Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

APPENDIX

VARIOUS
NUMERICAL PROCEDURES
SOLUTION OF
BEAM BENDING PROBLEM

INTRODUCTION

To illustrate the use of energy procedures and of the methods of weighted


residuals, in this appendix we an example of beam bending by
shall solve
using these procedures. Moreover, we and finite
also consider the Ritz
difference methods the former is based on the minimization concept and is
;

often considered a forerunner of the finite element method. The coverage of


different methods herein is intended only as an introduction to various
procedures and to give the reader an idea of the available schemes detailed ;

study of these procedures is beyond the scope of this elementary treatment.


Before the problem of beam bending is considered, we shall give further
details of the methods of weighted residuals (MWRs) introduced in Chapter
2. A good introduction and applications of MWR
are given by Crandall [1].
As stated in Eq. (2-12), the trial or approximation function in the MWR
is expressed as

u = ± «,?,. (2-12)

The undetermined parameters a, are chosen such that the residual R(x) over
the domain D vanishes. This is usually done in an average sense by weighting
R(x) (Fig. 2-7) with respect to weighting functions W t
(x). Thus,

f
Jd
R(x)W (x)dx
i
= 0, i = 1, 2, . .
.
, «. (Al-1)

398
Appendix 1 Solution of Beam Bending Problem 399

For a one-dimensional problem, the domain D is simply the linear extent of


the body.

VARIOUS RESIDUAL PROCEDURES

There are a number of ways to choose Wu and depending on the choice of


W ti we obtain different procedures.
In the case of the collocation method,

W = 3(x - x
t t). (Al-2a)
Then
RixdSix - x)
t
= 0, i = 1, 2, . .
.
, n, (Al-2b)
where
1, x =x if
3 = i

[0, x ^x t9

is the Dirac delta function. This means that the residual is equated to zero at a
selected number of points in the domain. For instance, as shown in Fig.
A 1-1 (a), the residual is equated to zero at n = 5 points.
Thedomain can be divided into a number of subdomains [Fig.
total
A 1 and the residual is integrated and equated to zero over each sub-
-1(b)],

domain. This yields the subdomain method. Here the weighting functions are

- ,+1 '
(Al-3a)
0, x 'J
< x and x t
> xt+l .

Then

p
1

R(x)dx = 0, i = 1, 2, ...,«- 1. (Al-3b)

In the case of the least-squares method [Fig. Al-l(c)], the weighting


functions are chosen to be

*-g. (Al-4)

and this leads to minimization of the integrated square residual as

f
R 2 (x)a dx = t
{
i 0. (Al-5)
Jd

In Galerkin's method [Fig. Al-l(d)], the weighting functions are chosen as


the coordinate functions from Eq. (2-12), and hence

f *(*)?,(*)= 0, / = 1,2, ...,«. (Al-6)


Jd

This expression implies that the functions q> t are made orthogonal to the

residual R{x). In finite element applications, the approximation or trial


400 Solution of Beam Bending Problem Appendix 1

Domain D ,

£
R(3)

12 ! 1
3
I
4 5 i
R
= 1,2
(x=) =
5

(a)

R(x)dx = 0, i
= 1,2,3, 4
-1 + 1 f:
(b)

2 = 1,2
(x)a, = 0, 5
JR
n

Ry?,dx = 0, i
= 1, 2, 3, 4

Figure Al-1 Methods of weighted residuals, (a) Collocation, (b)


Subdomain. (c) Least squares, (d) Galerkin.

functions are commonly expressed in terms of shape, interpolation, or basis


functions N t ;
then the N (
are usually chosen as the weighting functions.

BEAM BENDING BY VARIOUS PROCEDURES


The beam and details are shown in Fig. A 1-2. Assuming flexural rigidity to be
uniform, the governing differential equation is given by Eq. (7- lb), and the
residual is

R(x) = F^ - P(x), (Al-7)


Appendix 1 Solution of Beam Bending Problem 401

Collocation
5 Finite difference

Subdomain
L/2 3L/4

Galerkin,
(c) Least squares,
Ritz
0-L

(d)
A A Finite element

© © ©
Figure Al-2 Beam bending with different methods, (a) Collocation,
finite difference, (b) Subdomain. (c) Galerkin, least squares, Ritz.
(d) Finite element.

where F = £7 is the flexural rigidity, w is the assumed transverse displacement,


p(x) is the forcing function, and x is the coordinate. The boundary conditions
associated with Eq. (7- lb) can be expressed as

W (X = 0) = w(x =L)=0, (Al-8a)


2
d w, =
dx 2
0)7
dx Ax
:
L) 0. (Al-8b)

The first set [Eq. (Al-8a)] represents the essential or forced boundary condi-
tions and the second set [Eq. (Al-8b)] represents the natural boundary
conditions.
We now choose the following trial or approximation function for the
unknown w* :

w = a, sin ^+a 2 sin ^+ a3 sin ^+ a4 sin


^
= t *0M, (Al-9)

where the a are the undetermined parameters and the


t
(p t are the known
:

402 Solution of Beam Bending Problem Appendix 1

functions. Here we have chosen w* in terms of the trigonometric functions,


whereas in Chapter 7 the approximation function was chosen in terms of
interpolation functions N t.

Note that the function w in Eq. (A 1-9) satisfies the boundary conditions in
Eq. (A 1-8) at the two ends of the beam. In Chapter 7, when we used
Galerkin's method for each element, then only the geometric boundary
conditions [Eq. (Al-8a)] at the ends were used to modify the assemblage
equations; as explained in Chapter 3, the natural boundary conditions
[Eq. (Al-8b)] are satisfied automatically in an integrated sense.
To express R(x) in terms of w and its derivative, we differentiate the
expression in Eq. (A 1-9) four times as

-T-; = 1*cl x sin -j- + 16/ 4 a 2 sin -j- + 8U 4


a3 sin -j—

+ 256A 4 a 4 sin^, (Al-10)

where X = n/L.
Now we shall consider the solution of the beam bending problem by using
a number of different procedures. For this illustration, the following prop-
erties are assumed
E= 10 X 10 6 psi,

L = 10 in.,
A = in. x in. = in.
1 1 1
2
,

PA = 500 and PB = 1000 lb/in.

Collocation

As shown in Fig. A 1-2, we chose four points at x = L/5, {


2L/5, 3L/5, and
4L/5, i = 1, 2, 3, 4, from the end A. Note that the residual is identically zero
at the supports. Then we have
R(x = x )=0,t 1 = 1,2,3,4. (Al-lla)

For instance, for Xj = L/5, we have


X4a l
sin j- ~ + 16A a 4
2 sin
^ y + 8U a
4
3 sin ^~
Pb ~ Pa
+ 256A 4 a 4 sin
^^ - ±--Pa = (A-l lb)

and so on. The resulting equations are

4.77 x 103<xi + 1.23 x 10^2 + 6.25 x 10 4 a 3 + 1.22 x 10«a 4 = 600,

7.72 x 103<xi + 7.63 x 10 4 a 2 - 3.86 x 105a 3 - 1.98 x 10^4 = 700,


(Al-12)
7.72 x 10 3 ai - 7.63 x 10 4 a 2 - 3.86 x 10 5 a 3 + 1.98 x 10 6 a 4 = 800,

4.77 x 103<xi - 1.23 x 10*a 2 + 6.25 x 105<x 3 - 1.22 x 10 6 a 4 = 900.


:

Appendix 1 Solution of Beam Bending Problem 403

Solution of these equations gives values of a, as

a, = 0.11374399, a2 = -0.00105974,
(Al-13)
a3 = 0.00033150, a3 = -0.00001564.
Hence, the approximate solution according to the collocation method is

w 0.11374 sin ^ - 0.00106 sin ^+ 0.0003315 sin ^*


4nx
-0.00001564 sin (Al-14)

Subdomain Method

Here, we have [Fig. A 1


-2(b)]

rL/4
R(x)dx o,
Jo

r 2L/4
R(x)dx o,
J L/4
(Al-15)
r 3 LI 4
R(x)dx 0,
J 1L 4

L
[ R(x)dx 0.
J 3Z. 4

After integrations, the resulting four equations are

7.57 x lCPai + 2.07 x 105<x 2 + 1-19 x 10*a 3 + 3.31 x 10^4 = 1410,

1.83 x 10 4 ai + 2.07 x 105a 2 - 4.93 x 105a 3 - 3.31 x 10*a 4 = 1720,


(Al-16)
1.83 x 10 4 ai - 2.07 x 10 5 a 2 - 4.93 x 10 5 a 3 + 3.31 x 10 6 a 4 = 2030,

7.57 x 10 3 ai - 2.07 x 10 5 a 2 + 1.19 x 10 6 a 3 - 3.31 x 10 6 a 4 = 2340.

Solution of these equations leads to the following approximation

w = 0.12388 sin ^- 0.00151 18 sin ^+ 0.00078719 sin ^


4nx (Al-17)
0.00004724 sin

Least-Squares Method

In this procedure, the weighting functions are

&«*»• (Al-18)

Therefore,
2nx
W — sin-p,
x
W, sin
L
W 3 sin
3nx
W = sin Anx
jjr
A —f-
.

>
404 Solution of Beam Bending Problem Appendix 1

According to the least-squares method [Fig. A 1 -2(c)],

*(*)*§& =
1
or

R(x) sin -j- dx = 0,


I
R(x) sin —j- dx = 0,

(Al-19)
R(x) sin —j- dx = 0,

!*(*) sin ^<fr = 0.

The final four equations and the resulting approximate solution are
4.06 x 10 4 ai + 0xa 2 + 0xa 3 + 0xa 4 = 4774.65
x ai + 6.49 x 10 a 2
5
+ 0xa 3 + 0xa 4 = -795.77,
(Al-20)
x ai + x a2 + 3.29 x 10 6 a 3 + x a4 = 1591.55,
Oxai + 0xa 2 + 0xa 3 + 1.04 x 10 7 a 4 = -397.89,

w = 0.11760 sin ^ - 0.001226 sin


^+ 0.0004837 sin
^
- 0.00003830 sin ^ •
(Al-21)

Galerkins Method

It is incidental that the weighting functions W t


in the least-squares
method are the same as the functions (p t used for the Galerkin method. Hence,
in this specific case, both the Galerkin and least-squares methods yield the
same solutions.

Ritz Method

In the Ritz method [1,2] the potential energy in the body (beam) is

expressed in terms of the trial functions, and the resulting expression is

minimized with respect to a,. This leads to a set of simultaneous equations in


a For instance,
f
.

^ = \[ F &) dx -\j wdx < A1 - 22 >

with

p{x)=Pa + ^(Pb-Pa)- (Al-23)


Appendix 1 Solution of Beam Bending Problem 405

Minimization of IT, with respect to a gives t

dll p
= 0,

U
da 2 ~ '

(Al-24)
dUp
0,
da 3
dll p
= 0.
(?a 4

For this specific problem, the equations are the same as in the Galerkin and
least-squares methods, and the approximate solution is the same as in
Eq. (Al-21).
Comment: It may be noted that in the Ritz procedure the potential energy
is minimized for the entire beam in other words, the limit of the integral is
;

from to L. The concept is thus similar to the finite element method (Chapters
method, the minimization of
3-5), except that in the case of the finite element
11, is achieved for the domain composed of a patchwork of elements.

Finite Element Method

The method can be achieved in a manner


solution by the finite element
identical to that covered in Chapter 7, with the subdivision consisting of two
elements [Fig. A 1 -2(d)]. We can use Eq. (7-15) for generating the two-
element equations and then performing the assembly with boundary condi-
tions, and the solutions of the resulting assemblage equations are [3]

" 12 30 -12 30 ( W; '


1437.5'

30 100 -30 50 e, 1250.0

F -12 -30 24 -12 30 w2


>
= <
3750.0
(Al-25)
125 30 50 200 -30 50 417.0

-12 -30 12 -30 rw 3


2312.5

50-30
30 100_ [e 3l -1875.0,

Solution after introduction of w = w = x 3


gives

w = 0.000000,
t
w = 0.1171875,
rv 2 w 3
= 0.0000000,
(A 1-26)
0, = 0.036665, 6 = 0.0007295, 2
e -0.0383335.

Finite Difference Method

Before the era of the finite element method, the finite difference method
was the commonly used technique for problems in engineering and mathe-
406 Solution of Beam Bending Problem Appendix 1

matical physics. Details of this method are beyond the scope of this book.
However, we shall solve the beam problem using this method mainly to
complete the discussion of commonly used numerical methods. For study of
this method, the reader can refer to various textbooks [1].

The finite difference method is based on the concept of replacing the


continuous derivative in the governing differential equation by approximate
finite differences. For instance, various derivatives are approximated as (see
Fig. Al-3)

First derivative:
dw w i+1 — w t ^ w — w _i „w
~~
M—w
t f i
_ 1

dx ~ Ax Ax 2Ax

Second derivative:
d2 w w,_, — 2w t + w i+1
"
dx 2 Ax 2
(Al-27)
Third derivative:
d3 w _ — w,-_ + 2 2w,_ t
— 2w i+l + w i+2
dx 3 2(Ax) 3

Fourth derivative:
d4 w w i-2
-i
yv t — 4w f _, + 6vv,- — 4w /+1 + w t

dx 4 Ax 4
Use of Eq. (Al-27) to replace the fourth-order derivative in Eq. (7- lb) leads to

p Wj-2
— 4w -_
t 1 + 6w t
— 4w i+l + w i+2 (A 1-28)
Ax'
Pf

Substitution of / = 0, 1, 2, 3, 4, 5 for six points in the beam domain divided


into five segments [Fig. A 1 -2(a)] gives six simultaneous equations. Introduc-
tion of the boundary conditions [Eq. (Al-8)] finally gives four simultaneous

Figure Al-3 Finite difference approximation.

2 i-1 i i + 1 i + 2 i +3 x
:

Appendix 1 Solution of Beam Bending Problem 407

equations in wu w2 , vv 3 , and u> 4 :

5w - i
4w 2 + w 2 + =0.0115,
— 4h>! + 6h> 2 — 4w 3 + w4 = 0.0134,
(A 1-29)
w — \ 4w 2 + 6w 3
— 4w 4 = 0.0154,
+ w2 - 4w 3 + 5w 4 = 0.0173.
Solution.

w t
= 0.070656, w 2 =0.114432,
(A 1-30)
w 3
= 0.115968, w4 = 0.073344,
with w == w5 = 0.

COMPARISONS OF THE METHODS

To compare results from various methods, we first state the closed form
solution for the displacement based on the strength of materials theory [4]

w = ^(L - 2Lx + x
3 2 3
) -|-
(/7
f80ir^l
)X
(3^
4
- 10L 2 * 2 + 7L 4 ). (Al-31)

Results for displacements at typical locations on the beam by using


various methods and the closed form solution [Eq. Al-31] are compared in
Table AM.
TABLE Al-1 Comparisons for Displacements

Deflection,
Location from End A

Method 2 in. 4 in. 5 in. 6 in. 8 in.

Collocation 0.067 0.110 0.113 0.109 0.068


Subdomain 0.072 0.117 0.123 0.118 0.075
Galerkin 0.068 0.111 0.117 0.112 0.071
Least squares 0.068 0.111 0.117 0.112 0.071
Ritz 0.068 0.111 0.117 0.112 0.071
Finite element 0.067 0.110 0.117 0.116 0.069
Finite difference 0.071 0.114 — 0.116 0.073
Strength of 0.068 0.111 0.117 0.112 0.071
materials

The results for the first five methods and the last procedure are obtained
by substituting various values of* and
into Eqs. (Al-14), (Al-17), (Al-21),
(Al-31). Since we did not have a point at x = 5, there is no direct result at
that point for the finite difference method. In the case of the finite element
408 Solution of Beam Bending Problem Appendix 1

method, once the nodal displacements and slopes are obtained, values at
other points can be obtained by substitution of (local) coordinates into
Eq. (7-2). We can also compute moments and shear forces by using the
results with second and third derivatives, as we did in Chapter 7.
For examples of solutions of problems similar to the beam bending and
other problems solved by using different procedures presented herein, the
reader can refer to various publications such as Crandall [1].
For the foregoing beam bending problem, the results for displacements
from various procedures are close to each other and to the result from the
closed form solution. These comparisons are presented only for the sake of
introducing the reader to some of the available numerical procedures. The
merits of the finite element method become evident as we solve problems with
greater complexities in factors such as material and geometric properties and
loading characteristics.

REFERENCES

[1] Crandall, S. H., Engineering Analysis, McGraw-Hill, New York, 1956.

[2] Abel, J. F., (private communication).

[3] Desai, C. S., and Abel, J. F., Introduction to the Finite Element Method, Van
Nostrand Reinhold, New York, 1972.

[4] Timoshenko, S., Strength of Materials, Van Nostrand Reinhold, New York,
1956.

You might also like