Introduction to FEM

9
MultiFreedom
Constraints II

IFEM Ch 9 – Slide 1
 Introduction to FEM

Penalty Function Method

Physical Interpretation

Recall the example structure

u 1 , f1 u 2 , f2 u 3 , f3 u4 , f4 u5 , f5 u6 , f 6 u7 , f7
(1) (2) (3) (4) (5) (6)
x
1 2 3 4 5 6 7

under the homogeneous MFC

u2 = u6

IFEM Ch 9 – Slide 2
 Introduction to FEM

Penalty Function Method (cont'd)

u 1 , f1 u 2 , f2 u 3 , f3 u4 , f4 u5 , f5 u6 , f 6 u7 , f7
(1) (2) (3) (4) (5) (6)
x
1 2 3 4 5 6 7

(7)

"penalty element" of axial rigidity w

    
1 −1 u2 0
w =
−1 1 u6 0

w = the "penalty weight" assigned to the constraint

IFEM Ch 9 – Slide 3
 Introduction to FEM

Penalty Function Method (cont'd)

Upon merging the penalty element the modified stiffness

equations are
    
K 11 K 12 0 0 0 0 0 u1 f1
 K 12 K 22 + w K 23 0 0 −w 0     f2 
   u2   
 0 K 23 K 33 K 34 0 0 0     f3 
   u3   
 0 0     f4 
 0 K 34 K 44 K 45 0   u4  =  
 0 0 0 K 45 K 55 K 56 0     f5 
   u5   
 0 −w 0 0 K 56 K 66 + w K 67   u 6   f6 
0 0 0 0 0 K 67 K 77 u7 f7

This modified system is submitted to the equation solver.

Note that u retains the same arrangement of DOFs.

IFEM Ch 9 – Slide 4
 Introduction to FEM

But which penalty weight to use?

From IFEM
Log10(of solution error norm)
Exercise 9.1
0
(Mathematica 4.2 on Mac G4)

-2

-4
constraint
gap error best w solution error
-6 dominates
dominates
Log10(w)
4 6 8 10 12 14

Rough guideline: "square root rule"; see Notes.

IFEM Ch 9 – Slide 5
 Introduction to FEM

Penalty Function Method - General MFCs

3u 3 + u 5 − 4u 6 = 1
Rewrite as matrix product

u3
[ 3 1 −4 ] u 5 =1
u6
Premultiply both sides by [ 3 1 −4 ] T




9 3 −12 u3 3 "Penalty element"
3 1 −4 u5 = 1
stiffness equations
−12 −4 16 u6 −4

Scale by w and merge:

K K 12 0 0 0 0 0
u   f 
11 1 1
 K 12 K 22 K 23 0 0 0 0   u2   f2 
 0 K 23 K 33 + 9w K 34 −12w 0   u 3   f3 + 3w 
   
 3w 
 0 0 K K K 0 0   u4  =  f4 
 34 44 45    
 0 0 3w K 45 K 55 + w K 56 − 4w 0   u5   f5 + w 
    
0 0 −12w 0 K 56 − 4w K 66 + 16w K 67 u6 f6 − 4w
0 0 0 0 0 K 67 K 77 u7 f7

IFEM Ch 9 – Slide 6
 Introduction to FEM

Assessment of Penalty Function Method

ADVANTAGES
general application (inc' nonlinear MFCs)
easy to implement using FE library and
standard assembler
no change in vector of unknowns
retains positive definiteness
insensitive to constraint dependence

DISADVANTAGES
selection of weights left to user - big burden
accuracy limited by ill-conditioning

IFEM Ch 9 – Slide 7
 Introduction to FEM
Lagrange Multiplier Method
Physical Interpretation
u 1 , f1 u 2 , f2 u 3 , f3 u4 , f4 u5 , f5 u6 , f 6 u7 , f7
(1) (2) (3) (4) (5) (6)
x
1 2 3 4 5 6 7
−λ λ
force-pair that enforces MFC
    
K 11 K 12 0 0 0 0 0 u1 f1
 K 12 0     
 K 22 K 23 0 0 0   u 2   f2 − λ 
 0 K 23 K 33 K 34 0 0 0     
   u 3   f3 
 0 0     
 0 K 34 K 44 K 45 0   u 4  =  f4 
 0 0 0 K 45 K 55 K 56 0     
   u 5   f5 
 0 0 0 0 K 56 K 66 K 67   u 6   f 6 + λ 
0 0 0 0 0 K 67 K 77 u7 f7

IFEM Ch 9 – Slide 8
 Introduction to FEM

Lagrange Multiplier Method (cont'd)

Because λ is unknown, it is passed to the LHS
and appended to the node-displacement vector:
 
  u1  
K 11 K 12 0 0 0 0 0 0   f1
 K 12 K 22 K 23 0 0 0 0 1   u2   f2 
   u3   
 0 K 23 K 33 K 34 0 0 0 0     f3 
   u4   
 0 0     f4 
 0 K 34 K 44 K 45 0 0   u5  =  
 0 0 0 K 45 K 55 K 56 0 0     f5 
   u6   
 0 0 0 0 K 56 K 66 K 67 −1     f6 
 u7 
0 0 0 0 0 K 67 K 77 0 λ f7

This is now a system of 7 equations and 8 unknowns.

Need an extra equation: the MFC.

IFEM Ch 9 – Slide 9
 Introduction to FEM

Lagrange Multiplier Method (cont'd)

Append MFC as additional equation:
    
K 11 K 12 0 0 0 0 0 0 u1 f1
 K 12 K 22 K 23 0 0 0 0 1     
   u2   f2 
 0 K 23 K 33 K 34 0 0 0 0     
   u3   f3 
 0 0 K 34 K 44 K 45 0 0 0     
   u4  =  f4 
 0 0 0 K 45 K 55 K 56 0 0     
   u5   f5 
 0 0 0 0 K 56 K 66 K 67 −1     
   u6   f6 
 0 0 0 0 0 K 67 K 77 0   u7   f7 
0 1 0 0 0 −1 0 0 λ 0

This is the multiplier-augmented system. The new

coefficient matrix is called the bordered stiffness.

IFEM Ch 9 – Slide 10
 Introduction to FEM

Lagrange Multiplier Method - Multiple MFCs

Three MFCs: u 2 − u 6 = 0, 5u 2 − 8u 7 = 3, 3u 3 + u 5 − 4u 6 = 1

Recipe step #1:

append the 3 constraints
K K 12 0 0 0 0 0   f1 
11
 K 12 K 22 K 23 0 0 0 0    f2 
 0 K 23 K 33 K 34 0 0 0 
u1  f3 
   u2   
 0 0 K 34 K 44 K 45 0 0    f4 
  u3   
 0 0 0 K 45 K 55 K 56 0   u4  =  f5 
 0 
K 67     
 0 0 0 K 56 K 66  u5   f6 
 0 K 77    
 f7 
 0 0 0 0 K 67  
 0 1 0 0 0 −1 0 
u6  0
  u7  
0 5 0 0 0 0 −8 3
0 0 3 0 1 −4 0 1

IFEM Ch 9 – Slide 11
 Introduction to FEM

Lagrange Multiplier Method - Multiple MFCs

(cont'd)
Recipe step #2: append multipliers,
symmetrize & fill

 K 11 K 12 0 0 0 0 0 0 0 0   u1   f1 
 K 12 K 22 K 23 0 0 0 0 1 5 0   u2   f2 
 0 K 23 K 33 K 34 0 0 0 0 0 3     f3 
   u3   
 0 0 K 34 K 44 K 45 0 0 0 0 0   u4   f4 
    
 0 0 0 K 45 K 55 K 56 0 0 0 1   u5   f5 
 0 −1
  = 
0 −4   u 6  

 0 0 0 K 56 K 66 K 67 f6 
 0 −8 0     f7 
 0 0 0 0 K 67 K 77 0   u7   
 0 −1 0     0
 1 0 0 0 0 0 0   λ1   
0 5 0 0 0 0 −8 0 0 0 λ2 3
0 0 3 0 1 −4 0 0 0 0 λ3 1

IFEM Ch 9 – Slide 12
 Introduction to FEM

Assessment of Lagrange Multiplier Method

ADVANTAGES
general application
exact
no user decisions ("black box")

DISADVANTAGES
difficult implementation
additional unknowns
loses positive definiteness
sensitive to constraint dependence

IFEM Ch 9 – Slide 13
 Introduction to FEM
MFC Application Methods:
Assessment Summary
Master-Slave Penalty Lagrange
Elimination Function Multiplier

Generality fair excellent excellent

Ease of implementation poor to fair good fair

Sensitivity to user decisions high high small to none

Accuracy variable mediocre excellent

Sensitivity as regards high none high

constraint dependence

Retains positive definiteness yes yes no

Modifies unknown vector yes no yes

IFEM Ch 9 – Slide 14