Steel Plate Analysis by Finite Elements

N O R M A N L. F I R K I N S

Using the coordinates of the assumed element con

figuration in Fig. 1, the area of the element may b
expressed in determinate form as

T H E FINITE ELEMENT m e t h o d is becoming a popular as

well as practical way of analyzing plates, shells, a n d

related structures. T h e technique of modeling* an elastic
structure into an assemblage of discrete bars or sections
and interconnecting these elements by a finite number of
joints or nodal points is not new. 1 T h e growth a n d development of the finite element method has been greatly
enhanced by the high speed digital computer a n d matrix
methods in structural mechanics.
Extended finite element formulation is being carried
on to increase the potential of this method in the areas
of plasticity, vibration, a n d three-dimensional stress
applications.
Although continuous investigation is being undertaken, it might be in order to review the basic finite
element formulation a n d provide a simplified method of
evaluation for the practicing engineer.

Xi yi

A = V<

(2
X71 Jn

It can be shown 3 that through certain relationship;

substitutions, a n d elimination of a\, a2 , a n d az, a
area-coordinate matrix [B] will relate element strai
and nodal displacement, and may be expressed thus:

{} = [B]{A]

(2

In matrix form,

[B]

ELEMENT CHARACTERISTICS

Cj

Cn

di

dj

dn

cZ i

Ci

u, j

Cj

=h

UJI

a n d d = yj yn , Cj = yn yi, cn = yi yj

In order to analyze and express the element properties in

matrix form, certain characteristics 2 of an elastic element
must be assumed, i.e., the material making u p the
element must be isotropic a n d homogeneous; Hooke's
Law and Saint Venant compatibility equations apply
under small deflection theory.

a n d di = xn Xj, dj

A ^

sx 72, % Us yx

J\ j

*\ %

In this discussion, it is assumed that the elastic ek

ment will obey Hooke's Law and that the following stress
strain relationship 4 exists for plane stress as well as plan
strain. Using the following formulations:

GENERAL EQUATION FORMULATIONS

With these characteristics accounted for, a set of simultaneous equations may be used to describe the linear
relationship that exists between the nodes in Fig. 1
under rigid body displacements:
u% = a\ + a2Xi + a^yi
Uj = fli + a2Xj

a^yj

un a\ + aixn + a%yn

a)

* Providing the ability to reduce, synthesize, and properly represent

the structural configuration.

-x

Norman L. Firkins is a Civil Engineer with Henningson, Durham &

Richardson, Consulting Engineers and Architects, Omaha, Nebraska.
Figure 7

130
AISC

ENGINEERING

JOURNAL

(4

Cyi

1
= T, fax V <JV)

- fay -

(5)

V <TX)

2 (1 + v)
7xy

Expressed in matrix form:

(6)
Plane stress and plane strain may be determined by
using the appropriate matrix [D].

Figure 2

For plane stress,

[D] =

(1

0
0

"1

(1

T h e normal counter-clockwise assignment of z, j ,

and n to the nodes of assembled elements is shown in
Fig. 2.
T o develop the total stiffness of a particular structural
system, the elemental stiffnesses must then be assembled
into a total stiffness matrix to express the compatibility
of adjacent triangles and their respective influence on
nodal point displacements. This total stiffness assemblage
takes the form shown in matrix (12), See Fig. 3.

"
(7)

-v)

and for plane strain,

(i - sy (1 - v2)
vi\ + v)

MI

v{\ + v)
(1 - v2)

0
0
(i -

+ v)f

(8)

[*(1 + )]*
2 (1 + v)

v*y

Using the principal of virtual work and equating

internal and external work equations, a generalized
nodal point stiffness matrix [k] can be defined for an
element in the matrix form:
[k] =

[B]T[D][B]dVol

to

(12)

(9)

J Vol

or simply for an element:

[k] =

(10)

[BY[D][B]tA

or in final matrix expansion, adaptable for computer

processing, thus:

1
[*]. =" 2A

Ci

dil

di

Ci

dj

dj

1 v
V 1

<W)

dn
dn Cn]

Cn 0
0

2A
iXii

J^-ij

K-ji

**- 33

0
0
v

Figure 3

It will be noted in Eq. (11) that each Krc is a 2 by 2

matrix and that each row (r) and column (c) thereof
must contribute properly to the overall makeup of the
final stiffness matrix. This relationship must exist 5 with
respect to off diagonal as well as diagonal terms.
With the final stiffness matrix generated, the desired stiffness equation can be written in matrix form:

I
X

J
o 1

| Ci

Cj

cn

di

dj

dn

l_di

Ci

dj

Cj

un

cn j

tA

{F} = [K]{A]

Inverting the appropriate [K] matrix, the respective

nodal point displacements may be found:

-**- in
f^-in

(13)

(11)

{A} = [tf]-i{F}

(14)

131
O C T O B E R / 1969

Using these nodal displacements, the following m a y

be determined:
E
Z)
t
w

1. Element stresses:
{a} =

[D][B){A\

(15)

2. Element strains:
{e} = [B] {A}

(16)

36"

3. Angle of maximum and minimum normal stress:

(17)

tan 2d =
4. M a x i m u m and minimum principal stresses:
ax + a
0"max =

OV

30,000,000
0.25
0.10
100 llVin.

Figure 4

(18)

j *

<rx + a
0"min

+ rxy>

A
A
/ \
A
A
A
A

=
=
=
=

+ r, y

0"2

(19)

ELEMENT REPRESENTATION

In triangular finite element representation, it has been

established that for proper accuracy the element configuration should resemble or closely approximate an
equilateral triangle. Although quadrilateral elements
are currently being used in analysis with excellent results, triangular division seems to be desirable.
STEEL PLATE SAMPLE PROBLEMS

Figure 5

E x a m p l e 1Consider a typical steel plate shown in

Fig. 4.
It is required to compute the displacements and
internal stresses and strains on the loaded plate according
to the finite element method. T h e structure modeled into
finite elements is shown in Fig. 5.
Using matrices (2), (4), (7), compute the areas,
[B]T, [JD], [B], and [k] for elements Ei and E 2 , as in
matrices (20) and (21), Fig. 6.
With these elemental stiffness matrices, generate the
total stiffness matrix (22) (see Fig. 7).
T h e total stiffness matrix (22) may then be blocked
or deleted, according to boundary conditions, to provide
the appropriate stiffness matrix [K]; a normal inversion
, computer program will then provide the inverse of [K]:

[*] =

[*]-

0.2799
-0.9999
-0.3999
0.3999

"0.4394
0.2433
0.1064
0.4257

E
E
E
E

E
E
E
E

07
06
06
06

06
06
07
07

-0.9999
0.1966
0.5999
-0.1066

0.2433
0.9286
-0.1642
0.4541

E
E
E
E

E
E
E
E

06
07
06
07

06
06
06
06

-0.3999 E 06
0.5999 E 06
0.2799 E 07
0.0

0.1064
-0.1642
0.3938
-0.9124

E
E
E
E

07
06
06
07

0.3999 E 06
-0.1066 E 07
0.0
0.1966 E 07

0.4257
0.4541
-0.9124
0.7461

E
E
E
E

07
06
07
06.

With the distributed load, w, a load vector {F} i

assembled to express the required nodal forces. Usin
this load vector and (24), the displacements of node
points 3 and 4 are found to be
{A} =
[K]-i{F]
=
0.0008102
Xz
yz
=
0.0001423
x4 =
0.0007281
yA
= -0.0000875

ENGINEERING

JOURNAL

(2f

where
FxZ = F x 4 = 1800.0 lbs
Fyz = FyA =
0.0 lbs

(23)

Normally, for triangular elements, it is assumed tha

the computed stresses are to be applied at the centroi
of the element under consideration.
Using the computed nodal displacements (25), a:
element, such as E 2 , can be analyzed for point stresse
and strains. This is demonstrated as follows:

(24)

132
AISC

in.
in.
in.
in.

FOR

A, =

M
M,
M,

H,

ELEMENT

.432 000E 03

416666E-01 .000000E-99
000000E-99 277777E-01.000000E-99 . 000000E-99 000000E-99-".277777E-01
.416666E-01 . 000000E-99
.000000E-99 000000E-99

C-I-

.277777E-01
.416666E-01
277777E-01
000000E-99
000000E-99
.416666E-01

.320000E 08 .800000E 07 .000000E-99

.800000E 07 .320000E 08 .000000E-99
.000000E-99 .000000E-99 . 120000E 08

-.416 666E-01 .00 0000E-99 . 000000E-99 .0000 0OE-99 416666E-01 . 000000E-99

.00 00 00E-99 27 7 7 77E-01 .0 0000OE-99-.27777 7E-01 000000E-99 .000000E-99
.277777E-01-.416666E-01-.277777E-01 .000000E-99 i()0 0000E-99 . 41 6666 E-01

.279999E
999999E
.399999E
.399999E
.239999E
.599999E

FUR ELEMENT

A2=

El

07-.999999E
06 ,1966666
06 .599999E
06-.106666E
07 .399999E
06-.899999E

06-.399999E 06 .399999E 0607 .599999E 06-.106666E 07

06 .399999E 06-.000000E-9907-.000000E- 99 .106666E 0706-.000000E-99-.399999E 06
06-.599999E 06 .000000E-99

239999E 07 .599999E 06
.399999E 06-.899999E 06
.000000E-99-.599999E 06
.399999E 06 000000E-99
.239999E 07 .000000E-99
000000E-99 .899999E 06

(20)

E2

.432000E 03

-.416666E-01
.000000E-99
.416666E-01
.000000E-99.000000E-99
.000000E-99

.000000E-99
.000000E-99.000000E-99.277777E-01
, 000000E-99
.277777E-01

.000000E-99
.416666E-01
.277777E-01
>416666F~01
,277777E-(U
000000E-99

[>!

.320000E 08
800000E 07
000000E-99

[B>

416666E-01 .000000E-99 416666E-01 .000000E-99 .000000E-99

.000000E-99 .000000E-99 .000000E-99-.277777E-01 000000E-99
.000000E-99-.416666E-01-.277777E-01 .416666E-01 .277777E-01

&a

239999E 07-.000000F-99-.239999E
000000E-99 .899999E 06 .599999E
.239999E 07 .599999F 06 .279999E
399999E 06-.899999E 06-.999999E
000000E-99-.599999F 06-.399999F
.399999E 06-.000000E-99 .399999E

800000E 07 , 000000E-99
.320000E 08 .000000E-99
, 000000E-99 , 120000E 08

07 .399999E
06-.899999E
07-.999999E
06 .196666E
06 .599999E
06-.106666E

.000000E-99
.277777E-01
.000000F-99

06 .000000E-99-.399999E 06
06-.599999E 06-.000000E-99
06-.399999E 06 .399999E 06
07 .599999E 06-.106666E 07 (21)
06 .399999E 06 .000000E-99
07 .000000E-99 .106666E 07

Figure 6
133
OCTOBER/

1969

,279E 07-.999E 06- .399E 06 .399E 06 .000E-99 .000E-99-.239E 07 .599E 06

.999E 06 .196E 07 .599E 06-.1066 07 ,0006-99 .000E-99 .399E 06-.899E 06
.399E 06 ,599E 06 .280E 07-.000E-99-.2396 07 .399E 06 .000E-99-.100E 07

.3996 06-.106E 07 .0006-99 .1966 07 .5996 06-.8996 06-.100E 07 .0006-99

TOTAL

.000E-99 000E-99- .2396 07 .599E 06 .279E 07-.9996 06-.3996 06 .3996 06

.000E-99 .000E-99 .3996 06-.8996 06-.9996 06 .1966 07 .5996 06-.1066 07
.239E 07 .399E 06 .0006-99-.1006 07-.3996 06 .5996 06 .2806 07 .0006-99
.599E 06-.899E 06- .1006 07-.000E-99 .3996 06-.1066 07 .0006-99 .196E 07
Figure 7

For elemen t E 2 :

A ^

Compute stresses:
E2

<Jy
r

xy

\(JJ.

= [D][B]{A}
= 1029.198 psi
=
65.692 psi
=
43.795 psi

1/" 0

(26)

2K

A
4\ W.

Compute strains:
[B]{A]

#2

Jxy

I* (.)

0.3375 E - 04 in./in.
= 0.6387 E - 05 in./in.
= 0.3650 E - 05 in./in.

From (26), compute the maximum and

principal stresses:

<*x + TJL

_L

\((Tx

(27)

minimum

\ ,

ay

,-

17

IS

_ *z+ *y __ J(x
1\

- *y\
9

0/

/16

0/

'n
00

19

\0

\ /20

^\ 0/
\0

Figure 8

MESH CONFIGURATION

DISPLACEMENTS

3-ELEMENT

4-ELEMENT

* 5 = 0.0007698

x 2 = 0.0007777

Figure 9

134
JOURNAL

25

'}

M o d e l AccuracyIt will be noted in Eq. (25) that the

symmetrically loaded plate did not produce symmetrical
displacements. This situation can be eliminated when
finer mesh size (more elements) are used and relatively
constant triangle shape is maintained. T o represent
this, consider modeling Fig. 4 into a mesh size of 3, 4,
and 108 elements as shown in Figs. 8a, 8b, and 8c,
respectively. As indicated in Fig. 10, the nodal displacements of the fine mesh system are in close agreement with
those shown in Fig. 9.
T h e previous two, three and four element configurations were analyzed on an I B M 1620 computer and the
108 element fine mesh plate on an I B M 360/65 with a
2314 storage device.

ENGINEERING

/I8

22

= 63.706 psi

AISC

= 1031.184 psi

v mm

0/ \ 0 0/ \0 0/ \0
/0
0 / \(lL 0/ \0 0 /
^ \0
0 \ /0
/ 0 0\ /0

k2;

SAVPLP

PP^RLFM PLATE 108 - F L F M F M S

NO OF P A R T I T I O N S
10
NO OF NODAL POINTS
70
NO OF ELEMENTS
10S
NO PF BOUNDARY POINTS
10
NO OF LEADING
1
N ) 0^ ^ATFP IAL TYPFS
1
TYPE OF ANALYSIS PLANE STRESS
NO OF
FORCF
FORCE

L ' E C P F F S OF F R E E D r IM
X
PER U N I T
VOLUME
Y
ER U N I T
VOLUME

JOINT

COORDINATES
X
0.0
4.C0000

2
7.

2
0.0
0.0
Y
36.00000
36.00000
36.00000
36.C0000
36.00000
36.C0000
36.C0000
32.OOOOO
32.00000

8.00000
12.C0000
16.C0000

4
S

?0.ooooo

6
7

24.CO000

o.o

p
Q

4.00000
FLT^FNT
UESC.RIPTION
jni\'T
*ATERlAL
TYPF
CONNECTIVITY

ND
1
2
3
4

1
P

8
o

?
2

Q
10

2
2
10
3

NODE X - C I SPLACFMFN'TS
..1
C.00000000
2
'
C00C14710
3
0.00028421
4
0*00041395
5
C00054372
6
C.00067482
7
C.000RC776
3
"" C . 0 0 0 0 0 0 0 0
^
C.0001^270

ELEMENT NUMHER
X-SIRESS
1

1014.5633^008
3

2
7

J_^_A- A7
4

10

P C

10

_5_
c

10

1.3333330?
C

C O

0.10300
0.10000
0.10000
0.10000

0.0
C O
C O

SECOND NODE
__X.VST Rr SS

THIRD
NODE
STRESS-;!

2.666666C3

33.33332825

77.72729492
5.^3333302

6.66666603

AND
Y
COORDINATES OF CENTRQID
..STRESS-.?....
PR INC IPAL ANGLF

34.66665649
-205.4 00466)2

74.9Q78C273

THICKNESS

2 4.20507812

_?_ _L

1035.9^33^844

ANGLE
nr
_.FRTHCTRnpY

Y-D I SP L AC FMF N T S
-O.OCCOOOOO
-0.00006847
-0.C0010621
-C.C0C12849
-0.0CC14C75
-O.COC15C80
-CCCCI5997
-O.OOCOOCOO
-C.C0C04501

FIRST
NODE
.__YT_S_T.iESS

1 176.21C449?
2

1
1
1
1

-91.81152344

1 3 2 2 . 2 7*578 5 16

248.74609375

-77.520446*8

1023.57519531

68.81542969

- 84.45571899

10 0 6 . 3 2 8 1 2 5 0 0

72.41748047

10 3 8 . 7 3 5 8 3 9 8 4

27.81396484

33.33332825
-49.02273560

86.986 81641,

34.66665649

30.56640625

-52.67968750

9_.?3.3? _3302

34.66665649

-87.00878906

77

Figure 70
135
OCTOBER / 1969

Example 2Given the cantilever plate shown in Fig.

11a, determine the stress distribution on a stress block
taken from element 32 under the given loading condition.
T h e plate configuration is first modeled into a working finite element mesh (Fig. l i b ) . Next, the proper
boundary conditions and specific load vectors are
applied. Using the appropriate matrix formulation, the
state of stress at the centroid of element 32 is found to be
the following:
<TX
<Jy
T

xy

= 3175.4 psi
= - 7 3 3 . 7 psi
161.2 psi
=

-0 = 0.25

A unit square stress block is shown in Fig. 12a. T h e

state of stress at that point (centroid of element) may
be shown more clearly by expanding the point into a
free body and by arbitrarily verifying that the vertical
stress components are in equilibrium (Fig. 12b). Computations are as follows:
Compute maximum principal stress (positive):
<jx + <Jy

+ rx

+
_ 3175.4 "

733.7

*"

3175.4 + 733 .TV

(161.2) 2

= 3182.8 psi

Figure 77

Compute principal angle:

tan 2a =
ax o V
2 (161.2)

3175.4 -

(-733.7)

= 2.35

<

= 87.65 (from vertical)

Check vertical c omponents foi' equilibrium (SF ?

Down
161.1

0):

(1) (D = 161.0

Up:
733.76 (1) (0.042)
= 30.0
3182.13 (1) (1.00176) (0.041) = 131.0
161.0
Check
161.0 = 161.0

O.K.

It must be stated here that this mesh is usable but

it should be realized that high stresses will occur near
and at points of applied loads and boundary supports.
These areas of high stress concentration may be easily
examined by creating a finer mesh of elements near the
points in question and reanalyzing the structure.

136
AISC

ENGINEERING

JOURNAL

{A}

= Displacement vector

CONCLUSIONS

[D]

= Elastic stress-strain matrix

It has been found that many types of irregular plate

configurations may be analyzed for areas of high stress
concentration using the finite element approach. Basic
element stiffness matrices and solutions to "small elem e n t " structures may be programmed on small to
medium size digital computers. However, when trying
to evaluate irregular structures using fine mesh subdivision, it becomes necessary to go to large capacity
computers.
Basic investigation has shown that with regular mesh
modeling, the element stiffness equations are identical
to those of the finite difference approach and, in addition, provide a more systematic method for computer
programming.

[k)e

= Element stiffness matrix

2.

SYMBOL NOMENCLATURE

3.

hh

[KI total

= Total stiffness matrix

[K]

= Final deleted stiffness matrix

{F}

= Nodal force vector

= Principal angle

= Poisson ratio

= Shearing stress

1.

Element nodal points or joints

= Area of element

[B]

= Area-coordinate matrix

{e}

= Strain vector

[a]

= Stress vector

4.
5.

REFERENCES
Hrennikojf, A. Solution of Elasticity by the Framework
Method ASCE Journal of Applied Mechanics, Vol. 63, pp.
769-775, 7947.
Chou, Pei Chi and Pagano, Nicholas J. Elasticity Tensor,
Dyadic and Engineering Approaches D. Von Nostrand
Company, Inc., 7967, Chapters 7 and 2.
Dodson, Ronald G. Discussion of the Finite Element Method
of Structural Analysis'Report-CE 516 University of
Nebraska at Omaha, May 76, 7968.
Zienkiewicz, 0. C. and Cheung, Y. K. The Finite Element
Method in Structural and Continuum Mechanics McGrawHill Publishing Company, 7967, Chapters 2, 3 and 74.
Rubinstein, Moshe F. Matrix Computer Analysis of Structures Prentice Hall, Inc., 7966, Chapter 72.

137
OCTOBER / 1969