Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6:...

22
Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D: Simplified (little) BVP in 1D: that satisfy SD and SS relationships above for all x, and u(x) = u D on ϕ(A D ) and Find displacement field u(x), strain field ε(x), and stress field S(x) Recall Big Picture: Ch6-ApproxNumSols Page 1

Transcript of Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6:...

Page 1: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Chapter 6: Numerical solutions to boundary value problems

Governing equations in 3D:

Simplified (little) BVP in 1D:

that satisfy SD and SS relationships above for all x, and u(x) = uD on ϕ(AD)and

Find displacement field u(x), strain field ε(x), and stress field S(x)

Recall Big Picture:

Ch6-ApproxNumSols Page 1

Page 2: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

The Ritz Method

Ch6-ApproxNumSols Page 2

Page 3: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Ch6-ApproxNumSols Page 3

Page 4: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Example: 1D Problem:

Ch6-ApproxNumSols Page 4

Page 5: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Matlab code for 1D Ritz method

Ch6-ApproxNumSols Page 5

Page 6: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Matlab code for 1D Ritz method (continued)

Ch6-ApproxNumSols Page 6

Page 7: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Results

Ch6-ApproxNumSols Page 7

Page 8: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Results

Ch6-ApproxNumSols Page 8

Page 9: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Summary of the method for obtaining approximate solutions:

Ch6-ApproxNumSols Page 9

Page 10: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Derivation of weak forms

Ch6-ApproxNumSols Page 10

Page 11: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Approximation functions

When choosing functions to approximate the real displacement u(x) and the virtual displacement ū(x),Make sure that they satisfy the continuity requirements imposed by the weak form (H1 for PVW) and,The approximation functions hi(x) are completei.e. they can converge to the exact solution in the limit i → ∞.

Condition number of K

In addition, one should try to make sure that the approximation functions are sufficiently different from one another. If the functions are not sufficiently different, it can lead to poorly conditioned system of equations K a = f

If condition number is large (~105 or larger) the computer will not be able to solve the system accurately.In order to keep the condition number small, we should use hi(x) functions that are linearly independent.

Orthogonal vectors and Orthogonal functions

Ch6-ApproxNumSols Page 11

Page 12: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Gram-Schmidt orthogonalization

In order to get smaller (better) condition numbers for the system matrix K, one should choose different (possibly orthogonal) approximating functions.It is possible to generate a set of orthogonal vectors (or functions) from a given set of linearly independent vectors (or functions) which are not necessarily orthogonal to each other.

Similarly for functions:

Ch6-ApproxNumSols Page 12

Page 13: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

1D Finite Element Basis

Approximation (Alternative implementation of HEBC)

1D Finite Element Basis Functions

Ch6-ApproxNumSols Page 13

Page 14: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Ch6-ApproxNumSols Page 14

Page 15: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Similarly for the force vector

Ch6-ApproxNumSols Page 15

Page 16: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

MATLAB Code

Ch6-ApproxNumSols Page 16

Page 17: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Element-wise computations for finite elements

Ch6-ApproxNumSols Page 17

Page 18: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Ch6-ApproxNumSols Page 18

Page 19: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Discretized weak form

Ch6-ApproxNumSols Page 19

Page 20: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Ritz and finite element methods for 2D and 3D problems

Ritz method Approximation:

Note: It may not always be easy find such functions for complicated shapes and boundary conditions.

Ch6-ApproxNumSols Page 20

Page 21: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Discretized weak form:

Approximations of strains and stresses

Finite Element approximations

In 2D In 3D

Ch6-ApproxNumSols Page 21

Page 22: Chapter 6: Numerical solutions to boundary value problemsaprakas/CE570/Ch6... · Chapter 6: Numerical solutions to boundary value problems Governing equations in 3D : Simplified (little)

Finite Element "Element-wise" approximations 2D

Finite element "Element-wise" approximation 3D

Ch6-ApproxNumSols Page 22