Adeola A Adedeji

. aaadeji@unilorin.edu.n, g gaiadeji@gmail.com 0023CV

Simplifying the finite difference method

in structural analysis
A A Adedeji
Department of Civil Engineering, University of Ilorin, Nigeria
gaiadeji@gmail.com, aaadeji@unilorin.edu.ng

Theory
When considering a plane element, applying the method of finite difference gives us reliable results as
we assume the use of Laplace and Poisson equations from the theory of torsion, which can be obtained
by the derivation of differential equation in a linear algebraic form. It is assumed, as shown in Figure 1,
that difference Δ = Δx = Δy. From the first and second order derivations, Equation (1) holds:

Fi , j Fi 1, j  Fi 1, j  2 Fi , j Fi 1, j  2 Fi , j  Fi 1, j

 ,  (1a)
x 2 x x 2
x

As well as,

Fi , j Fi 1, j  Fi 1, j  2 Fi , j Fi 1, j  2 Fi , j  Fi 1, j

 ,  (1b)
y 2 y y 2
y

From the above equations we can obtain another high derivation and as shown in Figure1, the second
derivation can be defined now along x-axis as:

 F i 1, j
'' ''
 3 Fi , j   Fi , j
2
F i 1, j Fi  2, j  2 Fi 1, j  2 Fi 1 , j  Fi  2, j
 ( )  (2)
x 3 x x 2 2 x 2 x
3

equally along y-axis.

For higher derivation (4th order):

 2 F i , j  F i 1, j
'' '' ''
 4 Fi , j   Fi , j
2
F i 1, j Fi 2, j  4 Fi 1, j  6 Fi , j  4 Fi 1, j  Fi 2, j
 ( )  (3)
x 4 x 2
x 2 x
2
2 x
4

equally along y-axis.

Δx Δx = Δ

i, j+2

i-1, j+1 i, j+1 i+1, j+1

Δy = Δ
i-2, j i-1, j i, j i+1, j i+2, j

Δy = Δ
i-1, j+1 i, j+1 i+1, j-1

y i, j-2
x
Figure 1. Differential points from a node (i, j – coordinate)
Adeola A Adedeji
. aaadeji@unilorin.edu.n, g gaiadeji@gmail.com 0023CV

An elementary computation yields the following equations:

 F i 1, j
'' ''
 2 Fi , j   Fi , j
2
F i 1, j Fi 1, j 1  Fi 1, j 1  (2 Fi 1 , j 1  Fi  2, j 1 )
 ( )  (4)
xy x xy 2 x 4xy

Likewise:

 2 F i , j  F i 1, j
'' '' ''
 4 Fi , j   Fi , j
2
F i 1, j
 ( ) 
x 2 y 2 x 2
y 2 x
2

Fi 1, j 1  2 Fi 1, j  Fi 1, j 1  2( Fi , j 1  2 Fi , j  Fi , j 1 )  Fi 1, j 1  2 Fi 1 , j  Fi 1, j 1

 (5)
 x y 2
2

By the use of derivation we can establish differential substitutions for a differential equation, i.e. bi-
harmonic equation:

ΔΔF = 0 (6a)

Simplifying, it can be assumed that step difference is found in both directions for Δx = Δy= Δ. In such
a condition we can obtain the differential equations:

20Fi , j  8( Fi 1, j  Fi 1, j  Fi , j 1  Fi , j 1 )  2( Fi 1, j 1  Fi 1, j 1  Fi 1, j 1  Fi 1, j 1 ) +

 Fi2, j  Fi2, j  Fi , j 2  Fi , j 2  0 (6b)

Equation (6b) corresponds to the points in Figure 2. If we assumed a rectangular element, differences
are taken for all the internal points of the element, where we can obtain 12 linear algebraic equations
with the unknown functions F1 to F12. These include all the boundary and internal nodal points

Δx Δx = Δ

2 -8 2

Δy = Δ
1 -8 20 -8 1 =0
Δy = Δ
2 -8 2

1
y
x F
Figure 2. Coefficients for nodal functions

The element boundary values of Fx are calculated in terms of boundary conditions. This has been
expressed clearly in Figure 3 for a vertical plain object, so that at side AB, function F A is obtained by
the help of the values of the internal nodal points. Along AB, the following relationship holds:

y
dF (0)
F ( y )    P ( )dd   F (0) (7a)
0 0 x
dy

And the second condition of such function along the x axis perpendicular to the boundary side is
expressed as:
Adeola A Adedeji
. aaadeji@unilorin.edu.n, g gaiadeji@gmail.com 0023CV

F ( y )
y
dF (0)
   Py , x ( )d  (7b)
x 0
dy
span
The same method is applied to other boundaries. A__
Integration constants are: Py
__
dF (0) dF (0) Px
F (0), ,
dx dy y
x B
As in Equation (7), each of the 2 sides must be chosen
so the joint functions are secured for the two derivations Figure 3 Boundary conditions
for the function stresses. The course of derivation of function
stresses is obtained with respect to x, so that by its substitution,
we can now write, deriving from Equation (1) referring to Figure 4, that:

Fa F  F4
 a ,
x 2 (8)
Fa
where : Fa  2  F4
x

The same method is applied to the remaining boundaries.

Δ Δ

F1 F2 F3 F4 a Fa

F5 F6 F7 F8

F9 F10 F11 F12

B
y

x
Figure 4. Coefficients for nodal functions

After the boundary(end) conditions are fulfilled, the basic Equation (6b) becomes homogenic which we
can now solve so as to obtain values of the functions of all the internal points. Components of stresses
can be computed from the following equations:

2F 2F 2F

x  ,   ,  (9)
y 2 x 2 xy
y xy

This can be expressed by substitution as shown in Equations (10) to (12):

1
 x, y  ( Fi , j 1  2 Fi , j  Fi , j 1 ) (10)

2
Adeola A Adedeji
. aaadeji@unilorin.edu.n, g gaiadeji@gmail.com 0023CV

1
 y, x  ( Fi 1, j  2 Fi , j  Fi 1, j ) (11)

2

1
 xy , y  ( Fi 1, j 1  Fi 1, j 1  Fi 1, j 1  Fi 1, j 1  Fi 1, j 1 ) (12)
4
2

The diagrammatical representations relating to Equations (10) to (12) are represented in Figure 5.
1
1 -1
1 -2 1 τxy 4∆2
-2 x∆2 y∆2

1 -1 1

Figure 5. Stresses at nodal points

Example
A rectangular slab shown in Figure 6 has three fixed-end (continuous) sides and a free end with
hydrostatic pressure P acting as triangular force against the element.

y ¾L

x
x
Po
L

Figure 6 Ground water tank slab

Solution
Using the finite difference method, we create the differences by applying the nodal points as shown in
Figure 7.

16 15 16

9 8 7 8 9

0 6 5 6 0 10

0 16 4 15 3 4 16 0 11

0 2 1 2 0 12

0 16 0 150 0 16 0

13 14 13

Figure 7. Differential assessment of nodes

Adeola A Adedeji
. aaadeji@unilorin.edu.n, g gaiadeji@gmail.com 0023CV

From Figure 7, we have to compute the nodal loads in points (1) to (6) due to symmetry nature of the
differences so that:

P1 = P2
P3 = P4 = (13)
P5 = P6 + 0

From differential equation for linear algebraic equation of the nodal points, the force function:

Fi,j = (14)

where w is an unknown factor, D = element stiffness. Referring to Figure 2 and using this to formulate
our equations as in Equation (6b) so that:

20 F1 - 8(F3 +2F2) + 2(2F4) - F5 + F14 =

20 F2 - 8(F4 +2F1) + 2F3 – F6 + F2+ F13 + F12 =
20 F3 - 8(F5 +2F4 + F1) + 2(2F6 + 2F2) +F7 = (15)
20 F4 - 8(F6 +2F3 + F2) + 2(F5 + F1) – F8 + F4+ F11 =
20 F5 - 8(F7 +2F6 + F3) + 2(2F8 + 2F4) + F15 + F1 = 0
20 F46- 8(F8 +2F5+ F4) + 2(F7 + F3 + F9) + F16 + F8+ F12 + F10= 0

The values of bending at the external nodal points are expressed using the internal nodal points on the
based on the end conditions:

 At the fixed end, x = constant, so that Fn = 0 and FA = Fn

F10 = F6
F11 = F4
and,
F12 = F13 = F2
F14 = F1

 At the free end and the condition of zero bending moment:

Mx,5 = - (F7 + 2F5 + F3) = 0 , (16a)
F7 = 2F5 - F3 (16b)
Mx,6 = - (F8 + 2F6 + F4) = 0 , (17 a)
F8 = 2F6 - F4 (18b)
F9 = 0 (19)

 At the zero shear forces

F15- 2F7 +2F3- F1 + 2(2F8 – 2F7- 2F4+ F3 )= 0 (20a)

F15 = F1 - 12F3 + 8F4+ 12F5 – 8F6 (20b)
F16- 2F8 +2F7- F2 + 2(F9 – 2F8 + F7+ F7 + 2F4 – F3)= 0 (21a)
F16 = F2 + 4F3 - 12F4 - 4F5 + 12F6 (21b)

Combining Equations (16) t (20) and solve, we have:

21 F1 - 16F2 - 8F3 + 4F4 - 1F5 + 0F6 = =

-8F1 +3F2 + 23F3 - 8F4 + 0F5 + F6 = =
-8F1 + 4F2 +19F3 + 16F4 - 6F5 + 4F6 = =148PoL2 (22)
2
2F1 + 8F2 + 8F3 + 21F4 + 2F5 + 6F6 = = 148PoL
2F1 + 0F2 - 12F3 + 8F4 + 165F5 - 16F6 = 0
0F1 + 2F2 - 4F3 + 12F4 - 8F5 + 18F6 = 0

So that:
Adeola A Adedeji
. aaadeji@unilorin.edu.n, g gaiadeji@gmail.com 0023CV

F1 = 0.01PoL2
F2 = 0.0067PoL2
F3 = 0.0134PoL2
F4 = 0.008PoL2
F5 = 0.0117PoL2
F6 = 0.007PoL2

For instance, the actual bending at point 5 free end, is:

= PoL4 = 0.000733 PoL4D

See Figure 8 for details of the bending moments at the slab axis of symmetry
-0,0088PoL2

5
M7 Mx
3

-0,020PoL2

Figure 8. Bending moment diagram at node 5

free end