CHAPTER 2

FINITE DIFFERENCE METHOD

1
2
3
 Strong Formulation – Formulation of Differential Equation
A Soil-Pile Interface Problem
Let’s use a simplified soil-pile
interaction problem to kick off
the discussion.

4
5
6
 Soil-Pile Problem – Formulate a Model
(1) Strong Formulation

(2)

Combine (1) & (2):

Constant E with depth Varying A with depth (E constant)

7
 Solution to the PDE
Analytical Solution

Incorporating boundary condition(s) into the solution.

1. Top boundary (z = 0): F = P

2. Bottom boundary (z=L):

Case 1: End-bearing pile (w |z=L = 0)

Case 2: Pile rested on compressible material

Sign convention adopted herein: negative downward force causes positive strain.

8
Boundary Condition – Case 1

9
Boundary Condition – Case 2

10
 Example

Assume that a pile has

length L = 7.0 m,
perimeter S = 1.7 m, and
cross-sectional area A = 0.002 m2.

The modulus of elasticity of the pile

E = 2.1x107 kPa.

The Winkler coefficient of side

resistance
Ks = 1000 kN/m3,

and the tip Winkler coefficient

Kt = 20000 kN/m3.

The load applied at the top of the

pile is 1000 kN.

11
 Analytical Solution

Axial Loaded Pile

Axial Displacement (m)

0 0.05 0.1 0.15
0

2
Depth (m)

5
Fixed Pile
6
Floating Pile

7
Quick Question:
Pile displacement for a fixed and floating pile What happen if I do not know the
(exact solutions) analytical solution to the PDE?

12
 Collocation Method (1)

Simple to use and has reasonable accuracy in simple problem.

1. Express the unknown by a simple function, say polynomials. The

function should either satisfy the boundary condition(s)
automatically or by assigning specific values to the function
coefficient(s).

2. Substitute the function into the governing PDE. Try to satisfy the
function at as many points as possible.

3. Substitute f(x) into the governing differential equation of the

problem. Try to satisfy the equation at as many points as
possible by assigning specific values to the coefficients of the
equation f(x).

Again, we use the soil-pile interaction problem as an example!!

13
 Collocation Method (2)

Assume

Case 1 – End-bearing pile

14
Collocation Method (3)

15
Collocation Method (4)

16
Collocation Method (5)

Any comment on this method ?

17
Least Squares Method (1)

18
Least Squares Method (2)

We only need two equations among them.

(4 unknowns; 2 boundary conditions + 2 equations)

19
Least Squares Method (3)

20
Least Squares Method (4)

Quick Questions:
1. Would you comment on these
two methods?
2. Any unique solution?

21
 Introduction to Finite Difference

Taylor Expansion:

a=xi; x-xi = 

a=xi; x-xi = -

f(x)

Bf(x+)

f(x-)
P f(x)
A
0 x- x x+

22
Finite Difference – 1st Derivative

Adding

f(x)

Bf(x+)

f(x-)
P f(x)
A
0 x- x x+
23
24
 Finite Difference – 2nd Derivative

2nd derivative
Subtraction

Addition
1st derivative
f(x)

Bf(x+)

f(x-)
P f(x)
A
0 x- x x+
25
 FD Quick Summary

1st Derivative

2nd Derivative

Quick Questions:
Could you derive higher-order
derivative?

26
27
 Simple Illustrative Example
Simple Parabolic PDE

Finite Difference
Approximation

Rewritten as
Meaning !?

28
29
30
31
32
 FD Approach to the Pile Problem

Governing Equation

FD Approximation

Re-written

Discretization

33
 Discretization and Nodes

Option 1
Node 2, 3, 4 are considered
(3 eqts with 5 unknowns)

Option 2
Node 1-5 are considered
(5 eqts with 7 unknowns)
(Note 2 fictitious nodes)

Quick Questions:
Two more equations are
needed. Where are they?
34
 Boundary Conditions

BC 1 (Pile head) :
Force

BC 2 (Pile Toe)
Displacement (End-bearing pile)

35
System of Equations

Expand

+ BCs

36
 FD Results

Quick Questions:
Can we use different ?
Pros and Cons?

37
38
Numerical scheme and stability

The previous recurrence relation is called the explicit scheme. It has been found
that the solution is numerically stable and convergent only when 0<r < ½. Otherwise
the solution becomes unstable and starts to wildly oscillate.

Let’s rewrite the formula as

 
ui, j1 ui, j

1
2
 ui1, j1  2ui, j1  u i1, j1  1 u  2u u 
t x   about t j1 about t j 
 = 0 gives the explicit scheme
 = 0.5 gives the Crank-Nicolson scheme (implicit/explicit)
 = 1 gives the implicit scheme

An assignment will be given on this topic to help you

better understand the subject

39
 Flux Boundary

For example:
No Flow:
Heat Flow
Fluid Flow
Dissipation of excess pore pressure Flow:

40
41
42
 One-dimensional Consolidation

Terzaghi’s One-dimensional Consolidation Equation

surcharge surcharge

43
44
45
FD Approximation

46
FD Approximation

47
Consolidation Problem – FD Procedures

48
In-Class Exercise

49
 In-Class Exercise (Supplementary)

i Initial uex = 100

5 100 0 0 0 0 0 0 0 0 BC1: uex (Top) =0

4 100 100 BC2: no flux at bottom

3 100 100
ui, j 1 
1
2
ui1, j  u i1, j 
2 100 100

1 100 100

0 100 100 j
0 1 2 3 4 5 6 7 8
1 2 3 4 5 6 7 8
Tv 0
32 32 32 32 32 32 32 32

50
 In-Class Exercise

Tv 0 1/32 2/32 3/32 1/8 5/32 6/32 7/32 1/4

j 0 1 2 3 4 5 6 7 8
Quick Questions:
How about time-dependent loading?
51
 Summary
• Model Formulation

• Collocation Method

• Least Squares Method

• Solution Uniqueness

• Finite Difference Method

• Implementation

Bring home message:

A tool to approximate the solution to PDEs.
The solution is only an approximation!!!

52
 Non-End-bearing Pile

Governing Eqt: (Constant E & A) Collocation and

d 2w  2 SK s Least Square Method
 w  0 where  
dz 2 EA
- Assume a form of w=w(z)
- Put back into BCs & GovE

BC1: Pile Head

Finite Difference Method
dw
P  1000  EA 1000
dz z0
BC1:  w1  w0
EA
1000  AKt w5
BC2: Pile Toe
L BC2:  w1  w2  w2  w3 
P  AK t w |z  L  SKswdz  
2 2
SK s z  
0 w3  w4 w4  w5
  
2 2 
53