the elastic curve at point 2 (the tangent)
is equal to the moment of the area of the M/EI diagram between points
1and2, taken about point 1 (the point). Note that the moment of the area
is the area times a distance, x1, from point 1.
2
t12 =
1
M
x1dx
EI
RB
tAB
wL
wL2
2EI
3L/4
The reactions at the fixedend on the right can be written in terms of the
vertical reaction, RA, as follows from equilibrium on Figure4.9 (ignoring
the distributed load):
M B = 0 = R A L M B M B = R A L
Fy = 0 = RA + RB RB = RA
The M/EI diagram and the deflected shape for the vertical reaction, RA, is
shown in Figure4.11.
tAB
RAL
RA
RA
2L/3
RA L
EI
M
x dx = Ax1
EI A
A
A = 0 = t AB =
1 wL2 3L 1 RA L 2 L
3wL
0 = L
+ L
RA =
2 EI
3
8
3 2 EI 4
From statics on the original freebody diagram in Figure4.9, the other
reactions at B can be found.
Fy = 0 = RA + RB wL RB =
M B = 0 = R A L M B +
5wL
8
wL2
wL2
MB =
2
8
CONJUGATE
Pin
Pin
Roller
Roller
Fixed
Free
Free
Fixed
Internal pin
Hinge
Internal roller
Hinge
Hinge
Internal roller
CONJUGATE
observing that the shear in the conjugate beam equals the rotations in the
real beam and the moment in the conjugate equals the deflection in the
real beam. Therefore, the fixedend of the real beam becomes free in the
conjugate beam and the free end in the real beam becomes fixed in the
conjugate beam. The conjugate beam is loaded with the M/EI diagram of
the real beam. This is a triangular moment diagram due to the applied load
P divided by EI. The last beam in Figure4.14 is a freebody diagram of the
conjugate beam with the area of the M/EI diagram equated to a point load
located at the centroid of the diagram. Static equilibrium can then be performed on the conjugate beam to find shear and moment at point B, which
represents rotation and deflection in the real beam shown in Figure4.15.
P
A
L/2
L/2
P
P
M=PL/2
L/2
L/2
L/2
L/2
PL
2EI
L/6
5L/6
MB
VB
PL2
8EI
Fy = 0 =
qB =
PL2
PL2
VB VB =
8 EI
8 EI
PL2
8 EI
PL2 5 L
5 PL3
+ MB MB =
8 EI 6
48 EI
3
5 PL
B =
48 EI
M B = 0 =
B
B
q or =
0
mM
dx
EI
(4.7)
The value of M is the moment equation due to the real loads on the structure. The value of m is the moment equation of the virtual load. Rotation or
deflection may be found depending on whether a virtual moment or force
is applied at the point under consideration.
Example4.8 Virtual work
Determine the vertical deflection at the free end of the uniformly loaded,
cantilever beam in Figure4.16 using the virtual work method.
A freebody diagram is drawn of the righthand side of the beam to
determine the internal moment in the beam. The uniformly distributed
load is represented as a point load equal to the area under the load and
located at the centroid of the area.
From static equilibrium, we can determine the internal moment at any
point x measured from the right end of the beam.
M=
wx 2
2
L
wx
x/2
M=  wx
2
A virtual force is applied at the point and in the direction that the deflection is desired. This is shown in Figure4.17 along with a freebody diagram of the righthand side of the beam. In this case, the internal moment
is represented as m.
w
1 kN
L
1 kN
m= 1x
wx 2
L
2
wx3
wL4
dx =
dx =
2 EI
8 EI
EI
0
1x )
L (
mM
dx =
EI
0
0
B =
C
8ft
B
16ft
Solve for the reactions due to the applied loads. The freebody diagram is shown in Figure4.19.
M B = 0 = 8k (16 ft ) Ay ( 24 ft ) + 40k ft Ay = 7 k
Fy = 0 = Ay + By 8k By = 1k
The shear and moment diagrams are constructed in Figure4.19. The
moment diagram is broken into unique areas that will be continuous over
the area of the virtual moment diagrams shown in Figure4.20.
Apply a virtual force dQ = 1 at point C as shown in Figure4.20. Solve
for the reactions similar to the real loads. Draw the shear and moment
diagram locating the centroid of the real moment diagrams on the virtual
moment diagram.
8ft
1k
16ft
7
V(k)
2(8ft)/3
1
2(16ft)/3
56
16ft/2
40
M(kft)
Q=1
C
2
3
8ft
3.556
1
3
16ft
16
3
3.556
2.667
m(kft)
M B = 0 = 1k (16 ft ) Ay ( 24 ft ) Ay =
1
Fy = 0 = Ay + By 1k By = k
3
2
k
3
mM
dx
EI
0
dQ C =
1
1 (8 ft ) 56k ft (3.556k ft ) + (16 ft )16k ft (3.556k ft )
2
2
EI
1k C =
4.7CASTIGLIANOS THEOREMS
In 1879 Alberto Castigliano published his two theorems on elastic structures that are known as Castiglianos Theorems (Castigliano 1879). The
first theorem states that the first partial derivative of strain energy with
respect to a particular deflection component is equal to the force applied at
the point and in the direction corresponding to that deflection component.
This may be written in mathematical terms as shown in Equation 4.8. The
second theorem is used more often in statically indeterminate structural
analysis and states that the first partial derivative of strain energy with
respect to a particular force is equal to the displacement of the point of
application of that force in the direction of its line of action. This is shown
in Equation 4.9. The equations are written in terms for flexural energy,
M/EI, of a particular rotation, qA, and moment, MA, relationship and in
terms of a particular deflection, DA, and force, PA, relationship. They can
be written for any elastic force and deformation relationship.
L
M A = M
0
q A = M
0
M dx
M dx
and PA = M
qA EI
A EI
0
L
M dx
M dx
and A = M
M A EI
PA EI
0
(4.8)
(4.9)
L
P
x/2
M
V
M B = 0 = M wx
M
= x
P
L
B = M
0
B =
x
wx 2
Px M =
Px
2
2
L
L
wx 2
wx 3
dx
M dx
dx
=
Px ( x )
=
+ Px 2
P EI 0 2
EI 0 2
EI
wL4 PL3
+
8 EI 3EI
Since P is a fictitious applied load placed on the beam only to find the
deflection, it can be removed and the deflection at B becomes the following:
B =
wL4
8 EI
X
A
B
L
w
2
wL PL
6
wL P
2
P
wx
L
M
x
MB =
1
L
wL
2
3
wL2
PL
6
(4.10)
1
Fy = 0 = RB + P wL
2
RB =
wL
P
2
(4.11)
A = 0 = M
0
1 wx x
wx 3
x
M =
Px
2 L 3
6L
L
L
wx 4
wx 3
dx
M dx
dx
=
=
+ Px 2
Px ( x )
P EI 0 6 L
EI 0 6 L
EI
wL5 PL3
0=
+
3
30 L
wL
P = RA =
10
This result can be substituted back into Equations 4.10 and 4.11 to find the
reactions at A.
2 wL
5
wL2
MB =
15
RB =
4 EI
2 EI
6 EI
fi +
fj
b
L
L
L
(4.12)
M j = FEM ji +
2 EI
4 EI
6 EI
fi +
fj
b
L
L
L
(4.13)
In these two equations, f is the rotation at the joint and b is the lateral
translation between the ends divided by the length of the member. It will
be seen later that the values 4EI/L, 2EI/L, and 6EI/L are flexural stiffness
terms. The FEM terms are the fixedend moments due to the loads on the
member. A special case may be used if one end of the member is pinned.
Equation 4.14 is for the i end of a member when the j end is pinned.
M i = FEM ij +
FEM ji
2
3EI
(fi b )
L
(4.14)
Example4.12 Slopedeflection
Determine the moments at the ends of the members of the continuous
beam in Figure4.24 using the slopedeflection equations.
Z
8 kips
1.2 kips/ft
8 kips
B
20ft
5ft
5ft
The fixedend moments due to the loads on each span are computed
and can be found in most elementary structural analysis textbooks.
1.2k /ft ( 20 ft )
wL2
=
= 40k ft
12
12
2
FEM AB =
FEM BA =
8 (5 ft ) (5 ft )
Pb 2 a
= 10k ft
=
2
L
(10 ft )2
2
FEM BC =
Pa 2b 8 (5 ft ) (5 ft )
= 2 =
= 10k ft
L
(10 ft )2
2
FEM CB
M AB = FEM AB +
M BA
M BC
Equilibrium equations are written at each joint that has a real rotation. In
this case, that is only joint B.
M BA + M BC = 0 = 40 +
fB =
50
fB
EI
EI
3EI
f B 15 +
fB
5
10
Substituting the value of fB back into the moment equation will result in
the final memberend moments.
M AB = 40 +
M BA = 40 +
EI 50
= 45k ft
10 EI
EI 50
= 30k ft
5 EI
M BC = 15 +
3EI 50
= 30k ft
10 EI
K=
4 EI
L
The joint stiffness factor is the sum of all the member stiffness factors for
the members connected at a joint.
K T = K
The distribution factor for each memberend at a joint is the member stiffness factor divided by the joint stiffness factor.
DF =
K
K
=
KT K
If a member is connected to a support and not to other members, the distribution factor is dependent on the support type. If the support is fixed
against rotation, then DF=1. If the support is free to rotate, then DF=0.
The member relative stiffness factor is used when a continuous beam
or frame is made from the same material when calculating the distribution
factor. This can be used in place of the member stiffness factor for calculation of the other factors.
KR =
I
L
The final factor is the carryover factor, which represents the fraction of a
moment that is carried over from one end of a member to the other. If the
member is prismatic, then the ratio of the far end moment to the near end
moment of a member is onehalf ().
3K
4
I AB
1
=
= 0.05
LAB 20
K BC =
I BC
1
=
= 0.10
LBC 10
K AB
0.05
=
= 0.333
K AB + K BC 0.05 + 0.10
DFBC =
K BC
0.1
=
= 0.667
K AB + K BC 0.05 + 0.10
8 kips
1.2 kips/ft
8 kips
B
20ft
5ft
C
5ft
FEM AB
1.2k /ft ( 20 ft )
wL2
=
=
= 40k ft
12
12
2
FEM BA =
FEM BC =
8 (5 ft ) (5 ft )
Pb 2 a
=
= 10k ft
2
L
(10 ft )2
2
FEM CB
Pa 2 b 8 (5 ft ) (5 ft )
= 2 =
= 10k ft
L
(10 ft )2
The momentdistribution is shown in Table4.1. In the first line, the fixedend moments are recorded. In the second line, the distribution factor is
multiplied times the negative sum of the fixedend moments. This is a distribution. In each subsequent onehalf, the end moment is carried over the
opposite end of the members. This is a carryover. Then, the unbalanced
moments at the joint are distributed again and carried over again.
Table4.1. Example4.13 Momentdistribution
0.00
0.333
0.667
40.00
40.00
10.00
10.00 Fixedend
1.00 Distribution
Factor
0.00
10.00
20.00
10.00 Distribution
5.00
0.00
5.00
10.00 Carryover
0.00
1.67
3.33
10.00 Distribution
0.83
0.00
5.00
1.67 Carryover
Moment
0.00
1.67
3.33
1.67 Distribution
0.83
0.00
0.83
1.67 Carryover
0.00
0.28
0.56
1.67 Distribution
0.14
0.00
0.83
0.28 Carryover
0.00
0.28
0.56
0.28 Distribution
0.14
0.00
0.14
0.28 Carryover
0.00
0.05
0.09
0.28 Distribution
0.02
0.00
0.14
0.05 Carryover
0.00
0.05
0.09
0.05 Distribution
0.02
0.00
0.02
0.05 Carryover
45.00
30.00
30.02
Example4.14 Momentdistribution
Determine the moments in the beam shown in Figure4.25 by the modified
momentdistribution method.
There are two primary differences when using this method. First, use
a modified member stiffness factor for member BC.
K BC
=
3K BC 3 ( 0.1)
=
= 0.075
4
4
K AB
0.05
=
= 0.40
K AB + K BC 0.05 + 0.75
DFBC =
K BC
0.075
=
= 0.60
K AB + K BC 0.05 + 0.075
Second, the fixedend moment at the pinned support is balanced then carried over to the far end before the momentdistribution process begins.
The process is shown in Table4.2.
Table4.2. Example4.14 Momentdistribution
0.00
0.400 0.600
1.00 Distribution
40.00 10.00
10.00 Fixedend
40.00
5.00
10.00 Balance @ C
0.00
Distribution
10.00 15.00
0.00
0.00
Carryover
5.00
30.00 30.00
0.00 Final Moments
45.00
Factor
Moment
1
1
[ K ][ ] = [ P ]
The derivation of elastic member stiffness in the XZ system will be
derived in the two following examples using the conjugate beam method
and area moment method.
Example4.15 qiy stiffness
Derive the qiy stiffness using the conjugate beam method for a linear
member.
A freebody diagram is shown in Figure4.27 with an imposed rotation of 1 unit on the iend of the member. The moments are assumed in the
Piz
Miy
Mjy
iy
Miy
EIy
PizL
EIy
positive y direction using the righthand rule and the Cartesian righthand
coordinate system. The forces are shown consistent with the deformation.
The moment diagram divided by EI is shown for the reaction forces on the
iend of the member.
The conjugate beam can be constructed for the two basic assumptions. The shear in the conjugate is equal to the slope of the real beam, and
the moment of the conjugate is equal to the deflection of the real beam.
Viconj = qireal = qiy
M iconj = ireal = 0
Since moment in the conjugate does not exist but shear does, the conjugate
beam is pinned on the iend.
V jconj = q jreal = 0
M jconj = jreal = 0
Since both the shear and moment in the conjugate do not exist, the conjugate beam is free on the jend. The resulting conjugate beam is shown in
Figure4.28.
The conjugate beam method can be applied to find the reactions at the
iend of the conjugate beam, which are equal to deformations at the iend
of the real beam. The load from Figure4.27 is applied to the conjugate
beam in Figure4.28.
M iy L
EI y
Piz L2
2 EI y
M iy L L Piz L2 2 L
M iconj = ireal = 0 =
+
EI y 2 2 EI y 3
Solving the second equation for Piz in terms of Miy and then substituting
into the first equation, the stiffness value can be found.
Piz =
3M iy
qiy =
M iy =
Piz =
2L
M iy L
EI y
4 EI y
L
6 EI y
L2
3M iy L
4 EI y
M iy L
4 EI y
qiy
(4.15)
qiy
It should be noted that the force Piz was actually shown as negative in the
original freebody diagram so it does have a negative value for stiffness.
Piz =
6 EI y
L2
qiy
(4.16)
qij = 0 = q j qi =
Piz =
2M iy
L
M iy L
EI y
Piz L2
2 EI y
Mjy
Miy
Miy
EIy
PizL
EIy
The tangential deviation of a point at the iend from a tangent to the curve
on the jend is the implied deflection. This is equal to the moment of the
area of the M/EI diagram about the point at the iend.
j
tij = iz =
M iy L L Piz L2 2 L
M
xi dx =
+
EI
EI y 2 2 EI y 3
Substituting the first equation for Piz into the Diz equation results in one
of the stiffness terms. The second term is found by substituting the first
stiffness term back into the Piz equation. Note that Miy was assumed as
negative in the freebody diagram, so the sign must be switched.
iz =
M iy L2
6 EI y
6 EI y
M iy =
Piz =
L2
12 EI y
L3
iz
iz
(4.17)
(4.18)
The four terms given in Equations 4.15 through 4.18 are the flexural stiffness terms for the forces at the iend due to motions at the iend. This is
denoted as stiffness matrix [Kii] in Equation 4.19. The stiffness equation
and matrix form of this are as follows:
[ K ][ d ] = [ F ]
ii
6 EI y
12 EI y
2
3
L iz Piz
L
=
4 EI y qiy M iy
6 EI y
2
L
L
6 EI y
12 EI y
2
[ K ][ d ] = [ F ]
ii
6 EI y
12 EI y
2
3
L iz Piz
L
=
4 EI y qiy M iy
6 EI y
L
L2
6 EI y
12 EI y
2
3
[ Kii ] = 6LEI 4 EIL
y
y
2
L
L
(4.19)
The transmission matrix derived in Section 4.3 for the XZ system can be
used to find the forces at the jend.
1 0
1
[T ] = L
From statics equilibrium, the forces at the iend transmitted to the jend
plus the forces at the jend are equal to zero. The forces at the jend due to
motions at the iend are denoted as stiffness matrix [Kji].
[T ][ Kii ] + K ji = 0
K ji = [T ][ K ii ]
6 EI y
12 EI y
12 EI y
2
3
1 0
L 1 0 L3
L
K ji =
=
4 EI y L 1 6 EI y
L 1 6 EI y
2
2
L
L
L
12 EI y 6 EI y
3
L
L2
K ji =
6 EI y 2EI y
2
L
L
6 EI y
L2
4 EI y
(4.20)
For a prismatic member, it can be shown that the stiffness matrix is symmetric about the main diagonal. This results in stiffness [Kij] being equal
to the transpose of [Kji], where the forces at the jend due to motions at the
iend are denoted as stiffness matrix [Kji].
12 EI y 6 EI y
12 EI y
3
3
2
T
L
L
L
K ij = K ji =
=
6 EI y 2 EI y
6 EI y
2
L
L
L2
6 EI y
12 EI y
2
3
L
L
K ij =
2 EI y
6 EI y
L
L2
6 EI y
L2
2 EI y
(4.21)
From statics equilibrium, the forces at the iend transmitted to the jend
plus the forces at the jend are equal to zero. The forces at the jend due to
motions at the jend are denoted as stiffness matrix [Kjj].
[T ] Kij + K jj = 0
K jj = [T ] K ij
12 EI y
3
1 0
L
K jj =
6 EI
1
L
L2
12 EI y 6 EI y
3
L2
K jj = L
6 EI y 4 EI y
2
L
L
6 EI y
12 EI y
3
2
L
L 1 0
=
2 EI y L 1 6 EI y
L
L2
6 EI y
L2
2 EI y
(4.22)
Adding the axial stiffness terms that are derived in most strength of materials textbooks, the entire coplanar XZ frame stiffness matrix can be
found.
Ax E
ix
L
AE
Pjx = x ix
L
Ax E
Pix =
jx
L
AE
Pjx = x jx
L
Pix =
Ax E
L
Ax E
L
0
12 EI y
3
L
6 EI y
6 EI y
2
L
4 EI y
L2
12 EI y
6 EI y
2
L
6 EI y
L
2 EI y
L2
Ax E
L
0
0
Ax E
L
0
0
12 EI y
L3
6 EI y
L2
0
12 EI y
L3
6 EI y
L2
6 EI y
2 P
ixix
ix
L
2 EI y iz Piziz
L qiy Miyiy
==
jx Pjxjx
0
jz Pjz
jz
6 EI y q jy M jy
jy
L2
4 EI y
L
0
(4.23)
Pjy
Mjz
Miz
iz
Miz
EIz
PiyL
EIz
Since both ends of the beam are fixed for translation, the tangential
deviation of a point at the iend from a tangent to the curve on the jend is
zero. This is equal to the moment of the area of the M/EI diagram about
the point at the iend.
j
tij = 0 =
i
2
M L L Piy L 2 L
M
xi dx = iz +
EI
EI z 2 2 EI z 3
3M i z
Piy =
2L
The change in rotation from the iend to the jend is equal to the negative
of the implied rotation. This is the area under the M/EI diagram between
those points.
qij = 0 qiz = qiz =
2
M iz L Piy L
+
EI z
2 EI z
Substituting the first equation for Piy into the qiz equation results in one of
the stiffness terms. The second term is found by substitution of the first
stiffness term back into the Piy equation.
qiz =
M iz L
4 EI z
4 EI z
qiz
L
6 EI
Piy = 2 z qiz
L
M iz =
(4.24)
(4.25)
Mjz
Miz
Miz
EIz
PiyL
EIz
The conjugate beam can be constructed for the two basic assumptions. The shear in the conjugate is equal to the slope of the real beam, and
the moment of the conjugate is equal to the deflection of the real beam.
Viconj = qireal = 0
M iconj = ireal = iy
Since moment in the conjugate exists but shear is zero, the conjugate beam
is slotted in the y direction on the iend. This is a connection that is free to
move vertically, but restrained from rotation.
V jconj = q jreal = 0
M jconj = jreal = 0
Since both the shear and the moment in the conjugate do not exist, the conjugate beam is free on the jend. The resulting conjugate beam is shown
in Figure4.32.
The conjugate beam method can be applied to find the reactions at the
iend of the conjugate beam, which are equal to deformations at the iend
of the real beam. The load from Figure4.31 is applied to the conjugate
beam in Figure4.32.
Viconj = qireal = 0 =
2
M iz L Piy L
+
EI z
2 EI z
2
M L L Piy L 2 L
M iconj = ireal = iy = iz +
EI z 2 2 EI z 3
Solving the first equation for Piy in terms of Miz and then substituting into
the second equation, the stiffness value can be found.
2 M iz
L
M L2
iy = iz
6 EI z
Piy =
M iz =
6 EI z
iy
L2
(4.26)
This can be substituted back into the equation for Piy to obtain the last
stiffness value.
Piy =
12 EI z
iy
L3
(4.27)
The four terms given in Equations 4.24 through 4.27 are the flexural stiffness terms for the forces at the iend due to motions at the iend. This is
denoted as stiffness matrix [Kii] in Equation 4.28. The stiffness equation
and matrix form of this are as follows:
[ K ][ d ] = [ F ]
ii
12 EI z
L3
6 EI z
L2
6 EI z
L2 iy Piy
=
4 EI z iz M iz
L
12 EI z 6 EI z
L3
L2
[ Kii ] = 6 EI
4 EI z
z
L2
L
(4.28)
The transmission matrix derived in Section 4.3 for the XY system can be
used to find the forces at the jend.
1
[T ] = L
0
1
From statics equilibrium, the forces at the iend transmitted to the jend
plus the forces at the jend are equal to zero. The forces at the jend due to
motions at the iend are denoted as stiffness matrix [Kji].
[T ][ Kii ] + K ji = 0
K ji = [T ][ K ii ]
12 EI z 6 EI z
12 EI z
2
L3
1
0
1 0 L3
L
K ji =
L 1 6 EI z 4 EI z L 1 6 EI z
L2
L2
L
6 EI
12 EI z
2z
L3
L
K ji =
EI
6
2
EI
z
z
L2
L
6 EI z
L2
4 EI z
L
(4.29)
For a prismatic member it can be shown that the stiffness matrix is symmetrical about the main diagonal. This results in stiffness [Kij] being equal
to the transpose of [Kji], where the forces at the jend due to motions at the
iend are denoted as stiffness matrix [Kji].
12 EI z
L3
T
K ij = K ji =
6 EI z
L2
12 EI y 6 EI y
3
L2
L
K ij =
6 EI y 2 EI y
2
6 EI z
12 EI z
2
L3
L
=
2 EI z
6 EI z
L
L2
6 EI z
L2
2 EI z
L
(4.30)
From statics equilibrium, the forces at the iend transmitted to the jend
plus the forces at the jend are equal to zero. The forces at the jend due to
motions at the jend are denoted as stiffness matrix [Kjj].
[T ] Kij + K jj = 0
K jj = [T ] K ij
12 EI y
3
1 0
L
K jj =
L 1 6 EI y
2
L
6 EI z
12 EI z
6 EI y
12 EI y
3
2
L 1 0
L
=
2 EI y L 1 6 EI y
2
L
L
6 EI y
L2
2 EI y
[T ] Kij + K jj = 0
K jj = [T ] K ij
12 EI y
3
1 0
L
K jj =
6
L
1
EI
y
2
L
6 EI
12 EI z
2z
L3
L
K jj =
4 EI z
6 EI z
L2
L
6 EI y
12 EI y
3
2
L 1 0
L
=
2 EI y L 1 6 EI y
2
L
L
6 EI y
L2
2 EI y
L
(4.31)
If we add the axial stiffness terms that were shown in the previous section,
the entire coplanar XY frame stiffness matrix can be found.
Ax E
L
AE
x
L
12 EI z
L3
6 EI z
L2
6 EI z
L2
4 EI z
L
12 EI z
L3
6 EI z
L2
6 EI z
L2
2 EI z
L
Ax E
L
0
0
Ax E
L
0
0
0
12 EI z
L3
6 EI
2z
L
0
12 EI z
L3
6 EI
2z
L
6 EI z
P
L2 ix ixix
iy
2 EI z Piyiy
L qiz Miziz
== P
0 jx jxjx
jy Pjyjy
6 EI z q jz Mjz
2 jz
L
4 EI z
L
0
(4.32)
qix
L
IG
M jx = x qix
L
I xG
M ix =
qix
L
IG
M jx = x qix
L
M ix =
AE
Ax E
0
0
0
0
0 x
0
0
0
0
0
L
L
6 EI z
12 EI
6 EI z
0 12 EI z
0
0
0
0 3 z
0
0
0
3
2
2
L
L
L
L
12
6
12
6
E
I
EI
EI
EI
y
y
y
y
0
0
0
0
0
0
0
0
3
2
3
2
L
L
L
L
ix Pix
I xG
I xG
0
0
0
0
0
0
0
0
0
0 iy Piy
L
L
iz Piz
6 EI y
4 EI y
6 EI y
2 EI y
0
0
2
0
0
0
0
0
0 qix M ix
2
L
L
L
L
M
q
6 EI z
4 EI z
6 EI
2 EI z iy iy
0
0
0
0
0 2z
0
0
0
q
2
L
L
L
L iz = M iz
jx Pjx
AE
Ax E
x
0
0
0
0
0
0
0
0
0
0
Pjy
L
jy
L
12 EI
6 EI
12 EI z
6 EI
P
0
0
0 2z 0
0
0
0 2 z jz jz
0 3 z
3
L
L
L
L
q jx
M jx
12 EI y
6 EI y
12 EI y
6 EI y
q
M
jy
0 3
0
0
0
0
0
0 jy
0
2
3
2
L
L
L
L
q jz M jz
I xG
I xG
0
0
0
0
0
0
0
0
0
0
L
L
6 EI y
2 EI y
6 EI y
4 EI y
0
0
0
0
0
0
0
0
L2
L
L2
L
6 EI z
2 EI z
6 EI z
4 EI z
0
0
0
0
0
0
0
0
L2
L
L2
L
(4.33)
be transformed into this system. The basic process to set up and solve the
structure using this stiffness is summarized as follows:
1. Find the local member stiffness, [Km], using Equation 4.33 (4.23
or 4.32 for 2D systems). Rotate the local member stiffness to the
global joint stiffness system, [Kg]. This is shown in Equation 4.34.
K g = [ R ] [ K m ][ R ]
T
(4.34)
2. Assemble all of the members into the global joint stiffness matrix.
This is done by using joint labeling to order the matrix.
3. Determine the global joint loading, [Pg], from all direct loads on
joints, [P & Mg], and loads on members. The member loads are
applied as the opposite of the fixedend forces and moments,
[FEPMm]. These must be rotated from the local system to the global
system, [R]T.
Pg = P & M g [ FEPM m ][ R ]
(4.35)
(4.36)
5. Determine the reactions at the support, [Pg], from the global deformations. Any fixedend forces must be subtracted from the results.
Only the forces at the supports due to the free deformations need
to be found.
Pg = K g g + [ FEPM m ][ R ]
(4.37)
6. Solve for the local member forces and moments, [P & Mm], for each
member separately from the global joint deformations. The global
joint deformations must be rotated into the member system, [R].
The fixedend forces and moments must be added back to get the
final member end forces.
[ P & M ] = [ K ][ R ]
m
+ [ FEPM m ]
(4.38)
The general setup for the stiffness method of analysis represents a system of linear equations, where the displacement vector is the unknown.
Except for those designated as supports, there are six unknown joint displacement components for each joint in the structure in a 3D structure.
Each displacement released at a support is still an unknown displacement
component to the system. There is an equation for each degree of freedom
of the structure. Each nonrelated component at a support has a displacement that is set identically equal to zero, and as far as the system of equations is concerned, this particular equation may be omitted, along with
any coefficient in the other equations which corresponds to the dropped
displacement.
Sometimes the system solution is handled in six by six blocks of
coefficients or six rows of equations at a time, where each block representing the accumulated stiffness for a joint in the case of the diagonal,
or the carryover effects from other joints in the case of offdiagonals.
In this case, unless the support joint is fully restrained, its corresponding row of six by six blocks is maintained intact and a number of sufficient sizes to simulate infinite stiffness in the restrained direction
are added to the diagonal of the diagonal block in the master stiffness
matrix.
The building of the global joint stiffness matrix consists of various
stages of operations. First, the member stiffness matrix is defined in its
own system, giving due consideration to member end releases, for each
member in the structure. This will be discussed in Chapter 5. The member
stiffness matrix can be considered as four separate six by six blocks. These
represent the forces at the ends due to the motions at the end and were
discussed in Sections 4.10 and 4.11.
K ii
[K ] = K
m
ji
K ij
K jj
[ K ][ ] = [ P ]
m
This equation represents the force in the local system due to deformations
in the local system. To go from local to global we multiply by the rotation
transpose, [R]T, on both sides of the equation.
[ K m ][ R ] [ m ] = [ R ] [ Pm ]
T
The right side of the equation now represents the forces in the global system, [Pg]. The left side represents the force in the global system due to
local deformations. The equation needs to be written in terms of the global
deformation, [Dg]. From Section 4.3, the local deformation is the global
deformation multiplied by the rotation matrix, [R].
[ R ]T [ K m ][ R ] g = Pg
Example4.19 Global joint stiffness
Determine the global joint deformations, support reactions, and local
member forces for the pinconnected bracing structure loaded as shown
in Figure4.33.
The area of each member, Ax, is 10 in2 and the modulus of elasticity,
E, is 10,000 ksi. Note that the structure is in the XZ coordinate system.
Since this is a pin connected structure loaded only at the joint, it will
act as a true truss with only axial forces in the members. The stiffness
model will be simplified to only include the axial stiffness components,
AE/L. Furthermore, rotation at the joints will be excluded since there is no
rotational stiffness imparted by the members. Rotation could be included
Z
8k
8k
9k
12k
15'0"
X
20'0"
4
1
2
4
if it is desired to find the rotation of the member ends, but this can easily
be found from the final deformed position of the structure.
A numbering system must be assigned to the joint and members for
easy bookkeeping. Figure 4.34 shows a numbering system for the five
joints and the seven members of the structure. The member numbers are
circled for clarity.
The member stiffness for each member is first found from Equation 4.34
(step 1).
K g = [ R ] [ K m ][ R ]
T
K g = [ b ] [ K m ][ b ]
T
Expand the general local member stiffness to include just the x and z
forces and motions. Also note that rotation transformation of members
will be about the yaxis or b.
AE
0
0 L
cos b sin b
sin b cos b
0
0 0
K g =
0
0
cos b sin b AE
0
sin b cos b L
0
0
cos b sin b
sin b cos b
0
0
0
0
AE
L
0
AE
L
0
0
0
0
0
0
0
0
0
cos b sin b
sin b cos b
cos 2 b
sin b cos b
sin 2 b
AE sin b cos b
K g =
L cos 2 b
sin b cos b
sin 2 b
sin b cos b
cos 2 b
sin b cos b
cos 2 b
sin b cos b
sin b cos b
sin 2 b
sin b cos b
sin 2 b
0 555.6
0
0
0 555.6 K11
=
0
0 K 41
0 555.6
K14 K 33
=
K 44 K 53
K 35
K 55
Member 3 does not need rotations since it is already in the global system
orientation. The global stiffness will be the same as the member stiffness.
Selecting the left end as the iend, the rotation is 0. The axial stiffness is
416.7 k/in.
416.7
0
K
=
[ 45 ] 416.7
0 416.7
0
0
0 416.7
0
0
0
0 K 44
=
0 K 54
K 45
K 55
Members 4 and 5 have the same orientation. Selecting the bottomleft end
as the iend, the rotation is 36.87 or 323.13. Instead of using angles, in
this case it is easier to use trigonometry directly. The cosine is 0.8 and the
sine is 0.6. The axial stiffness is 666.7 k/in.
426.7 320.0 426.7 320.0
320.0
240.0 320.0 240.0 K
[ K12 ] = [ K 25 ] = 426.7 320.0 426.7 320.0 = K11
21
240.0
320.0 240.0 320.0
K 22
=
K 52
K 25
K 55
K12
K 22
Members 6 and 7 have the same orientation. Selecting the topleft end as
the iend, the rotation is 36.87 or 323.13. Instead of using angles, in
this case it is easier to use trigonometry directly. The cosine is 0.8 and the
sine is 0.6. The axial stiffness is 666.7 k/in.
426.7 320.0 426.7 320.0
320.0 240.0 320.0 240.0
K
[ K 23 ] = [ K 42 ] = 426.7 320.0 426.7 320.0 = K 22
32
320
.
0
240
.
0
320
.
0
240
.
0
K 44
=
K 24
K 23
K 33
K 42
K 22
The global joint stiffness matrix can be assembled using each of the members contributions (step 2).
K11
K
21
K g = 0
K 41
0
427 320
320 796
427 320
320 240
0
0
K g =
0
0
0
0
0
566
0
0
0
0
427
320
1707
0
427
320
427
320
427
320
K12
K 22
K 32
K 42
K 52
0
K 23
K 33
0
K 53
K14
K 24
0
K 44
K 54
0
K 25
K 35
K 45
K 55
0
0
0
0
0
0
320
0
0
0
0
0
240
566
0
427 320 427 320 427 320
0
0
0
240 320 796
566
320
0
0
843 320 417
0
0
0
240
320 796
0
0
0
0
0
843 320
320
417
240
0
566
0
0
320
795
The global joint loading can be determined from Equation 4.35 directly
since all of the applied loads are at the joints and in the global system
(step 3). Note that there are no loads applied directly to the members, so
there are no fixedend forces and moments. The load matrix is in units of
kips (k).
Pg = P & M g [ FEPM m ][ R ]
P1x 0
P
1y 0
P2 x 0
P2 y 0
P3 x 0
= =
P3 y 0
P 12
4x
P4 y 8
P5 x 9
P6 y 8
g = K g Pg
1
0
427 320 427 320 0 0.0246
2 x 1707
960 320 240 320 240 0 0.0057
2y 0
4 x 427 320
843 320 4117
0 12 0.0595
=
320
240
320
796
0
0 8 0.0022
4y
5 x 427 320 417
0
843 320 9 0.0602
0
0
320
795 8 0.0261
6 y 320 240
The reactions at the support can be found using the solution of the global
deformation with Equation 4.37 (step 5). Only the terms in the rows corresponding to the restrained degrees of freedom and in the columns of
the unrestrained degrees of freedom need to be included. Since there are
no applied loads on the members, the fixedend forces and moments are
omitted. In addition, there are no applied loads at the support locations so
the global applied forces and moments are omitted. The reaction forces
are in kips (k).
[ P ] = K g g P & M g [ FEPM m ][ R ]T
0
0
427 320 0
320 240 0 566 0
[ P ] = K g g = 427 320 0 0 0
0.0246
0 0.0057
0 0.0595
0 0.0022
[ P ] = K g g P & M g [ FEPM m ][ R ]T
427 320
320 240
[ P ] = K g g = 427 320
320 240
0
0
0 566
0
0
0
0
0
0
0
0
0.0246
0 0.0057
0 0.0595
0 0.0022
566 0.0602
0.0261
8.67
7.75
=
12.33
23.75
The final step is finding the member forces for each of the members using
Equation 4.35 (step 6). Since there are no applied forces on the members, the fixedend forces and moments can be omitted. Note that the local
member stiffness matrix is used here and not the global matrix. The member force will be in kips. If the iend is positive, the member is in compression and if it is negative the member is in tension.
[ P & M ] = [ K ][ R ]
[ P & M ] = [ K ][ b ]
m
+ [ FEPM m ]
555.6
0
555.6
0
0
0
0 0
0 1
0 0
0 0
1 0
0 0
0 0
0 1
0 0 1.24
0 0 0
=
1 0.0595 1.24
0 0.0022 0
555
.
6
0
555.6
0
555.6
0
0
0
0 0
0 1
0 0
0 0
1 0
0 0
0 0
0 1
0 0 14.5
0 0 0
=
1 0.0602 14.5
0 0.0261 0
416.7
0 416.7
0
0
0 416.7
0
0
0 1
0 0
0 0
0 0
0
1
0
0
0
0
1
0
0 0.0595 0.3326
0 0.0022 0
=
0 0.0602 0.326
1 0.0261 0
666.7
0 666.7
0
0
0 666.7
0
0
0 0.8 0.6
0
0 0 10.8
0
0 0 0
0 0.6 0.8
=
0
0.8 0.6 0.0246 10.8
0 0
666.7
0 666.7
0
0
0 666.7
0
0
0 0.8 0.6
0
0 0.0246 10.8
0
0 0.0057 0
0 0.6 0.8
=
0
0.8 0.6 0.0602 10.8
0 0
666.7
0 666.7
0
0
0 666.7
0
0
0 0.8 0.6
0
0 0.0246 15.4
0
0 0.0057 0
0 0.6 0.8
=
0
0.8 0.6 0 15.4
0 0
0 0.6 0.8 0 0
0 0
666.7
0 666.7
0
0
0 666.7
0
0
0 0.8 0.6
0
0 0.0595 15.4
0
0 0.0024 0
0 0.6 0.8
=
0
0.8 0.6 0.0246 15.4
0 0
3 k/ft
18k
2
16'0"
24'0"
K g = [a ] [ K m ][a ]
T
Using the local member stiffness for the XY system from Section 4.11
which has translation in x and y direction and rotation about the z direction, the rotation transformation of members will be about the zaxis or a.
[a ]T
cos sin
sin cos
0
0
=
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0 cos sin
0 sin cos
0
0
0
0
0
0
0
0
AE
L
[ K m ] = AE
12 EI z
L3
6 EI z
L2
6 EI z
L2
4 EI z
L
12 EI z
L3
6 EI z
L2
0
0
6 EI z
L2
2 EI z
L
cosa sina
sina cosa
0
0
[a ] = 0
0
0
0
0
0
AE
L
AE
L
0
0
0
12 EI z
L3
6 EI
2z
L
0
12 EI z
L3
6 EI
2z
L
0
0
0
0
0
0
1
0
0
0 cosa sina
0 sina cosa
0
0
0
6 EI z
L2
2 EI z
L
6 EI z
2
L
4 EI z
L
0
0
0
0
0
0
Members 1 and 3 have the same orientation. Selecting the bottom end as
the iend, the rotation is 90. Table4.3 contains the local member stiffness
matrix, the rotation matrices, a and aT, and the global member stiffness
matrix.
K11
[K ] = [K ] = K
12
43
21
K12 K 44
=
K 22 K 34
K 43
K 33
Member 2 is already in the global system and does not need rotation.
Selecting the left end as the iend, the rotation is 0. Table4.4 contains
the local member stiffness matrix, the rotation matrices, a and aT, and the
global member stiffness matrix.
K 22
[K ] = K
23
32
K 23
K 33
0
49.17
4720
0
49.17
4720
0
4720
604166.67
0
4720
302083.33
0
1
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
49.2
0
0
1510
4720
0
4720
49.2
0
4720
0
0
1510
0
604166.67
4720
0
302083.33
KL
1510.4
0
0
1510.4
0
0
0
49.17
4720
0
49.17
4720
0
4720
302083.33
0
4720
604166.67
0
0
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
1
KG
49.2
0
0
1510
4720
0
4720
49.2
0
4720
0
0
1510
0
302083.33
4720
0
604166.67
aT
The global joint stiffness matrix can be assembled using each of the members
contributions (step 2). Table4.5 contains the global joint stiffness matrix.
The global joint loading is determined from Equation 4.35. In this
case, member 2 is loaded with a uniformly distributed load. The fixedend
forces and moments due to the load must be calculated. Normally, the
fixedend forces and moments are rotated into the global system before
they are placed in the global joint loading, but in this case, the member is
already in the global system and no rotation is necessary (step 3). The load
matrix is in units of kips and inches (kin).
1006.9
14.57
2098
14.57
2098
2098
402777.78
2098
201388.89
1006.9
1006.9
14.57
2098
14.57
2098
2098
201388.89
2098
402777.78
a
1
aT
1
1006.9
1006.9
15
2098
15
2098
2098
402777.78
2098
201388.89
1006.9
1006.9
KG
15
2098
15
2098
2098
201388.89
2098
402777.78
K12
K 22
K 32
0
K 23
K 33
K 43
0
0
K 34
K 44
FEP23 =
FEM 23 =
FEM 32 =
Pg = P & M g [ FEPM m ][ R ]
P1x 0
P 0
1y
M 1z 0
P2 x 18
P2 y 36
M 2 z 1728
=
=
P 0
3x
P3 y 36
M 3 z 1728
P4 x 0
P4 y 0
M 0
4z
The global deformations can be found from the global stiffness Equation 4.36 (step 4). The rows and columns corresponding to the support
constraint degrees of freedom must be deleted prior to the solution. This
would be all three motions at 1 and 4. The resulting matrix is just joints 2
and 3. This is shown in Table4.6 along with the reduced load. The solution
for the deformations will be in inches and radians.
2 x 0.3040
2 y 0.0207
q 0.0034
1
g = K g Pg = 2 z =
3 x 0.2852
3 y 0.0270
q 3 z 0.0010
PFEPM
1056
4720
1007
18
1525
2098
15
2098
36
4720
2098
1006944
2098
201389
1728
1007
1056
4720
15
2098
1525
2098
36
2098
201389
4720
2098
1006944
1728
The reactions at the supports can be found using the solution of the global
deformation with Equation 4.37 (step 5). Only the terms in the rows corresponding to the restrained degrees of freedom and in the columns of the
unrestrained degrees of freedom need to be included. Table4.7 shows the
appropriate stiffness terms and deformations needed to find the reactions.
Since there were no fixedend forces and moments at the support joints,
the solution is complete. The reaction forces are in kips and inches (kin).
[ P ] = K g g P & M g [ FEPM m ][ R ]T
[ P ] = K g g
Table4.7. Example4.20 Global joint stiffness
KG2
49
DG
4720
0.3040
P
0.92
1510
0.0207
31.22
4720
302083
0.0034
419.18
49
4720
0.2852
18.92
1510
0.0270
40.78
4720
302083
0.0010
1659.74
The final step is finding the member forces for each of the members using
Equation 4.35 (step 6). The member force will be in kips and inches.
The local member stiffness matrix and the rotation matrix were shown
in step1 and are omitted here. The sign convention for the XY system
applies when interpreting the finalend forces and moments.
[ P & M ] = [ K ][ R ]
[ P & M ] = [ K ][a ]
m
+ [ FEPM m ]
For member 1, the deformations at joints 1 and 2 are used. Table4.8 contains the final memberend forces in the local system along with global
deformations used to find those end forces.
Table4.8. Example4.20 Global joint stiffness
PL
DG
0
Dx1
31.22
kips
Dy1
0.92
kips
z1
419.18
kipin
0.3040
Dx2
31.22
kips
0.0207
Dy2
0.92
kips
0.0034
z2
596.37
kipin
For member 2, the deformations at joints 2 and 3 are used. Table4.9 contains the final memberend forces in the local system along with global
deformations used to find those end forces.
Table4.9. Example4.20 Global joint stiffness
DG
PL
0.3040
Dx2
18.92
0.0207
kips
Dy2
4.78
kips
0.0034
qz2
1131.63
kipin
0.2852
Dx3
18.92
kips
0.0270
Dy3
4.78
kips
0.0010
qz3
245.46
kipin
DG
0
Dx4
40.78
kips
Dy4
18.92
kips
qz4
1659.74
0.2852
Dx3
40.78
kips
0.0270
Dy3
18.92
kips
qz3
1973.46
0.0010
kipin
kipin
REFERENCES
Castigliano, A. 1879. Theorie de lEquilibre des Systemes Elastiques et ses Applications. Turin, Italy.
Cross, H. 1930. Analysis of Continuous Frames by Distributing FixedEnd
Moments. Proceedings of the American Society of Civil Engineers (ASCE),
Cleveland, Ohio.
Leibniz, G. 1695. Specimen Dynamicum.
Lewis, D.C.; Review: Frazer, R.A., Duncan, W.J., and Collar, A.R. 1939. Elementary Matrices and some Applications to Dynamics and Differential Equations.
New York, New York: Bulletin of the American Mathematical Society 45.
Maney, G.A. 1915. Studies in Engineering. Minneapolis, Minnesota: University
of Minnesota
MllerBreslau, H. 1875. Elementary Handbook of Strength of Materials.
Timoshenko, S.P. 1953. History of Strength of Materials. New York, New York:
McGrawHill.
CHAPTER 5
Advanced Structural
Stiffness
In this chapter, concepts learned in previous chapters are expanded and
applied to advanced structural stiffness. This method is applied in computers to solve complex structures that are statically determinate or statically
indeterminate. Matrices will be utilized to determine internal member
forces and displacements within a structure. Small pieces of the structure are analyzed and then compiled into a larger matrix in order to view
the structure as a whole. This procedure is the basis for finite element
analyses.
k1,1
0
0
0
0
k
7 ,1
0
0
0
0
0
0
k2, 2
0
0
0
k6 , 2
0
k8, 2
0
0
0
k12, 2
0 0
0 0
k3 , 3 0
0 k4, 4
k5 , 3 0
0 0
0 0
0 0
k9 , 3 0
0 k10, 4
k11,3 0
0 0
0
0
k3 , 5
0
k5 , 5
0
0
0
k9 , 5
0
k11,5
0
0
k2, 6
0
0
0
k6 , 6
0
k8, 6
0
0
0
k12, 6
k1, 7
0
0
0
0
0
k7 , 7
0
0
0
0
0
0
k 2 ,8
0
0
0
k6 ,8
0
k8,8
0
0
0
k12,8
0
0
0
0
k3, 9 0
0 k4,10
k5 , 9 0
0
0
0
0
0
0
k9 , 9 0
0 k10,10
k11,9 0
0
0
0
0
k3,11
0
k5,11
0
0
0
k9,11
0
k11,11
0
0 ix PPixix
k2,12 iy PPiyiy
0 iz PPiziz
0 qix Mixix
0 qiy Miyiy
k6,12 qiz Miziz
=
=
0 jx PPjxjx
k8,12 jy PPjyjy
0 jz PPjzjz
0 q jx M jxjx
0 q jy M jyjy
k12,12 q jz M jzjz
(5.1)
Example5.1 iz end release
Derive the local member stiffness for a Diz member end release using the
conjugate beam method.
A freebody diagram of the released beam is shown in Figure5.1.
Since the beam is allowed to move at the iend in the z direction, the
reaction Piz is equal to zero. The loaded conjugate beam is also shown in
Figure5.1. Note that the shear in the conjugate beam is equal to the rotation in the real beam and the moment in the conjugate beam is equal to the
deflection in the real beam.
Miy
Piz=0
Miy
iy
Miy
EIy
Pjz
Miy
EIy
iz=0
iz=0
iy
iy=0
Piz = M iy = Pjz = M jy = 0
k3,3 = k5,3 = k9,3 = k11,3 = 0
If a motion Djz is imposed, there is no resistance and therefore no forces.
The resulting forces are the stiffness values due to the motion. The following are forces and stiffness due to Djz:
Piz = M iy = Pjz = M jy = 0
k3,9 = k5,9 = k9,9 = k11,9 = 0
If a motion qiy is imposed, there is resistance and therefore forces. The
resulting forces are derived using conjugate beam.
Fz = qiy
M iy =
k5 , 5 =
EI y
L
EI y
L
M iy L
EI y
=0
qiy
qiy
M y = M iy + M jy = 0
M jy = M iy =
k11,5 =
EI y
L
Piz = Pjz = 0
EI y
L
qiy
qiy
k3 , 5 = k9 , 5 = 0
From symmetry of the stiffness matrix, the following terms can be
found:
k5,11 = k11,5
k5,11 =
EI y
L
q jy
M iy = M jy =
EI y
EI y
L
q jy
q jy
L
Piz = Pjz = 0
k11,11 =
k3,11 = k9,11 = 0
The resulting stiffness matrix is shown in Equation 5.2 with only the
affected terms replaced with the new values.
k1,1
0
k7 ,1
0
0
k2, 2
0
0
0
0
0
0
0
0
0
k4, 4
0 0 0
k6 , 2
0
k8, 2
0
0
0
0
0
0
0
0
0
0
0
k10, 4
0 0 0
k12, 2 0 0
0
0
0 k2, 6
0
0
0
0
EI y
0
L
0 k6 , 6
0
0
0 k8, 6
0
0
0
0
EI y
0
L
0 k12, 6
k1, 7 0
0 k 2 ,8
0 0
0 0
0
0
k7 , 7
0
0
0
0
0
0
0
0
0
0
0
k4,10
0 0 0
k6 ,8
0
k8,8
0
0
0
0
0
0
0
0
0
0
0
k10,10
0 0 0
0 k12,8 0 0
0
0
0 k2,12 ix
0
0 iy
0
0 iz
qix
EI y
0
L
qiy
0 k6,12 qiz
0
0 jx
0 k8,12 jy
0 jz
0
0
0 q jx
q jy
EI y
0
L
q jz
0 k12,12
Pix
P
iy
Piz
M ix
M
iy
M iz
=
P
jx
Pjy
Pjz
M jx
M jy
M
jz
(5.2)
Example5.2 qiy end release
Derive the local member stiffness for a qiy member end release using the
conjugate beam method.
A freebody diagram of the released beam is shown in Figure5.2.
Since the beam is allowed to rotate at the iend in the y direction, the
reaction Miy is equal to zero. The loaded conjugate beam is also shown in
Figure5.2. Note that the shear in the conjugate beam is equal to the rotation in the real beam and the moment in the conjugate beam is equal to the
deflection in the real beam.
If a motion Diz is imposed, there is resistance and therefore forces. The
resulting forces are derived using conjugate beam.
Miy=0
Mjy
iz
Piz
Pjz
PizL
EIy
iz
jz=0
iy
jy=0
M jy = iz +
Piz =
k3 , 3 =
3EI y
L3
3EI y
L3
Piz L L 2 L
=0
EI y 2 3
iz
iz
3EI y
3EI y
L2
iz
iz
L2
Fz = Piz + Pjz = 0
Pjz = Piz =
k9 , 3 =
M iy = 0
k5 , 3 = 0
3EI y
L3
3EI y
L3
iz
iz
jz
Piz
Pjz
Piz =
3EI y
k3 , 9 =
L3
3EI y
L3
jz
jz
M jy = Piz L =
k11,9 =
3EI y
L2
Pjz = Piz =
k9 , 9 =
3EI y
L3
3EI y
L2
jz
jz
3EI y
L3
jz
jz
Mjy
Miy=0
jy
Piz
Pjz
Mjy
EIy
iz=0
jz=0
iy
jy
3EI y
L
3EI y
L
M jy L 2 L
=0
EI y 2 3
q jy
q jy
M jy = Piz L + M jy = 0
Piz =
M jy
3EI y
L
3EI y
L2
q jy
L2
Fz = Piz + Pjz = 0
k3,11 =
Pjz = Piz =
k9,11 =
3EI y
M iy = 0
k5,11 = 0
L2
3EI y
L2
q jy
q jy
q jy
The resulting stiffness matrix is shown in Equation 5.3 with only the
affected terms replaced with the new values.
k1,1
0
0
0
k7 ,1
0
0
k 2 ,2
0
0
0
k 6 ,2
0
k8,2
0
0
0
k12,2
0
0
3EI y
0 0 0 k1,7 0
0 0 k2,6 0 k2,8
0
L3
k 4 ,4
0
0
0
0
0
0
0
0
0
3EI y
3
0
L
k10,4
0
3EI y
0
2
L
0
0
0 0
0
0
0
0
0
L3
0
0
0
0
0
3EI y
0 0 0
0 0 0
k6,6 0 k6,8
0 k7 ,7 0
k8,6 0 k8,8
0 0
0 0
0 0
0
0
3EI y
0
0
0
k4,10
0
0
0
0
0
L3
k10,10
0
3EI y
0
L2
0
0
0 k12,6 0 k12,8
0
0
0
k2,12
P
3EI y
0 ix ix
L2
iy Piy
0
0 iz Piz
0
0 q ix M ix
0
k6,12 q iy M iy
0
0 q iz = M iz
0
k8,12 jx Pjx
jy Pjy
3EI y
0
jz Pjz
L2
q M
0
0 jx jx
q jy M jy
3EI y
0 q jz M jz
L
k12,12
0
(5.3)
The member stiffness for releasing Djz and qjy can be derived in a similar manner to Diz and qiy. The resulting stiffness matrices are shown in
Equations 5.4 and 5.5 with only the affected terms replaced with the new
values.
k1,1
0
k7 ,1
0
0
k 2 ,2
0
0
0
0
0
0
0
0
0
k 4 ,4
0 0 0
k 6 ,2
0
k8,2
0
0
0
0
0
0
0
0
0
0
0
k10,4
0 0 0
k12,2 0 0
0
0
0
0
EI y
0 k1,7 0
k2,6 0 k2,8
0 0 0
0 0 0
0
L
0 k6,6
0
0
0 k8,6
0
0
0
0
EI y
0
L
0 k12,6
0 0
0 k6,8
k7 ,7 0
0 k8,8
0 0
0 0
0
0 0
0 0
0 0
0 k4,10
0
0 0
0 0
0 0
0 0
0 k10,10
0 0
0 k12,8 0
0
0
0 k2,12 ix Pix
0
0 iy Piy
P
0
0 iz
iz
q
EI y
ix M ix
0
M
L
qiy
iy
0 k6,12 qiz M iz
=
0
0 jx Pjx
0 k8,12 jy Pjy
0
0 jz Pjz
M
0
0 q jx
jx
EI y
q jy M
0 jy
L
q jz M jz
0 k12,12
(5.4)
k1,1
0
0
k
7 ,1
0
0
0
0
k 2 ,2
0
0
0
k 6 ,2
0
k8,2
0
0
0
k12,2
0
0
3EI y
0
0
0
L3
k 4 ,4
0
3EI y
0
2
L
0
0
0
0
0
0
3EI y
3
0
L
k10,4
0
0
0
0
0
0
0
0
k2,6
3EI y
2
0
L
0
0
3EI y
0
L
k6,6
0
0
0
0
k8,66
3EI y
0
L2
0
0
0
0
0
k12,6
k1,7 0
0 k2,8
0
0 k6,8
k7 ,7 0
0 k8,8
0
0 0
0 0
0 k12,8
0
0
3EI y
0
0
0
L3
0
3EI y
k4,10
L2
0
0
0
3EI y
0
0
0
L3
0
0
0
0
k10,10
0
0
0 0
0 k2,12
0 0
0 0
0 0
0 k6,12
0 0
0 k8,12
0 0
0 0
0 0
0 k12,12
ix Pix
P
iy iy
iz Piz
q ix M ix
q M
iy iy
q iz M iz
= P
jx jx
jy Pjy
jz Pjz
q M
jx jx
q jy M jy
q M
jz jz
(5.5)
When more than one of the four degrees of freedom is released on a flexural member, two conditions may exist. The first is that the member will
provide no joint stiffness. This occurs when both deflection and rotation
at either end are released, when rotation is released at both ends, or when
deflection is released at one end and rotation is released at the other end.
The stiffness matrix for this condition is shown in Equation 5.6. The
resulting beam is either a cantilever beam, a pinnedpinned beam, or a
pinnedslotted beam.
k1,1
0
0
0
0
k
7 ,1
0
0
0
0
0
0
k2, 2
0
0
0
k6 , 2
0
k8, 2
0
0
0
k12, 2
0 0
0 0
0 0
0 k4, 4
0 0
0 0
0 0
0 0
0 0
0 k10, 4
0 0
0 0
0
0
0
0
0
0
0
0
0
0
0
0
0 k1, 7
k2, 6 0
0
0
0
0
0
0
k6 , 6 0
0 k7 , 7
k8, 6 0
0
0
0
0
0
0
k12, 6 0
0
k 2 ,8
0
0
0
k6 ,8
0
k8,8
0
0
0
k12,8
0 0
0 0
0 0
0 k4,10
0 0
0 0
0 0
0 0
0 0
0 k10,10
0 0
0 0
0
0
0
0
0
0
0
0
0
0
0
0
0 ix Pix
k2,12 iy Piy
0 iz Piz
0 qix M ix
0 qiy M iy
k6,12 qiz M iz
=
0 jx Pjx
k8,12 jy Pjy
0 jz Pjz
0 q jx M jx
0 q jy M jy
k12,12 q jz M jz
(5.6)
qiy
Djz
qjy
Equation
0
1
0
0
0
1
1
1
0
0
0
0
1
1
1
1
0
0
1
0
0
1
0
0
1
1
0
1
0
1
1
1
0
0
0
1
0
0
1
0
1
0
1
1
1
0
1
1
0
0
0
0
1
0
0
1
0
1
1
1
1
1
0
1
4.33
5.2
5.3
5.4
5.5
5.6
Unstable
5.6
5.6
5.6
5.6
Unstable
Unstable
Unstable
Unstable
Unstable
values. Equation 5.11 contains the stiffness matrix used when the member
has no stiffness contribution in the case of the cantilever beam, a pinnedpinned beam, or a pinnedslotted beam.
k1,1
0
0
0
k7 ,1
0
0
0
0
0
0
0
0
0
0
k3,3
0
k5,3
0
0
0
k 4 ,4
0
0
0
k3,5
0
k5,5
0
0
0
k10,4
0
0
0
k9,5
0
k11,5
0 0
0 0
0 k9,3
0 0
0 k11,3
0
0
0
0
0
0
EI z
L
0
0
0
0
0
EI z
k1,7
0
0
0
0
0
0
0
0
0
0
0
k3,9
0
k5,9
k4,10
0
k3,11
0
k5,11
0
0
0
0
0
k7 ,7
0
0
0
0
0
0 0
0 0
0 k9,9
0 0
0 k11,9
0
0
0
0
0
0
k10,10
0
k9,11
0
k11,11
0
0 ix Pix
0 iy Piy
P
0 iz iz
0 qix M ix
EI z qiy M iy
q
M
L iz iz
=
0 jx Pjx
0 jy Pjy
0 jz Pjz
q M
0 jx
jx
0 q jy M jy
EI z q jz M jz
(5.7)
0
k1,1
EI
3
z
0
L3
0
0
0
0
0
0
0
0
0
k7 ,1
3EI
0 3z
L
0
0
0
0
0
0
3EI z
0
L2
0 k1,7
0 0
k3,3 0 k3,5
0 k 4 ,4 0
k5,3 0 k5,5
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
k7 ,7
0 0
k9,3 0 k9,5 0 0
0 k10,4 0 0 0
k11,3 0 k11,5 0 0
0
0 0
0
3EI z
3
L
0
0
0
0
0
3EI z
L3
0
0
0
3EI
2z
L
k3,9 0 k3,11
0 k4,10 0
k5,9 0 k5,11
0
0
0
0
0
0
0
k9,9 0 k9,11
0 k10,10 0
k11,9 0 k11,11
0
0
3EI z
L2
0
0
0
0
3EI
2z
L
0
0
0
3EI z
ix Pix
P
iy iy
iz Piz
q ix M ix
q M
iy iy
q iz M iz
= P
jx jx
jy Pjy
jz Pjz
q M
jx jx
q jy M jy
q M
jz jz
(5.8)
k1,1
0
0
0
k7 ,1
0
0
0
0
0 0
0
0
0 0
0
0
0 k3,3 0 k3,5
0 0 k 4 ,4 0
0 k5,3 0 k5,5
0
0 0
0
0
0 0
0
0
0 k9,3 0 k9,5
0 0 k10,4 0
0 k11,3 0 k11,5
0
0 k1,7
0
0
0
0
0
0
0
0
EI z
0
L
0 k7 ,7
0
0
0
0
0
0
0
0
EI
z 0
L
0 0
0
0
0 0
0
0
0 k3,9 0 k3,11
0 0 k4,10 0
0 k5,9 0 k5,11
0
0
0 ix Pix
0 iy Piy
0 iz Piz
0 qix M ix
q
EI z iy M iy
q
M
L iz = iz
0 jx Pjx
0 jy Pjy
0 jz Pjz
0 q jx M jx
0 q jy M jy
EI z q jz M jz
L
0 0
0
0
0 0
0
0
0 k9,9 0 k9,11
0 0 k10,10 0
0 k11,9 0 k11,11
0
(5.9)
0
k1,1
3EI z
0
L3
0
0
0
0
0
0
3EI z
0
L2
k7 ,1
0
3
0 EI3 z
L
0
0
0
0
0
0
0
0
k3,3 0 k3,5
0 k 4 ,4 0
k5,3 0 k5,5
0
k9,3 0 k9,5
0 k10,4 0
k11,3 0 k11,5
0 0 0
0
k1,7
0
3EI z
3EI z
0 3
L2
L
0
0
0
0
0
0
0
0
0
3EI z
3EI
0 2z
L
L
0
k7 ,7
0
3EI
3EI z
2z 0
L
L3
0
0
0
0
0
0
0
0
0
0
0
0
k3,9 0 k3,11
0 k4,10 0
k5,9 0 k5,11
0
k9,9 0 k9,11
0 k10,10 0
k11,9 0 k11,11
0
0
0
0 ix Pix
iy Piy
0
iz Piz
0 q M
ix
ix
0 q M
iy
iy
0 q iz M iz
jx = Pjx
0 P
jy jy
0 jz Pjz
q M
0 jx jx
q M
0 jy jy
0 q jz M jz
0
(5.10)
k1,1
0
0
0
0
0
k
7 ,1
0
0
0
0
0 0
0 0
0 k3,3
0 0
0 k5,3
0 0
0 0
0 0
0 k9,3
0 0
0 k11,3
0 0
0
0
0
k 4 ,4
0
0
0
0
0
k10,4
0
0
0
0
k3,5
0
k5,5
0
0
0
k9,5
0
k11,5
0
0 k1,7
0 0
0 0
0 0
0 0
0 0
0 k7 ,7
0 0
0 0
0 0
0 0
0 0
0 0
0 0
0 k3,9
0 0
0 k5,9
0 0
0 0
0 0
0 k9,9
0 0
0 k11,9
0 0
0
0
0
k4,10
0
0
0
0
0
k10,10
0
0
0
0
k3,11
0
k5,11
0
0
0
k9,11
0
k11,11
0
0 ix Pix
0 iy Piy
0 iz Piz
0 qix M ix
0 qiy M iy
0 qiz M iz
=
0 jx Pjx
0 jy Pjy
0 jz Pjz
0 q jx M jx
0 q jy M jy
0 q jz M jz
(5.11)
qiz
Djy
qjz
Equation
0
1
0
0
0
1
1
1
0
0
0
0
1
1
1
1
0
0
1
0
0
1
0
0
1
1
0
1
0
1
1
1
0
0
0
1
0
0
1
0
1
0
1
1
1
0
1
1
0
0
0
0
1
0
0
1
0
1
1
1
1
1
0
1
4.33
5.1
5.2
5.3
5.4
5.5
Unstable
5.5
5.5
5.5
5.5
Unstable
Unstable
Unstable
Unstable
Unstable
Table5.2 summarizes all the flexural stiffness conditions for the XY system. In this table, 1 indicates that the degree of freedom is released and 0
indicates that the degree of freedom is not released.
Hinge
3
X
2
3
12 ft
8 ft
12 ft
The member stiffness for each member is first found from Equation 4.34
(step 1). Member 2 will have a released stiffness matrix and will be created last. Selecting member 2 instead of member 1 will result in not having
to use modified fixedend forces and moments.
K g = [ R ] [ K m ][ R ]
T
K g = [a ] [ K m ][a ]
T
Using the local member stiffness for the XY system from Section 4.11,
which has translation in the x and y directions and rotation about the z direction, the rotation transformation of members will be about the zaxis or a.
[a ]T
cos sin
sin cos
0
0
=
0
0
0
0
0
0
AE
L
0
0
0
0
0
0
0
0
1
0
0
0 cos sin
0 sin cos
0
0
0
12 EI z
L3
6 EI z
6 EI z
L2
4 EI z
AE
L
0
0
0
0
0
0
0
1
0
12 EI z
L3
6 EI
2z
6 EI z
L2
2 EI z
[a ]T
cos sin
sin cos
0
0
=
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0 cos sin 0
0 Advanced
sin cosStructural
0
Stiffness
0
0
0
1
AE
AE
0
0
0
L
L
12 EI z
6 EI z
12 EI
0
3 z
0
L2
L
L3
6
6
EI
4
EI
EI
z
z
0
2z
0
L
L2
L
[ K m ] = AE
AE
0
0
0
L
L
12 EI
6 EI
12 EI z
0
3 z 2z
0
L
L
L3
6 EI
6 EI z
2 EI z
2z
0
0
L
L2
L
0
0
0
cos sin 0
sin cos 0
0
0
0
0
0
1
0
0
0
[a ] = 0
0
0 cos sin 0
0
0
0 sin cos 0
0
0
0
0
1
0
219
6 EI z
L2
2 EI z
L
6 EI z
2
L
4 EI z
L
0
Members 1, 4, and 5 are already in the global system and do not need
rotation. Selecting the left end as the iend, the rotation is 0. Tables 5.3
through 5.5 contain the local member stiffness matrices, the rotations
matrices, a and aT, and the global member stiffness matrices.
K 22
[K ] = K
21
12
K 44
[K ] = K
42
24
K 55
[K ] = K
53
35
K 21
K11
K 42
K 22
K 53
K 33
Member 3 is vertical and selecting the bottom end as the iend, the rotation
is 90. Table5.6 contains the local member stiffness matrix, the rotations
matrices, a and aT, and the global member stiffness matrix.
K 33
[K ] = K
32
23
K 32
K 22
Table5.3. Example5.3 M
ember stiffness, member 1
KL
1388.9
0.0
0.0
1388.9
0.0
0.0
0.0
40.2
2893.5
0.0
40.2
2893.5
0.0
2893.5
277777.8
0.0
2893.5
138888.9
1388.9
0.0
0.0
1388.9
0.0
0.0
0.0
40.2
2893.5
0.0
40.2
2893.5
0.0
2893.5
138888.9
0.0
2893.5
277777.8
a
1
aT
1
KG
1388.9
0.0
0.0
1388.9
0.0
0.0
0.0
40.2
2893.5
0.0
40.2
2893.5
0.0
2893.5
277777.8
0.0
2893.5
138888.9
1388.9
0.0
0.0
1388.9
0.0
0.0
40.2
2893.5
2893.5
277777.8
0.0
40.2
2893.5
0.0
0.0
2893.5
138888.9
0.0
For member 2, we will release the iend at joint 1 for rotation. Selecting
the iend at joint 1 the rotation is 213.7. Table5.7 contains the local member stiffness matrix, the rotation matrices, a and aT, and the global member stiffness matrix. The general released stiffness matrix is also shown.
K11
[K ] = K
13
31
K13
K 33
0.0
0.0
833.3
0.0
0.0
0.0
8.7
1041.7
0.0
0.0
1041.7
166666.7
0.0
8.7
1041.7
83333.3
833.3
0.0
0.0
833.3
0.0
0.0
8.7
0.0
1041.7
1041.7
0.0
8.7
1041.7
0.0
83333.3
0.0
1041.7
1041.7
166666.7
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
833.3
0.0
0.0
833.3
0.0
0.0
0.0
8.7
1041.7
0.0
8.7
1041.7
0.0
1041.7
166666.7
0.0
1041.7
83333.3
833.3
0.0
0.0
833.3
0.0
0.0
0.0
8.7
1041.7
1041.7
0.0
8.7
83333.3
0.0
1041.7
1041.7
166666.7
aT
KG
0.0
AE
L
0
[ K m ] = AE
0
3EI z
L3
0
0
3EI z
L3
3EI z
L2
AE
L
AE
L
3EI z
L3
0
0
3EI z
L3
3EI
2z
L
3EI z
L2
3EI z
2
L
3EI z
L
0
0.0
0.0
1388.9
0.0
0.0
0.0
40.2
2893.5
0.0
40.2
2893.5
0.0
2893.5
277778
0.0
2893.5
138888.9
1388.9
0.0
0.0
1388.9
0.0
0.0
0.0
40.2
2893.5
0.0
40.2
2893.5
0.0
2893.5
138888.9
0.0
2893.5
277778
aT
1
0.0
1388.9
0.0
0.0
40.2
2893.5
KG
1388.9
0.0
0.0
40.2
2893.5
0.0
0.0
2893.5
277778
0.0
2893.5
138888.9
1388.9
0.0
0.0
1388.9
0.0
0.0
40.2
2893.5
2893.5
277778
0.0
40.2
2893.5
0.0
0.0
2893.5
138888.9
0.0
The global joint stiffness matrix can be assembled using each of the members contributions (step 2). Table5.8 contains the global joint stiffness
matrix.
K11
K
21
K g = K 31
0
0
K12
K 22
K 32
K 42
0
K13
K 23
K 33
0
K 53
0
K 24
0
K 44
0
0
0
K 35
0
K 55
0.0
0.0
2083.3
0.0
0.0
0.0
135.6
6510.4
0.0
135.6
6510.4
0.0
6510.4
416666.7
0.0
6510.4
208333.3
2083.3
0.0
0.0
2083.3
0.0
0.0
0.0
135.6
6510.4
0.0
135.6
6510.4
0.0
6510.4
208333.3
0.0
6510.4
416666.7
a
6.12574E17
1
6.12574E17
6.12574E17
1
6.12574E17
aT
6.12574E17
1
6.12574E17
6.12574E17
1
6.12574E17
135.6
0.0
6510.4
135.6
0.0
6510.4
0.0
2083.3
0.0
0.0
2083.3
0.0
6510.4
0.0
416666.7
6510.4
0.0
208333.3
135.6
0.0
6510.4
135.6
0.0
6510.4
0.0
2083.3
0.0
0.0
2083.3
0.0
6510.4
0.0
208333.3
6510.4
0.0
416666.7
KG
The global joint loading is determined from Equation 4.35. In this case,
members 1 and 4 are loaded with a uniformly distributed load. The fixedend forces and moments due to the load must be calculated. Normally, the
fixedend forces and moments are rotated into the global system before
they are placed in the global joint loading, but in this case the member is
already in the global system and no rotation is necessary (step 3). The load
matrix is in units of kips and inches (kin).
0.0
0.0
1155.6
0.0
0.0
0.0
5.8
0.0
0.0
5.8
1001.6
0.0
1155.6
0.0
0.0
0.0
0.0
0.0
1155.6
0.0
0.0
0.0
0.0
0.0
5.8
0.0
0.0
5.8
0.0
1001.6
0.0
0.0
1001.6
1001.6
173343.8
0
0
0.832050294 0.554700196
0.554700196 0.832050294
0
0
0
0
1
0
0.832050294 0.554700196
0.554700196 0.832050294
0
0.554700196
0
0
aT
0.832050294 0.554700196
0.554700196 0.832050294
0
0
0
0
1
0
0.832050294
0.554700196 0.832050294
KG
801.8
530.7
530.7
359.6
0.0
0.0
801.8
530.7
530.7
359.6
555.6
833.4
0.0
801.8
530.7
555.6
0.0
530.7
359.6
833.4
0.0
0.0
0.0
0.0
0.0
801.8
530.7
555.6
0.0
530.7
359.6
833.4
0.0
555.6
833.4
173343.8
FEP42 =
wL 0.125k/in ( 240in )
= 15k
=
2
2
FEM 42 =
0.125k/in ( 240in )
wL2
= 600kin
=
FEM 24 =
12
12
wL 0.125k/in (144in )
FEP21 =
= 9k
=
2
2
2
FEM 21 =
wL2
0.125k/in (144in )
531
1389
531
400
2894
2894 138889
1389
0
0
802
531
556
2894 277778
802 531
0
556
0
0.125k/in
) 833 0 0
wL
0
6510
0 in6510
136 ( 240
=
FEP42 =
= 15k
0
2132
0
2 1852 0 2 2083 0
40 2894
9 1042
2
2861111 6510
0
208333
0
1042
83333
2894 138889 6510 1852
0.125k/in ( 240in )
wL
= 600kin
FEM 42 = 0
=
0
2326
531
0
531
136
1389 0
12 6510
12 7066
0
0 2083
0
531 2483 2060
0
360
40 2894
)
wL 0.125k/in ( 240in
=
=
=
FEP
15
k
0
6510
0
208333
867788
0
2894
138889
24
833
7066 2060
2
2
0
2358
833
833
2
2
0.125k/in ( 240in )0 9 1042
wL1042
9
= 600kin
=
=
24
12
0 1042 12
83333
0 1042 166667
144in )0
wL 0.125k/in
1389
1389 ( 0
FEP21 =
= 9k
=
2
0 2 40 2894
0
0
FEM
(
)
wL
=
= 216kin
12
12
wL 0.125k/in (144in )
FEP12 =
=
= 9k
2
2
FEM 21 =
0 2894
138889
0.125k/in
144in
0.125k/in (144in )
wL2
= 216kin
=
12
12
2
FEM12 =
Pg = P & M g [ FEPM m ][ R ]
P1x 0
P 9
1y
M 1z 216
P2 x 0
P2 y 24
M 2 z 384
P 0
3x
= P3 y = 0
M 3z 0
P4 x 0
P4 y 15
M 600
4z
P5 x 0
P 0
5y
M 5 z 0
40
2894
2894 277778
The global deformations can be found from the global stiffness Equation 4.36 (step 4). The rows and columns corresponding to the support
constraint degrees of freedom must be deleted prior to the solution. This
would be all three motions at 4 and the X and Z motions at joint 5. The
resulting matrix is shown in Table5.9 along with the reduced load. The
solution for the deformations will be in inches and radians.
1x 0.07505
2.45212
1y
q1z 0.00004
2 x 0.06168
2.25618
1
g = K g Pg = 2 y =
q 2 z 0.00261
0.03701
3x
3 y 2.25254
q 3 z 0.00449
q 5 z 0.02122
The reactions at the supports can be found using the solution of the global
deformation with Equation 4.37 (step 5). Only the terms in the rows corresponding to the restrained degrees of freedom and in the columns of the
unrestrained degrees of freedom need to be included. Table5.10 shows the
appropriate stiffness terms and deformations needed to find the reactions.
Since there are fixedend forces and moments at support joint 4, they must
Table5.9. Example5.3 Member stiffness
KG
2191
531
531
400
1389
0
0
FEPM
1389
531
556
802
2894
40
2894
531
360
833
9
2894 277778
2894 138889
216
6510
2083
24
208333
384
2358
6510
136
40
2894
2132
1852
802
531
531
136
360
2083
556
833
6510
6510
2326
531
7066
531
2483
2060 2894
0.0750
51.4
51.4
0 0 0
9
1042
2.4521
16.9
15
31.9
0 0 0
1042 83333
0.0000
0 0 0
1389
0.0617
51.4
51.4
0 0 0
2.2562
16.1
16.1
0.0026
0.0370
2.2525
0.0045
0.0212
be added back to get the final reactions. The reaction forces are in kips and
inches (kin).
[ P ] = K g g P & M g [ FEPM m ][ R ]T
[ P ] = K g g
The final step is finding the member forces for each of the members using
Equation 4.35 (step 6). The member force will be in kips and inches.
The local member stiffness matrix and the rotation matrix were shown
in step1 and are omitted here. The sign convention for the XY system
applies when interpreting the final end forces and moments.
[ P & M ] = [ K ][ R ]
[ P & M ] = [ K ][a ]
m
+ [ FEPM m ]
For member 1, the deformations at joints 2 and 1 are used. Since this
member had a load, the fixedend forces and moments must be added to
the results. Table5.11 contains the final member end forces in the local
system along with global deformations used to find those end forces.
For member 2, the deformations at joints 1 and 3 are used. Table5.12
contains the final member end forces in the local system along with global
deformations used to find those end forces.
For member 3, the deformations at joints 3 and 2 are used. Table5.13
contains the final member end forces in the local system along with global
deformations used to find those end forces.
0.06
2.26
0.00
0.08
2.45
0.00
Dx2
Dy2
PL
Final P
18.56
18.56
0.45
9.45
64.27
18.56
8.55
0.00
151.73
qz2
18.56
Dx1
0.45
Dy1
216.00
qz1
kips
kips
kipin
kips
kips
kipin
DG
0.08
2.45
0.00
0.04
2.25
0.00
20.19
Dx1
3.18
Dy1
0.00
qz1
20.19
Dx3
3.18
Dy3
550.21
qz3
Final P
20.19
kips
kips
3.18
0.00
kipin
kips
20.19
3.18
kips
550.21 kipin
DG
0.04
2.25
0.00
0.06
2.26
0.00
Dx3
Dy3
qz3
Dx2
Dy2
qz2
7.58
32.84
1772.68
7.58
32.84
1379.89
Final P
7.58
kips
kips
32.84
1772.68 kipin
kips
7.58
32.84
kips
1379.89 kipin
For member 4, the deformations at joints 4 and 2 are used. Since this
member had a load, the fixedend forces and moments must be added to
the results. Table5.14 contains the final member end forces in the local
system along with global deformations used to find those end forces.
For member 5, the deformations at joints 5 and 3 are used. Table5.15
contains the final member end forces in the local system along with global
deformations used to find those end forces.
Releases can also occur for the axial and torsional components of the
member stiffness. If either end of the member is released for Dx or qx, then
all the corresponding stiffness components are zero. If both ends of the
DG
0.00
0.00
0.00
0.06
Dx4
2.26
Dy2
0.00
Dy4
qz4
Dx2
qz2
51.40
16.87
2132.90
51.40
16.87
1915.62
Final P
kips
31.87
kips
2732.90 kipin
51.40
kips
kips
1.87
1315.62 kipin
51.40
0.00
0.00
Dx5
0.02
qz5
0.04
2.25
0.00
Dy5
Dx3
Dy3
qz3
PL
Final P
51.40
16.13
0.00
51.40
16.13
0.00
51.40
16.13
2322.89
kips
kips
kipin
51.40 kips
16.13 kips
2322.89 kipin
member are released, then the member is unstable. The axial components
of the member stiffness matrix are k1,1, k1,7, k7,1, and k7,7. The torsional
components of the member stiffness matrix are k4,4, k4,10, k10,4, and k10,10.
5.4NONPRISMATIC MEMBERS
Nonprismatic members have crosssectional properties that vary along
the length of the member. The stiffness of these members can be handled
in two ways. First, the member could be divided into prismatic sections
and modeled with several different members of constant crosssection.
Second, the member can be modeled with stiffness derived from the mathematical model of the crosssectional variation. The following derivation
is for the stiffness of a nonprismatic member in the XZ system.
Example5.4 Nonprismatic member stiffness
Derive the local member stiffness in the XZ system for a nonprismatic
crosssection using Castiglianos theorems.
Figure5.6 shows the beam with an applied deflection and an applied
rotation at the iend. These can be used simultaneously to derive the stiffness of the member.
Miy
iy
Piz
The internal moment, Mx, at any point, x, can be found from statics
and the partial derivatives of that moment can be found with respect to the
applied force and moment at the iend.
M x = Piz x M iy
dM x
= x
d Piz
dM x
= 1
d M iy
Castiglianos second theorem states that the first partial derivative of strain
energy with respect to a particular force is equal to the displacement of the
point of application of that force in the direction of its line of action. This
can be applied for both Diz and qiy.
iz = M x
d M x dx
dx
dx
= P x2
+ M x
d Piz EI y iz EI y iy EI y
qiy = M x
d M x dx
dx
dx
= Piz x
+ M iy
d M iy EI y
EI y
EI y
Since the crosssectional properties vary, the moment of inertia, Iy, varies. Let the values S1, S2, and S3 be used and substituted into the previous
two equations. These values can be prederived for the crosssectional
variation.
S1 =
dx
EI y
S2 = x
dx
EI y
S3 = x 2
dx
EI y
iz = Piz S3 + M iy S 2
qiy = Piz S2 + M iy S1
S1 =
dx
EI y
dx
S2 =Advanced
x EI y Structural Stiffness 231
S3 = x 2
dx
EI y
iz = Piz S3 + M iy S 2
qiy = Piz S 2 + M iy S1
Written in matrix form this is the flexibility matrix:
iz S3 S 2 Piz
q =
iy S 2 S1 M iy
[d i ] = [ f ii ][ Fi ]
We can solve this matrix equation for the forces Piz and Miy by the cofactor
method. Substituting the determinant of the flexibility matrix, D=S3S1
S22, to simplify the equation. This is the stiffness form of the equation.
Piz 1 S1
M =
iy D S 2
[ Fi ] = [ K ii ][d i ]
S 2 iz
S3 q iy
We can use the transmission matrix Equation 4.6 to find the forces at
the jend of the member, where the values of x are cause minus effect or
x=xixj=0L=L.
1 0 1 0
=
x 1 L 1
[T ] =
K ji = [T ][ K ii ]
1 0 S1
K ji =
L 1 S 2
S 2 1 S1
S2 1
=
S3 D S 2 S1 L S 2 L S3 D
S 2 iz
Pjz 1 S1
M =
jy D S 2 S1 L S 2 L S3 iy
Fj = K ji [d i ]
The carryover factor (COF) used in the moment distribution method can
be found by observing the ratio of the moment at the jend to the moment
at the iend.
COFi j
1
( S 2 S1 L ) iz + ( S 2 L S3 ) iy
D
=
=
1
M iy
S2 iz + S3iy
D
M jy
S 2 L S3
S3
The distribution factor (DF) used in the moment distribution method is the
ratio of the rotational stiffness of a member to the sum of the rotational
stiffness of all members at the joint. The rotational stiffness is the moment
at a joint due to the rotation at a joint. This is the term Kii for rotation and
moment only.
K Miy, q iy =
S3
D
We could find the deflection and rotation at the jend using the same
method. Alternatively, since the stiffness matrix is symmetric we can find
the forces at the iend due to motions at the jend directly.
T
S2
T
1 S1
1 S1
Kij = K ji =
= S
D S2 S1L S2 L S3
D 2
Piz 1 S1 S2 S1L jz
M =
iy D S2 S2 L S3 jy
S2 S1L
S2 L S3
Fi = Kij d i
We can use the transmission matrix equation again to find the forces at the
jend of the member.
K jj = [T ] K ij
1 0 S1
K jj =
L 1 S2
P jz 1 S1
=
M jy D S1L S2
Fj = K jj d j
S2 S1L 1 S1
=
S2 L S3 D S1L S2
S1L S2
jz
2
S1L 2 S2 L + S3 q jy
S1L S2
S1L 2 S2 L + S3 D
2
The COF from the jend to the iend is as follows considering the deformations at the jend:
COF j i =
M iy
M jy
1
S2 jz + ( S2 L S3 ) q jy
( S L S ) + S L2 2 S L + S q 1
jz
3
jy
2
1
2
1
D
S 2 L S3
S1L 2 S2 L + S3
S1 L2 2 S 2 L + S3
D
DF =
S1L2 2 S2 L + S33
D
S1L2 2S2 L + S3
D
The axial stiffness terms can be derived directly from strength of materials. The axial stiffness is the inverse of the flexibility, which can be written
as follows. The torsional stiffness would look the same as the axial stiffness with GIx substituted for EAx:
f =
dx
EA
0
[ Km ] = 1
S1
D
S
2
D
S2
D
S3
D
S1
D
S 2 S1L
D
S2
D
S 2 L S3
D
1
f
0
0
1
f
0
0
0
S1
D
S2
D
0
S1
D
S1L S2
D
S 2 S1L
S 2 L S3
S1L S 2
2
S1L 2 S2 L + S3
D
0
(5.12)
FIXEDEND MOMENTS
The fixedend forces and moments must be derived for a nonprismatic
member. The changes in stiffness along the length of the member will
change how the forces and moments are distributed by the member.
The following examples derive two of the most common member
loads.
Example5.5 Nonprismatic member stiffness
Derive the fixedend forces and moments due to a uniformly distributed load in the XZ system for a nonprismatic crosssection using
Castiglianos theorems.
FEMiy
FEPiz
FEMjy
FEPjz
L
The freebody diagram of the beam is shown in Figure5.7. The procedure for Castiglianos second theorem used in Example5.4 will be
repeated here.
The internal moment, Mx, at any point, x, can be found from statics
and the partial derivatives of that moment can be found with respect to the
applied force and moment at the iend.
Mx =
wx 2
FEPiz x FEM iy
2
dM x
= x
d FEPiz
dM x
= 1
d FEM iy
iz = 0 = M x
d M x dx
dx
dx
wx3 dx
= FEPiz x 2
+ FEM iy x
d FEPiz EI y
EI y
EI y
2 EI y
qiy = 0 = M x
d M x dx
dx
dx
wx 2 dx
= FEPiz x
+ FEM iy
EI y
EI y
d FEM iy EI y
2 EI y
Observing that there is a new term that varies with x, we will substitute S4
as follows:
S 4 = x 3
dx
EI y
w
S4
2
w
0 = FEPiz S 2 + FEM iy S1 S3
2
S3 S 2 FEPiz S 4 w
S S FEM = S 2
iy
2
1
3
0 = FEPiz S3 + FEM iy S 2
FEPiz 1 S1
FEM =
D S2
iy
S2 S4 w
w S1S 4 S 2 S3
=
S3 S3 2 2 D S32 S 2 S 4
wL2
FEPiz L FEM iy
2
FEPiz
FEMjy
FEPjz
The internal moment, Mx, at any point, x, can be found from statics
and the partial derivatives of that moment can be found with respect to the
applied force and moment at the iend. In this case, two moment equations
must be written. The first is Mx1, with x from the iend to the point load
(0 x a) and the second is Mx2, from the point load to the jend (a x L).
M x1 = FEPiz x FEM iy
M x 2 = P ( x a ) FEPiz x FEM iy
The partial derivatives are the same for either of the two moment equations.
dM x
= x
d FEPiz
dM x
= 1
d FEM iy
iz = 0 = M x
qiy = 0 = M x
L
d M x dx
dx
dx
dx
= FEPiz x 2
+ FEM iy x
Px ( x a )
d FEPiz EI y
EI y
EI y a
EI y
L
d M x dx
dx
dx
dx
= FEPiz x
+ FEM iy
P ( x a )
d FEM iy EI y
EI y
EI y a
EI y
Observing that there are two new terms that vary with x, we will substitute
S5 and S6 as follows:
S5 = ( x a )
dx
EI y
S6 = ( x 2 ax )
dx
EI y
D S 2 S3 S5
D S3 S5 S 2 S6
iy
X
100in
100in
E=10,000 ksi
S1 =
dx
=
EI y
S2 = x
100
dx
=
EI y
200
dx
dx
1 100 0 200 100
+
=
+
= 0.00015
100 E 100 200 E 10000 100
200
100
200
1 1002 02 2002 1002
xdx
xdx
+
=
+
2 ( 200 )
100 E 100 200 E 10000 2 (100 )
= 0.0125
S3 = x 2
dx
+
EI y
100
x 2 dx
1 1003 03 2003 1003
=
+
= 1.5
200 E 10000 3 (100 )
3 ( 200 )
100
200
x 2 dx
100 E +
0
100
200
dx 100 dx
dx 200 dx
dx
10000 200
200100
100
11 100
dx
f ==
.0015
==
++
==
10 ++
==00.0015
10EE 100
20
E
10000
20
20
10000
10
20
EEAx x 00 10
E
100
K ==
11
11
666.67
.67
==
==666
f 00.0015
.0015
Since the jend of the member is fixed, there is no need to build the entire
member stiffness matrix. The jend motions will be eliminated and only
the iend of the member stiffness needs to be developed.
1
f
[ Kii ] = 0
0
S1
D
S
2
D
0
666.67
0
0
S2
= 0
2.1818 181.81
D
0
181.81 21818
S3
D
The fixedend forces and moments must be derived and applied to the
system.
S4 = x 3
dx
EI y
100
200
x 3 dx
x 3 dx
1 1004 04 2004 1004
+
+
=
= 212.5
4 ( 200 )
100 E 100 200 E 10000 4 (100 )
0
0
w
Pg = [ P ] [ FEPM ] = FEPiz =
( S1S 4 S 2 S3 )
2D
2
FE
EM iy
( S3 S 2 S 4 )
0
0.05
Pg =
0.00015 ( 212.5 ) 0.0125 (1.5 ) = 4.7727
2 ( 0.00006875 )
2
147.73
The general stiffness equation can be setup and solved as follows. The
units are in inches and radians:
[ Pi ] = [ Kii ][d i ]
1
[d i ] = [ Kii ] [ Pi ]
0
0
666.67
2.1818 181.81
[d i ] = 0
0
181.81 21818
0 0
4.7727 = 5.3123
147.73 0.0375
Mjy
X
iy
Pjz
Piz
L
Miy
M
Piz
The internal shear, Vx, and moment, Mx, at any point, x, can be found
from statics and the partial derivatives of that shear and moment can be
found with respect to the applied force and moment at the iend.
Vx = Piz
dVx
= 1
d Piz
dVx
=0
d M iy
M x = Piz x M iy
dM x
= x
d Piz
dM x
= 1
d M iy
Castiglianos second theorem can be applied noting that at the iend the
deflection is zero and the rotation is qiy.
iz = 0 = M x
0=
d M x dx
d Vx dx
dx
dx
+ Piz
+ Vx
= ( Piz x 2 + M iy x )
d Piz EI y
d Piz GAz
EI y
GAz
2
Piz L3 M iy L Piz L
+
+
3EI y 2 EI y GAz
M iy
EI y
2 Piz L 2 Piz
3EI y LGAz
q iy = M x
q iy =
d M x dx
d Vx dx
dx
dx
+ 0
+ Vx
= ( Piz x + M iy )
d M iy EI y
d M iy GAz
GAz
EI y
Piz L2 M iy L
+
EI y
2 EI y
Substituting the first equation into the second equation results in the following:
q iy =
= iz iz
2 EI y 3EI y GAz
6 EI y GAz
L2
2
q iy =
Piz
6 EI y GAz
6 EI y Az G
Piz = 2
q iy
L Az G + 12 EI y
(5.13)
Substituting Equation 5.13 into the following equation repeated from earlier, results in Equation 5.14 for the second stiffness value.
M iy
EI y
2 Piz L
P
iz
3EI y LGAz
4 EI y L2 Az G + 3EI y
M iy =
L3 Az G + 12 LEI y
) q
iy
(5.14)
iz
Piz
Pjz
Miy
M
Piz
The internal shear, Vx, and moment, Mx, are exactly the same as in
Example5.8. The partial derivatives are also the same.
Vx = Piz
dVx
= 1
d Piz
dVx
=0
d M iy
M x = Piz x M iy
dM x
= x
d Piz
dM x
= 1
d M iy
Vx = Piz
dVx
= 1
d Piz
dVx
=0
d M iy
M x = Piz x M iy
dM x
= x
d Piz
dM x
= 1
d M iy
Castiglianos second theorem can be applied noting that at the iend the
rotation is zero and the deflection is Diz.
qiy = 0 = M x
0=
dVx dx
d M x dx
dx
dx
+ Vx
= Piz x + M iy
+ 0
d M iy EI y
d M iy GAz
EI y
GAz
Piz L2 M iy L
+
EI y
2 EI y
Piz L
2
d M x dx
dVx dx
dx
dx
iz = M x
= Piz x 2 + M iy x
+ Piz
+ V
EI y GAz
d Piz EI y x d Piz GAz
M iy =
iz =
2
Piz L3 M iy L Piz L
+
+
3EI y 2 EI y GAz
Substituting the first equation into the second equation results in the
following:
iz =
+
= iz
+ iz
3EI y 4 EI y GAz 12 EI y GAz
L3
L
+
iz =
Piz
12E
EI y GAz
12 EI y Az G
Piz = 3
L A G + 12 LEI iz
z
y
(5.15)
Substituting Equation 5.15 into the following equation repeated from earlier, results in Equation 5.16 for the second stiffness value.
Piz L
2
6 EI y Az G
M iy = 2
iz
L Az G + 12 EI y
M iy =
(5.16)
The forces at the jend of the member due to the motions at the iend can
be found using the transmission matrix. Then, the forces at the iend due
to the motions at the jend can be found by symmetry of the stiffness
matrix. Finally, the forces at the jend due to motions at the jend can be
found using the transmission matrix. This process was illustrated previously in Chapter 4. The resulting terms are shown in the matrices given as
Equations 5.21 through 5.24.
iz
Piy
Mjz
X
Pjy
Miz
M
Piy
The internal shear, Vx, and moment, Mx, at any point, x, can be found
from statics and the partial derivatives of that shear and moment can be
found with respect to the applied force and moment at the iend.
Vx = Piy
dVx
= 1
d Piy
dVx
=0
d M iz
M x = Piy x M iz
dM x
=x
d Piy
dM x
= 1
d M iz
Castiglianos second theorem can be applied noting that at the iend the
deflection is zero and the rotation is qiz.
iy = 0 = M x
0=
Piy L3
3EI z
d M x dx
dV dx
dx
dx
+ Vx x
= Piy x 2 M iz x
+ Piy
d Piy EI z
d Piy GAy
EI z
GAy
M iz L2 Piy L
+
2 EI z GAy
M iz 2 Piy L 2 Piy
=
EI z 3EI z LGAy
iz = M x
iz =
d M x dx
dVx dx
dx
dx
+ Vx
= Piy x + M iz
+ 0
d M iz EI z
d M iz GAy
EI z
GAy
Piy L2
2 EI z
M iz L
EI z
Substituting the first equation into the second equation results in the
following:
q iz =
Piy L2
2 EI z
2 Piy L2
3EI z
2 Piy
GAy
Piy L2
6 EI z
2 Piy
GAy
L2
2
Piy
+
q iz =
EI z GAy
6E
6 EI z Ay G
Piy = 2
q iz
L Ay G + 12 EI z
(5.17)
Substituting Equation 5.17 into the following equation repeated from earlier, results in Equation 5.18 for the second stiffness value.
M iz 2 Piy L 2 Piy
=
EI z 3EI z LGAy
4 EI z L2 Ay G + 3EI z
M iz =
L3 Ay G + 12 LEI z
) q
(5.18)
iz
Miz
iy
Piy
Pjy
Miz
M
Piy
The internal shear, Vx, and moment, Mx, are exactly the same as in
Example5.10. The partial derivatives are also the same.
Vx = Piy
dVx
= 1
d Piy
dVx
=0
d M iz
M x = Piy x M iz
dM x
=x
d Piy
dM x
= 1
d M iz
Vx = Piy
dVx
= 1
d Piy
dVx
=0
dM
246NUMERICAL STRUCTURAL izANALYSIS
M x = Piy x M iz
dM x
=x
d Piy
dM x
= 1
d M iz
Castiglianos second theorem can be applied noting that at the iend the
rotation is zero and the deflection is Diy.
qiz = 0 = M x
0=
M iz =
Piy L2
2 EI z
M iz L
EI z
Piy L
2
iy = M x
iy =
d M x dx
dVx dx
dx
dx
+ Vx
= Piy x + M iz
+ 0
d M iz EI z
d M iz GAy
EI z
GAy
Piy L3
3EI z
d M x dx
dV dx
dx
dx
= Piy x 2 M iz x
+ Piy
+ Vx x
d Piy EI z
d Piy GAy
EI z
GAy
M iz L2 Piy L
+
2 EI z GAy
Substituting the first equation into the second equation results in the
following:
iy =
Piy L3
3EI z
Piy L3
4 EI z
Piy L
GAy
Piy L3
12 EI z
Piy L
GAy
L3
L
Piy
+
iy =
EI z GAy
12E
12 EI z Ay G
Piy = 3
iy
L Ay G + 12 LEI z
(5.19)
Substituting Equation 5.19 into the following equation repeated from earlier, results in Equation 5.20 for the second stiffness value.
M iz =
Piy L
2
6 EI z Ay G
M iz = 2
iy
L Ay G + 12 EI z
(5.20)
The resulting terms are shown in the matrices given as Equations 5.21
through 5.24.
K ij
K jj
[ K ] = K ii
Ax E
0
0
L
12 EI z Ay G
0
0 L3 A G + 12 LEI
y
z
12 EI y Az G
0
0
3
L
A
z G + 12 LEI y
[ Kii ] =
0
0
0
6 EI y Az G
0
0
2
L Az G + 12 EI y
6 EI z Ay G
0
0
L2 Ay G + 12 EI z
ji
0
I xG
L
0
0
6 EI y Az G
L2 Az G + 12 EI y
0
4 EI y L2 Az G + 3EI y
L3 Az G + 12 LEI y
0
6 EI z Ay G
L2 Ay G + 12 EI z
2
4 EI z L Ay G + 3EI z
L3 Ay G + 12 LEI z
0
(5.21)
Ax E
0
0
0
0
0
L
12 EI z Ay G
6 EI z Ay G
0
0
0
2
0 L3 A G + 12 LEI
2
EI
+
1
L
A
G
z
y
z
y
12 EI y Az G
6 EI y Az G
0
3
2
0
0
0
L Az G + 12 LEI y
L Az G + 12 EI y
K ji =
I
G
0
x
0
0
0
0
2
2 EI y ( L Az G + 6 EI y )
6 EI y Az G
0
0
0
0
L2 Az G + 12 EI y
L3 Az G + 12 LEI y
2
+
EI
L
A
G
EI
2
6
EI
6
A
G
z (
y
z )
z y
0
0
0
0
L2 Ay G + 12 EI z
L3 Ay G + 12 LEI z
(5.22)
K ij =
Ax E
L
0
12 EI z Ay G
L3 Ay G + 12 LEI z
0
0
12 EI y Az G
I xG
L
6 EI y Az G
L2 Az G + 12 EI y
6 EI z Ay G
L2 Ay G + 12 EI z
L3 Az G + 12 LEI y
6 EI y Az G
L2 Az G + 12 EI y
0
2 EI y L Az G 6 EI y
L3 Az G + 12 LEI y
6 EI z Ay G
2
L Ay G + 12 EI z
2
2 EI z L Ay G 6 EI z
L3 Ay G + 12 LEI z
0
(5.23)
Ax E
0
0
L
12 EI z Ay G
0
0 L3 A G + 12 LEI
y
z
12 EI y Az G
0
0
L3 Az G + 12 LEI y
K jj =
0
0
0
6 EI y Az G
0
0
L2 Az G + 12 EI y
6 EI z Ay G
0
0
L2 Ay G + 12 EI z
0
I xG
L
0
6 EI y Az G
L2 Az G + 12 EI y
0
4 EI y L2 Az G + 3EI y
L3 Az G + 12 LEI y
6 EI z Ay G
2
L Ay G + 12 EI z
4 EI z L2 Ay G + 3EI z
L3 Ay G + 12 LEI z
0
(5.24)
5.7.1 SHEAR AREA
The shear area is the crosssectional property that is used for shear energy
resistance. It can be found for a crosssection using the shear stress equation for a beam derived in most strength of materials textbooks. The basic
equations for shear in the Y and Z direction are given as follows:
t xy =
Vy Qy
I z tz
Vy Ay '
I z tz
V Q V Az '
t xz = z z = z
I yt y
I yt y
If these equations are written using a single term to represent all the
crosssectional properties, it results in the following:
t xy =
t xz =
Vy Qy
I z tz
Vy Ay '
I z tz
Vy
Ay
Vz Qz Vz Az ' Vz
=
=
I yt y
I yt y
Az
The equations for the shear areas Ay and Az can then be found.
Ay =
I z tz
A y '
Az =
I yt y
Az '
There are three terms in these equations. The first is the centroidal moment
of inertia. The second is the moment of the area, Ay or Az, between the
centroid and the extreme fiber taken about the centroid. The third is the
value t, which is the width at the centroid.
Example5.12 Shear area
Determine the shear areas Ay and Az for the a rectangular section.
Figure5.14 shows the rectangle in the orientation to calculate Az that
corresponds to bending about the Y axis. To find Ay the area to the left of
the Z axis will be used.
y h
hb3
h
It
2bh
Ay = z z = 12 =
bh b
A y '
3
2 4
bh3
b
I yt y
2bh
Az =
= 12 =
Az ' bh h
3
2 4
For a rectangle, the shear areas are both twothirds the crosssectional
area.
Example5.13 Shear area
Determine the shear area Az for the Tshaped section shown in Figure5.15.
The centroid and the centroidal moment of inertia are found using the
moment of area principles.
z=
Az
A
2 (12 ) 6 + 12 ( 2 )13
2 (12 ) 2
= 9.50 in
2 (12 )
12 ( 2 )
2
2
+ 2 (12 ) ( 9.5 6 ) +
+ 12 ( 2 ) ( 9.5 13)
12
12
= 884.0 in 4
3
I y = I y + Ad z2 =
12"
z
2"
y
12"
2"
The area of the web below the centroid will be used to find Az.
Az =
I yty
Az '
884.0 ( 2 )
= 19.59 in 2
9.5
2 ( 9.5 )
2
b
12
2
h
z
2
13.0
12
6
z = 9.5
Iy = 884.00
A
Az
24
312
24
144
48
456
A'z' = 90.25
I
8
288
296
d
3.5
3.5
Ad2
294
294
588
Az = 19.590
Miz
Pix
Pix
iy
Miz
Piy
Piy
L
moment in the beam is a function of only the end shears and moments, as
given by the following equation:
M x = Piy x M iz
The second beam in Figure5.16 shows the bending deformations. In this
case, the internal moment is a function of not only the end shears and
moments, but also a function of the axial force multiplied by the beams
lateral deflection, y.
M x = Piy x M iz + Pix y
This additional moment, the product of axial force, Pix, and lateral deflection, y, is usually called the Pdelta effect. To derive a stiffness matrix
that includes the Pdelta effect, equilibrium of the deformed beam must
be considered.
Example5.14 Geometric stiffness
Derive the Diy stiffness using Castiglianos theorems for a linear member
including the geometric effects.
Using the principle of superposition, consider a beam with an applied
deflection while the rotation is held to zero. Figure 5.17 shows the
deformed beam with applied end forces. Also shown is a lefthand freebody of the beam cut at any distance x from the iend.
The internal bending moment in the beam is found from equilibrium.
M x = Piy x M iz + Pix iy y
Miz
Pix
iy
Piy
L
Miz
Pix
iyy
M
P
Piy
V
x
y ' ( 0) = 0
y ' ( L) = 0
2 x3 3x 2
y = iy 3 2 + 1
L
L
Substituting this equation into the internal moment equation yields the
following:
2 x3 3x 2
M x = Piy x M iz + Pix iy 3 + 2
L
L
It is assumed that axial shortening is caused only by the axial force, Pix.
This is the same assumption used for the ordinary elastic stiffness derivation. The geometric stiffness derivation considers the lateral and rotational
deformations, Diy and qiz. From Castiglianos theorem, the general deflection and rotation of the free end are as follows:
iy = M x
d M x dx
d Piy EI z
qiz = M x
d M x dx
d M iz EI z
dM x
=x
d Piy
dM x
= 1
d M iz
qiz = 0 =
M iz = Piy
L
1
+ Pix iy
2
2
2 x 3 3 x 2 dx
Piy x + M iz + Pix iy 3 2
L EI z
L
L
L
L2
+ M iz
Pix iy
0 = Piy
2 EI z
2 EI z
EI z
2 x 4 3 x 3 dx
2
Piy x M iz x + Pix iy 3 + 2
L EI z
L
7 L2
L3
L2
+ Pix iy
iy = Piy
M iz
10 EI z
3EI z
2 EI z
iy =
Substituting the first equation into the second equations yields the
following:
iy = Piy
L3
L2
+ Pix iy
12 EI z
10 EI z
Piy = iy
12 EI z
6
+ Pix iy
3
5L
L
(5.25)
Substituting Equation 5.25 into the following equation repeated from earlier results in Equation 5.25 for the second stiffness value.
1
L
+ Pix iy
2
2
6 EI
1
(5.26)
M iz = iy 2 z + Pix iy
10
L
M iz = Piy
Take note that the first terms in each of these stiffness equations are the
same as the elastic stiffness values derived in Equations 4.26 and 4.27.
The second term is the geometric component due to the deflected shape
and the axial thrust.
iz
L
Miz
Pix
M
Piy
iz
y = ax 3 + bx 2 + cx + d
y ( 0) = 0
y ( L) = 0
y ' ( 0) = qiz
y ' ( L) = 0
x3 2 x 2
y = qiz 2
+ x
L
L
x3 2 x 2
x
M x = Piy x M iz + Pix qiz 2 +
L
L
The general deflection and rotation of the free end are the same as
Example5.14.
iy = M x
d M x dx
d Piy EI z
qiz = M x
d M x dx
d M iz EI z
d Mx
= 1
d M iz
Setting Diy to zero and solving for Piy and Miz in terms of the deflection will
result in two terms of the stiffness matrix.
x3 2 x 2
dx
+ x
Piy x + M iz + Pix qiz 2
L
L
EI z
L
L2
L2
+ M iz
+ Pix qiz
qiz = Piy
2 EI z
EI z
12 EI z
qiz =
EI
L
L
Pix qiz qiz z
2
12
L
x 4 2 x3
dx
iy = 0 = Piy x 2 M iz x + Pix qiz 2 +
x2
L
L
EI z
M iz = Piy
0 = Piy
L2
L3
L3
M iz
Pix qiz
3EI z
2 EI z
30 EI z
Substituting the first equation into the second equation yields the following:
Piy = q iz
6 EI z
1
+ Pixq iz
2
10
L
(5.27)
Substituting Equation 5.27 into the following equation repeated from earlier results in Equation 5.28 for the second stiffness value.
EI
L
L
Pix qiz qiz z
2
12
L
4 EI z
2 L
M iz = qiz
+ Pix qiz
L
15
M iz = Piy
(5.28)
Take note that the first terms in each of these stiffness equations are the
same as the elastic stiffness values derived in Equations 4.24 and 4.25.
The second term is the geometric component due to the deflected shape
and the axial thrust. All four terms can be written in matrix form.
12 EI z
L3
6 EI z
2
L
6 EI z
6
2
5L
L
+ Pix
4 EI z
1
10
L
1
10 iy Piy
=
2 L qiz M iz
15
The first matrix on the left side of the equation in the basic elastic stiffness
matrix will be called [K]. The second matrix on the left side of the equation in the geometric stiffness matrix will be called [G]. In general terms,
the equation may be written as follows:
([ K ] + P [G ]) [ ] = [ F ]
ix
The same transformation used for previous stiffness matrix derivations can
be applied to find the rest of the geometric stiffness matrix. The geometric
stiffness matrix for the coplanar XY system is given as Equation 5.29.
The sign convention on Pix is positive for tension.
0
0
6
0
5L
1
0
10
[ K m ] = Pix 0 0
0 6
5L
1
0
10
0
1
10
2L
15
0
1
10
L
30
0
6
5L
1
10
0
6
5L
1
10
0
0
0
0
0
0
1
10
L
30
0
1
10
2L
15
(5.29)
6
0
5L
1
0
10
[ K m ] = Pix 0 0
0 6
L
5L
1
0
10
1
10
2L
15
0
1
10
L
30
0
6
5L
1
10
0
6
5L
1
10
0
0
0
0
0
0
1
10
L
30
0
1
10
2L
15
(5.30)
Z
50kips
1kip
A=10 in2
I=100 in4
200in
E=10,000 ksi
only the iend of the member stiffness needs to be developed. The value of
Pix in this case is known to be 50 kips.
EAx
L
0
12 EI z
L3
6 EI
2z
L
0 0
6 EI
6
2 z + Pix 0
5L
L
4 EI z
0 1
10
L
0
ix Pix
1
iz = Piz
10
q M
2 L iy iy
15
500
0
0 0 0
0 ix 50
0
1.5
150 + 0 .3
5 iz = 1
0
0
ix 50
500
0
1.2
145 iz = 1
([ K ] + P [G ]) [ ] = [ F ]
[ ] = ([ K ] + P [G ]) [ F ]
ix
ix
ix 0.1000
iz = 13.5757
qiy 0.10545
If the solution was performed with the geometric stiffness omitted, the
result would be as follows:
ix 0.1000
iz = 2.6667
qiy 0.02000
A=10 in2
I=100 in4
200in
E=10,000 ksi
This is the same as Example5.16, except the axial force has been
increased to 100 kips.
500
0
0 0 0
0 ix 100
0
1.5
150 + 0 .6
10 iz = 1
0
0
ix 100
500
0
0.9
140 iz = 1
iz = 4.33333
qiy 0.00350
Observe that the iend of the beam moved in the negative Z direction. This
does not make logical sense. The reason of the backward motion is that the
member has buckled elastically. The actual elastic buckling load of this
column is 61.685 kips.
0
Pix
0
0
6
5L
0
0
0
1
10
0
6
5L
0
0
0
1
10
6
1
0
5L
10
0 0 0
1
2L
0
10
15
0
6
1
0
5L
10
0 0 0
L
1
0
10
30
0 0 0
1
6
0
10
5L
1
2L
0
15
10
0 0 0
1
6
0
10
5L
0
0
L
1
0
30
10
6
1
0
5L
10
0 0 0
L
1
0
10
30
0
6
1
0
5L
10
0 0 0
1
2L
0
10
15
0
0
1
10
0
0
L
30
0
1
10
0
2L
15
ix Pix
iy Piy
iz Piz
q ix M ix
q M
iy iy
q iz M iz
jx = Pjx
P
jy jy
jz Pjz
q M
jx jx
q jy M jy
q M
jz jz
(5.31)
The geometric stiffness of a member can also be derived based on a general transcendental equation. The full derivation is published by Blette
(1985).
np
np
y = asin + bcos + cx + d
2L
2L
px
px px p
2
y = iy
+
sin
+ cos
4 p
2L
2L 2L 2
The particular solution for qiz is as follows:
px
px px
4 L p
1 sin
y = qiz
+ cos
+
1
2
4p p 2
2L
2 L 2 L
These two relationships can be used to develop the geometric stiffness
matrix.
0
Pix
0
0
0
0
1.2036
L
0
0
0
0.1018
0
1.2036
0.1018
1.2036
L
0
0.1018
0
0
0
1.2036
0
L
0
0
0
0.1018
0.1018
0
0
1.2036
0
L
0 0.1018
0
0
0 0.1379 L
0
0
0
0
0
0
0.1379 L
0
0
0
0
0
0.1018 0
0.1018
0
0
0
0
0 0.0361L
0
0
0
0
0.0361L 0
1.2036
0
L
0
0
0
0.1018
0
0.1018
0
0
1.2036
0
L
1.2036
0
L
0
0
0
0.1018
0
0.1018
0
0
0
0
0
0
0
0
0
0
0
0.1018 ix Pix
iy Piy
0.1018
0
iz Piz
0
0
M
ix ix
0.0361L
0
M
iy
iy
0
0.0361L iz M iz
jx = Pjx
0
0
P
0
0.1018 jy jy
jz Pjz
jx
M jx
0.1018
0
M
jy
jy
0
0
M
jz
jz
0.1379 L
0
0
0.1379 L
(5.32)
If more digits are desired for accuracy, the following substitutions can be made:
1.2036=1.20362445
0.1018=0.1018122226
0.1379=0.1378809597
0.0361=0.03606873710
ay =
L3
12 EI y
az =
L3
12 EI z
by =
L
GAz
bz =
L
GAy
Ax E
L
Ax E
L
L
2 (a z + b z )
L
2 (a y + b y )
1
a y + by
1
a z + bz
IxG
L
L
2 (a y + b y )
I xG
L
L
2 (a z + b z )
1
a y + by
1
a z + bz
L
2 (a y + b y )
L (a y + b y )
EI y ( 2 a y b y )
L
2 (a y + b y )
L (a y + b y )
EI y ( 4 a y + b y )
L
2 (a z + b z )
L (a z + b z )
EI z ( 2 az b z )
L (a z + b z )
EI z ( 4 az + b z )
L
2 (a z + b z )
Ax E
L
Ax E
L
1
a z + bz
L
2 (a z + b z )
1
a z + bz
2 (a z + b z )
1
a y + by
L
2 (a y + b y )
I xG
L
IxG
L
1
a y + by
L
2 (a y + b y )
L (a y + b y )
EI y ( 4 a y + b y )
L
2 (a y + b y )
L (a y + b y )
L
2 (a y + b y )
EI y ( 2 a y b y )
2 (a z + b z )
P
0
ix
Piy
Piz
0
M ix
EI z ( 2 a z b z ) M iy
M
L (a z + b z ) = iz
Pjx
P
0
jy
Pjz
L
M
2 (a z + b z ) jx
M
jy
0
M jz
EI z ( 4 a z + b z )
L (a z + b z )
(5.33)
0
0
0
0
( 6a z + 5b z )
5L ( a z + b z )
0
0
0
0
0
0
L
0
2 (a z + b z )
0
0
0
( 6a z + 5b z )
5L ( a z + b z )
0
0
0
0
0
0
az
0
a
10
( z + bz )
0
0
( 6a y + 5b y )
5L (a y + b y )
0
ay
10 a y + b y
0
0
0
ay
( 6a y + 5b y )
5L ( a y + b y )
10 a y + b y
0
0
0
0
)
60 (a y + b y )
)
60 (a y + b y )
L 2a y + 5b y
10 a y + b y
ay
L 8a y + 5b y
10 a y + b y
ay
0
0
60 (a z + b z )
0
0
0
az
10 (a z + b z )
60 (a z + b z )
L ( 8a z + 5b z )
L ( 2a z + 5b z )
0
0
0
az
10 (a z + b z )
0
0
0
0
0
0
0
0
10 a y + b y
0
ay
( 6a y + 5b y )
5L ( a y + b y )
( 6a z + 5b z )
5L (a z + b z )
0
0
10 a y + b y
az
10 (a z + b z )
0
ay
( 6a y + 5b y )
5L (a y + b y )
0
0
az
10 (a z + b z )
0
0
5L (a z + b z )
( 6a z + 5b z )
0
0
0
0
0
0
)
60 (a y + b y )
10 a y + b y
ay
L 8a y + 5b y
)
60 (a y + b y )
L 2a y + 5b y
10 a y + b y
ay
0
0
(5.34)
L ( 2a z + 5b z )
60 (a z + b z )
az
10 (a z + b z )
L ( 8a z + 5b z )
60 (a z + b z )
0
az
10 (a z + b z )
5.12TORSION
The torsional stiffness of slender linear members is composed of two parts.
The first is known as St. Venants torsion, which is uniform on a member
at any distance, r, from the longitudinal axis. This is the torsional stiffness
that is included in the elastic member stiffness Equation 4.33. This is the
primary resistance to torsion for circular crossessections that have area
distributed uniformly about the longitudinal axis. The second type of torsional stiffness is known as warping torsion. Warping torsion is the primary
stiffness in thinwalled open crosssections such as angles shapes and wide
flange shapes. The warping torsion will cause longitudinal deformations in
the crosssection that will cause certain portions to elongate and other portions to shorten. This warping effect can be included in the derivation of the
stiffness. Structural Analysis and Design, by Ketter, Lee, and Prawel, Jr.,
covers this derivation (Ketter, Lee, and Prawel 1979). The torsional stiffness
at each degree of freedom is represented as two components instead of the
single values used in the normal elastic stiffness. Equation 5.35 shows the
elastic torsional member stiffness in the 3D Cartesian coordinate system.
Ax E
0
0
L
12 EI z
0
0
L3
12
EI
y
0
0
L3
0
0
0
0
0
0
6 EI y
2
0
0
L
6 EI z
0
0
L2
A
E
x
0
0
L
12 EI z
0
0 L3
12 EI y
3
0
0
L
0
0
0
0
0
0
6 EI y
2
0
0
L
6 EI z
0
0
L2
6 EI z
L2
T1 T2
T2 T3
6 EI y
L2
0
0
4 EI y
0
0
0
0
4 EI z
L
T1 T2
T2 T4
0
0
0
0
L
0
12 EI z
L3
0
0
L2
0
0
2 EI y
6 EI y
Ax E
L
12 EI y
L3
0
0
6 EI y
0
0
T1 T2
T2 T4
6 EI y
L2
0
0
2 EI y
12 EI z
L3
0
0
0
0
0
0
6 EI
2z
L
6 EI z
L2
2 EI y
L
Ax E
L
6 EI z
L2
L2
12 EI z
L3
0
0
6 EI y
L2
0
T1 T2
T2 T3
0
6 EI y
L2
0
0
4 EI y
L
0
6 EI z
2
L
0
0
2 EI z
6 EI
2z
L
0
0
4EI z
L
0
(5.35)
The resulting member stiffness matrix is 1414 in size with the terms T1,
T2, T3, and T4 defined as follows:
T1 = GT
T2 = GT
l sinh ( l L )
2 cosh ( l L ) 1 l Lsinh ( l L )
cosh ( l L ) 1
2 cosh ( l L ) 1 l Lsinh ( l L )
sinh ( l L ) l Lcosh ( l L )
GT
l 2 cosh ( l L ) 1 l Lsinh ( l L )
l L l sinh ( l L )
GT
T4 =
l 2 cosh ( l L ) 1 l Lsinh ( l L )
T3 =
l=
GT
EIw
In the equation for l, kT is the St. Venant torsion constant, which is typically the polar moment of inertia, Ix. The value of Iw is known as the warping constant. Both of these values are normally tabulated in handbooks or
specifications.
5.13SUBSTRUCTURING
When a structure is of large enough size that the contents for the global
joint stiffness matrix cannot be contained in the RAM of a computer, the
matrix can be transformed into segments by reduction or decomposition.
The resulting transformed matrix can take many forms depending on the
process used. One of the common transformations is the Nmatrix, due to
the configuration of resulting values. The following is a general description of the operation used to solve large systems using the Nmatrix. This
method can be used by operating on the individual degrees of freedom or
on the entire joint as matrix operations.
The original equation set is normally a sparse matrix with most of
the values near the main diagonal. Equation 4.34 represents the original
stiffness solution set. For clarity, the zero values are left out of the matrices
and X indicates where values exist.
K g g = Pg
X
X
X
X
X
X
X
X
X
X
X
X
X
=
FEPM
X
X
X
(5.36)
Kg
X
X
X
K g
X
X
X
X
K g g Pg
X d S
X d A
=
X d V
X d E
K g g Pg
(5.37)
The information contained in the Nmatrix is used in two ways. The values
in the extreme boundaries of the matrices represent the stiffness, deformation, or force (Kg, Dg, and Pg) of the structure related to the retained degrees
of freedom or joints. These values can be combined with other members
in a stiffness analysis. The interior portions of the matrices represent the
values of the stiffness, deformation, or force of the degrees of freedom or
joints eliminated in the reduction. These values are used to find the deformations and forces of these degrees of freedom or joints once the main analysis
results are known. The Nmatrix can be compressed as shown in Equation
5.38. The interior stiffness values can be inverted for later processing.
Kg
X
X
X
K g
X
X 1
X 1
X 1
K g g Pg
X S
X A
=
X V
X E
K g g Pg
(5.38)
The main analysis is performed and the global deformation of the retained
degrees of freedom of joint is found.
g = K g
Pg
g
X
X
X
X
g
(5.39)
(5.40)
The local member forces are found from the deformation in Equation
5.40, the same as in step 6 of the general stiffness procedure given in
Section 4.13. The values of D are used for Dg in Equation 4.38 and
repeated here.
P & M = K [R ] + FEPM
m m
m
REFERENCES
Blette, K.J. 1985. Numerical Methods for Architectural Engineers. Stillwater,
Oklahoma: Oklahoma State University.
Ketter, R.L., G.C. Lee, and S.P. Prawel. 1979. Structural Analysis and Design.
New York, New York: McGrawHill Book Company.
Timoshenko, S.P. 1921. On the Correction Factor for Shear of the Differential
Equation for Transverse Vibrations of Bars of Uniform Crosssection. Philosophical Magazine 41, pp. 7446.
Index
A
Adjoint matrix, 60, 6263
Algebraic equation, 1
Alpha rotation matrix, 148149
Area moment method, 155158
B
Bairstows method, 3846
Basket weave method, 5556
Beta rotation matrix, 149150
Bisection method, 1315
C
Carryover factor, 171
Castiglianos second theorem,
165168
Cholesky decomposition method,
7378
Cofactor matrix, 57, 59, 61
Column buckling, difference
operators, 134136
Column matrix, 49
Conjugate beam method, 158161
Coordinate systems, 147148
Cramers rule
definition, 56
eigenvalues by, 8790
example, 57
Crouts method, 7378
D
Descartes rule, 36
Diagonal matrix, 4950
Difference operators
column buckling with,
134136
fixed beam with, 128133
partial differential equations,
136140
simple beam with, 123128
Taylor series expansion,
118123
Distribution factor, 171
Double integration
Gaussian quadrature, 114115
Simpsons onethird rule,
112114
E
Eigenvalues, Cramers rule, 8790
Elastic member stiffness
threedimensional (3D)
system, 186187
XY system, 181186
XZ system, 174181
Equations
algebraic, 1
homogeneous, 47
linear algebraic, 47
nonhomogeneous, 47
simultaneous, 4748
transcendental, 12
Error equations, 7880
F
FaddeevLeverrier method, 9091
274Index
J
Joint stiffness factor, 171
L
Linear algebraic equations, 47
Linear interpolation method,
1517
Lower triangular matrix, 50
M
Matrix
adjoint, 60, 6263
cofactor, 57, 59, 61
column, 49
definition, 48
diagonal, 4950
identity, 50
inverted, 51
minor, 51
row, 49
square, 49
transposed, 51
triangular, 50
Matrix factorization, 7378
Matrix inversion method, 8086
Matrix operations
matrix addition and
subtraction, 5253
matrix determinants, 5456
matrix multiplication, 5354
scalar multiplication, 53
Matrix structural analysis
area moment method, 155158
Castiglianos second theorem,
165168
conjugate beam method,
158161
global joint stiffness, 187203
momentdistribution method,
170174
slopedeflection method,
168170
transformation matrix
rotation matrix, 148152
Index 275
transmission matrix,
153155
virtual work method, 161163
visual integration method,
163165
Member end releases
three dimensional (3D)
system, 217229
XY system, 214217
XZ system, 205214
Member relative stiffness factor,
171
Member stiffness factor, 171
Method of false position, 1517
Momentdistribution method,
170174
N
NewtonRaphson method, 1821
Newtons second order method,
2124
Newtons tangent method, 1821
Nonhomogeneous equation, 47
Nonprismatic member stiffness,
229234
Numerical integration and
differentiation
difference operators
column buckling with,
134136
fixed beam with, 128133
partial differential
equations, 136
140
simple beam with, 123128
Taylor series expansion,
118123
double integration
Gaussian quadrature,
114115
Simpsons onethird rule,
112114
Gaussian quadrature method,
109112
276Index
FaddeevLeverrier method,
9091
GaussJordan elimination
method, 6871
GaussSeidel iteration method,
8687
improved GaussJordan
elimination method,
7273
iterative method, 9194
matrix inversion method,
8086
matrix operations, 5256
power method, 9194
Slopedeflection method, 168170
Square matrix, 49
Stiffness
definition, 174
elastic member
threedimensional (3D)
system, 186187
XY system, 181186
XZ system, 174181
flexural, 171
geometric
flexural stiffness
derivations,
262264
three dimensional (3D)
system, 260262
XY system, 251257
XZ system, 258260
global joint, 187203
shear
flexural stiffness
derivations,
262264
shear area, 248251
three dimensional (3D)
system, 247248
XY system, 243247
XZ system, 239243
torsion, 265266
Stiffness factor, 171
Index 277
Substructuring, 266268
Synthetic division, 710
T
Taylor series polynomial
expansion, 116118
Threedimensional (3D) system
elastic member stiffness,
186187
geometric stiffness, 260262
member end releases, 217229
shear stiffness, 247248
Torsional stiffness, 265266
Transformation matrix
rotation matrix
alpha, 148149
beta, 149150
gamma, 150152
transmission matrix
cause and effect, 153155
orthogonal forces, 153
Transcendental equation, 12
Transmission matrix
cause and effect, 153155
orthogonal forces, 153
Transposed matrix, 51
Trapezoidal rule, 9799
Triangular matrix, 50
U
Upper triangular matrix, 50
V
Virtual work method, 161163
Visual integration method,
163165
X
XY system
elastic member stiffness,
181186
geometric stiffness, 251257
member end releases, 214217
shear stiffness, 243247
XZ system
elastic member stiffness,
174181
geometric stiffness, 258260
member end releases, 205214
shear stiffness, 239243
The Momentum Press digital library is very affordable, with no obligation to buy in future years.
For more information, please visit www.momentumpress.net/library or to set up a trial in the US,
please contact mpsales@globalepress.com.
Manufacturing
Engineering
Mechanical
& Chemical
Engineering
Materials Science
& Engineering
Civil &
Environmental
Engineering
Electrical
Engineering
THE TERMS
Perpetual access for
a one time fee
No subscriptions or
access fees
Unlimited
concurrent usage
Downloadable PDFs
Free MARC records
For further information,
a free trial, or to order,
contact:
sales@momentumpress.net
THE CONTENT
OHARA RAMMING
EBOOKS
FOR THE
ENGINEERING
LIBRARY
SUSTAINABLE STRUCTURAL
SYSTEMS COLLECTION
Mohammad Noori, Editor
Numerical
Structural
Analysis
Steven OHara
Carisa H. Ramming