JN Reddy - 1 Lecture Notes on NONLINEAR FEM The Finite...
Transcript of JN Reddy - 1 Lecture Notes on NONLINEAR FEM The Finite...
JN Reddy
The Finite Element Method
Read: Chapter 6
2D Nonlinear Finite Element Analysis
CONTENTS
• Weak Form development• Finite element model• Computation of tangent
coefficients• Review of numerical integration• Computer implementation• Summary
JN Reddy - 1 Lecture Notes on NONLINEAR FEM
JN Reddy Nonlinear Problems (2-D): 2
Finite Element Formulation of a Model 2D Nonlinear Equation
Model Equation
Weak Form
Primary and Secondary Variables
11 22
11 11 , , 22 22 , ,
0 (1)
( , , , , ), ( , , , , )x y x y
u ua a f
x x y y
a a x y u u u a a x y u u u
11 220 (2)e
e
i h i hi i n
w u w ua a w f dx w q ds
x x y y
11 22PV: SV: h hh n x y
u uu q a n a n
x y
JN Reddy - 2 Lecture Notes on NONLINEAR FEM
JN Reddy Nonlinear Problems (2-D): 3
FINITE ELEMENT FORMULATION …(continued)
Finite Element Model
11 221
1 1
11 22
0
( ) ( )
e
e
e
e
e
e
nj ji i
jj
i i n
n ne e e e e e eij j i i ij j i
j j
j je i iij
ei i i n
u a a dxdyx x y y
f dxdy q ds
K u f Q K u F
K a u a u dxdyx x y y
F f dxdy q d
s
The coefficient matrix [K] is a function
of the nodal values of u
1
( , ) ( , ) ( , )n
e e eh j j
j
u x y u x y u x y
Finite Element Approximation
x
y
Γ
Γ
Ω e
e
nα
Ω
∧
•
•
•
• eu2
Degrees of freedom
12
34
eu1
JN Reddy - 3 Lecture Notes on NONLINEAR FEM
JN ReddyNonlinear Problems (2-D): 4
COMPUTATION OF TANGENT COEFFICIENTS
1
1 1
or [ ( )]n
e e e e e e e ei ij j i
j
e en nip p ipe e e e ei
ij p ip p ijj j j jp p
R K u F R K u u F
K u KRT u K u K
u u u u
0 1 0 1
11 11 11 22 22 22
11 22
1
1 1 1 1 111 2211 11 11 , 22 22 ,
1
1
Suppose that and . Then
e
np pe e e i i
ij ij pj jp
n
p p j jj j j j jp
e eij ij
a a a u a a a u
a aT K u dxdy
u x x u y y
a au ua a u a a a
u u u u u
T K a
1 11 22
e
i ij
u ua dxdy
x x y y
JN Reddy - 4 Lecture Notes on NONLINEAR FEM
JN Reddy
NUMERICAL EVALUATION OF INTEGRAL COEFFICIENTS
• Transformation of the integrals posed on arbitrary-shaped element to the master element domain so that evaluationof the integrals is made easy.
• The Gauss integration rule that evaluates an integral expression as a linear sum of the integrand evaluated at certain points (Gauss points) and weights (Gauss weights)is used.
x
y
Γ
Γ
Ω e
e
nα
Ω
∧
•
•
•
• eu2i
Degrees of freedom
12
34
eu1
1
( , ) ( , ) ( , )n
e e eh j j
j
u x y u x y u x y
1
( ( , ), ( , ))
( , )
eh
ne ej j
j
u x y
u
Numer Integration: 5
JN Reddy - 5 Lecture Notes on NONLINEAR FEM
JN Reddy
ξ
η
ξ = 1
η = 1
x
y
eΩ
1Ω 2Ω 3ΩΩ
ξ
η
eΩ
ηξ ddJdydx =
)1(),1( ηη ,yy,xx ==
)1(),1( ,yy,xx ξξ ==)()(
y,xy,x
ηηξξ
==
)()(
ηξηξ,yy,xx
==
y
x
NUMERICAL INTEGRATION
1
1
ˆ( , ) ( )
ˆ( , ) ( )
me ej j
j
me ej j
j
x x ,
y y ,
Numer Integration: 6
JN Reddy - 6 Lecture Notes on NONLINEAR FEM
JN Reddy
11 22
ˆ
1 1ˆ
( , ) ( ( , ), ( , ))
ˆ( , ) ( , )
e
e
j je i iij
ij ij
NGP NGP
ij I J ij I JI J
K a a dxdyx x y y
F x y dxdy F x y J d d
F J d d W W F
NUMERICAL EVALUATION OF INTEGRALS
[ ]
i i i i i i
i ii i i i
x y x yx y x xJ
x yx yy yx y
Computer Implementation: 7
Using the chain-rule, we obtain
JN Reddy - 7 Lecture Notes on NONLINEAR FEM
JN Reddy
NUMERICAL INTEGRATION implementation point of view
1 1
1 1
1 11 2
2 2
1 2
ˆ ˆ
ˆ ˆ
ˆˆ ˆ
,ˆˆ ˆ
m mi ii ii i
m mi ii ii i
m
m
m m
x y x yJ
x yx yη η η η
x yx y
d
x yη η η
xdy J d d
1 *
i ii
i i i
x J J
y
Jacobian matrix
Global derivatives in terms of the local derivatives
Computer Implementation: 8
JN Reddy - 8 Lecture Notes on NONLINEAR FEM
JN Reddy
ELEMENT CALCULATIONS
11 22
* * * *11 11 12 11 12
ˆ
* * * *22 21 22 21 22
( , )
( , )
e
j je i iij
j ji i
j ji i
K a a dxdyx x y y
a J J J J
a J J J J
1 1
ˆ 1 1
1 1
ˆ ˆ( , ) ( , )
ˆ ( , )
e eij ij
NGPNGPeij I J I J
I J
J d d
F d d F d d
F WW
Computer Implementation: 9
JN Reddy - 9 Lecture Notes on NONLINEAR FEM
JN ReddyComputational Issues: - 10
η
ξ
ξ = − 13 ξ = 1
3
η = 13
η = − 13
η= 0
η
ξ
ξ = − 35 ξ = 3
5
η = 35
η = − 35
ξ= 0
η= 0.861...
ξ= − 0.861...
ξ
ηξ= 0.861...
η= 0.339...
η= − 0.339...η= − 0.861...
GAUSS QUADRATURE
Domain of the masterelement
Domain of the physical element
Gauss points Gauss weights
ˆ
, 1
ˆ,
ˆ ,
( ) ( )
( )
e e
ij ij
N
I J ij I JI J
F x y dxdy F ξ,η dξ dη
W W F ξ η
JN Reddy - 10 Lecture Notes on NONLINEAR FEM
JN Reddy
Variables used in the program, FEM2DRead the input data description for FEM2D to know the meaning of all input variables; other variables are described below.
Computer Implementation: 11
COMPUTER IMPLEMENTATION (2D)
NPE - nodes per element, ELXY( , ) - Global coordinates of the th node of element , ( , )ELK( , ) - Element coefficient, ELF( ) - Element coefficient, ELU( ) - Element solution,
e ei i
eij
ei
ei
ni j i e x y
i j Ki fi u
( )tanTAN( , ) - Element coefficient,
0, 1, Coefficients in the definition of ( ) :( ) 0 1 * *
eiji j K
AX AX AU a xa x AX AX x AU u
−= + +
JN Reddy - 11 Lecture Notes on NONLINEAR FEM
JN Reddy 2-D Problems: 12
A11 = A10 + A1X*X + A1Y*Y; A22 = A20 + A2X*X + A2Y*Y; A00=CONSTIEL = 1, Linear rectangular elementIEL= 2, Quadratic rectangular elementGLXY(I,1) – Global x coordinate of the Ith global nodeGLXY(I,2) – Global y coordinate of the Ith global nodeNEQ – Number of equations in the mesh
SFL( ) Element hape (or approximation) funtion, DSFL( , ) Derivative of the th shape function with respect to
the local (normalized) coordinates and
( 1) : , ( 2) :
GDSFL( , ) Derivative o
ei
i i
i si j i
j j
i j
−−
∂ ∂= =
∂ ∂− f the th shape function with respect
to the global coordinates and : andi i
i
x yx y ∂ ∂
∂ ∂
COMPUTER IMPLEMENTATION (2D)JN Reddy - 12 Lecture Notes on NONLINEAR FEM
JN Reddy
Variables used in the program, FEM2D (continued)
1 2
1 2
( ) , , , , ( , ) , ( , )
( , ) , ( , )
i ii
i i
SF i XI ETA DET J DSF i DSF i
GDSF i GDSF ix y
1 2( ) , ( ) , ( , ) , ( , ) , ( , )i i ij i iELU i u ELF i f ELK i j K ELXY i x ELXY i y
Computer Implementation: 13
( , ) 11 * (1, ) * (1, ) 22 * (2, ) * (2, )F i j A GDSF i GDSF j A GDSF i GDSF j
11 22
ˆ
1 1
11 22
( , )
( , )
( , )= ( , ) ( , )
e
j je i iij ij
NGP NGP
I J ij I JI J
j ji iij
K a a dxdy F J d dx x y y
W W J F
F a ax x y y
CONST
COMPUTER IMPLEMENTATION (continued)
JN Reddy - 13 Lecture Notes on NONLINEAR FEM
JN Reddy
Initialize global Kij, fi
Iter = iter + 1
DO 1 to n
Iter = 0
Impose boundary conditionsand solve the equations
CALL ELKF to calculate Kij(n)
and fi(n), and assemble to form global Kij and Fi
Transfer global information(material properties, geometry and solution)
to element
Yes Iter < itmax no
Error < ε yesno
STOP
Print Solution
Write a message
STOP
GENERAL LOGIC FOR THE NONLINEAR ANALYSIS
Logic in the MAIN program
JN Reddy - 14 Lecture Notes on NONLINEAR FEM
JN Reddy
COMPUTER IMPLEMENTATION for the nonlinear analysis
ITER=0160 ITER=ITER+1
IF(ITER.GT.ITMAX)THENWRITE(IT,1200)ITMAXSTOP
ENDIFC Initialize the global coefficient matrices and vectors
DO 180 I=1,NEQGLF(I)=0.0DO 180 J=1,NHBW
180 GLK(I,J)=0.0C Compute element matrices and assemble the matrices
DO 250 N=1,NEMDO 200 I=1,NPENI=NOD(N,I)IF(NONLN.GT.0)THEN
ELU(I)=GLU(NI)ENDIFELXY(I,1)=GLXY(NI,1)ELXY(I,2)=GLXY(NI,2)
200 CONTINUE……
250 CONTINUE
Logic in the MAIN program
Computer Implementation: 15
JN Reddy - 15 Lecture Notes on NONLINEAR FEM
JN Reddy
COMPUTER IMPLEMENTATION for the nonlinear analysis (continued)
C Save the previous iteration solution and update the current oneIF(NONLN.GT.0)THEN
DO 270 I=1,NEQGPU(I)=GLU(I)IF(NONLN.EQ.1)THEN
GLU(I)=GLF(I)ELSE
GLU(I)=GLU(I)+GLF(I)ENDIF
270 CONTINUEC Test for the convergence of the solution
IF(NONLN.GT.0)THENDNORM=0.0DINORM=0.0DO 280 I=1,NEQDNORM=DNORM+GLU(I)*GLU(I)IF(NONLN.EQ.1)THEN
DINORM=DINORM+(GLU(I)-GPU(I))**2ELSE
DINORM=DINORM+GLF(I)*GLF(I)ENDIF
280 CONTINUE
ERROR=DSQRT(DINORM/DNORM)
IF(ERROR.GT.EPS)GOTO 160
DO I=1,NEQ
GLF(I)=GLU(I)
ENDDO
WRITE(ITT,1100)ITER,ERROR
ENDIF
ENDIF
C Print the solution (i.e., nodal values of the primary variables)
Logic in the MAIN program
Computer Implementation: 16
JN Reddy - 16 Lecture Notes on NONLINEAR FEM
JN Reddy
SUBROUTINE ELMATRCS2D (NPE,NN,INTF,NONLN,ITYPE)IMPLICIT REAL*8(A-H,O-Z)COMMON/STF/ELF(9),ELK(9,9),ELXY(9,2),ELU(9) COMMON/PST/A10,A1X,A1Y,A20,A2X,A2Y,A00,F0,FX,FY,
* A1U,A1UX,A1UY,A2U,A2UX,A2UYCOMMON/SHP/SF(9),GDSF(2,9)DIMENSION GAUSPT(5,5),GAUSWT(5,5),TANG(9,9)COMMON/IO/IN,ITT
CDATA GAUSPT/5*0.0D0, -0.57735027D0, 0.57735027D0, 3*0.0D0,
2 -0.77459667D0, 0.0D0, 0.77459667D0, 2*0.0D0, -0.86113631D0,3 -0.33998104D0, 0.33998104D0, 0.86113631D0, 0.0D0, -0.90617984D0,4 -0.53846931D0,0.0D0,0.53846931D0,0.90617984D0/
CDATA GAUSWT/2.0D0, 4*0.0D0, 2*1.0D0, 3*0.0D0, 0.55555555D0,
2 0.88888888D0, 0.55555555D0, 2*0.0D0, 0.34785485D0,3 2*0.65214515D0, 0.34785485D0, 0.0D0, 0.23692688D0,4 0.47862867D0, 0.56888888D0, 0.47862867D0, 0.23692688D0/
Logic in the ELMATRCS2D subroutine
COMPUTER IMPLEMENTATIONfor the nonlinear analysis (continued)
Computer Implementation: 17
JN Reddy - 17 Lecture Notes on NONLINEAR FEM
JN Reddy
COMPUTER IMPLEMENTATIONfor the nonlinear analysis (continued)
Logic in the ELMATRCS2D subroutine (continued)
CC Initialize the arraysC
DO 100 I = 1,NPEELF(I) = 0.0
DO 100 J = 1,NPEIF(ITYPE.GT.1)THENTANG(I,J)=0.0ENDIF
100 ELK(I,J)= 0.0CC Do-loops on numerical C integration begin here.C Subroutine INTERPLN2DC is called hereC
DO 200 NI = 1,IPDFDO 200 NJ = 1,IPDF
XI = GAUSPT(NI,IPDF)ETA = GAUSPT(NJ,IPDF)CALL INTERPLN2D (NPE,XI,ETA,DET,ELXY)CNST = DET*GAUSWT(NI,IPDF)*GAUSWT(NJ,IPDF)X=0.0Y=0.0U=0.0UX=0.0UY=0.0DO 140 I=1,NPE
U=U+ELU(I)*SF(I)UX=UX+ELU(I)*GDSF(1,I)UY=UY+ELU(I)*GDSF(2,I)X=X+ELXY(I,1)*SF(I)
140 Y=Y+ELXY(I,2)*SF(I)
Computer Implementation: 18
JN Reddy - 18 Lecture Notes on NONLINEAR FEM
JN Reddy
COMPUTER IMPLEMENTATIONfor the nonlinear analysis (continued)
Logic in the ELMATRCS2D subroutine (continued)C Define the coefficients of theC differential equation
FXY=F0+FX*X+FY*YA11=A10+A1X*X+A1Y*YA22=A20+A2X*X+A2Y*YIF(NONLN.GT.0)THEN
A11=A11+A1U*U+A1UX*UX* +A1UY*UY
A22=A22+A2U*U+A2UX*UX* +A2UY*UY
ENDIFC Define the element source vector C and coefficient matrix
DO 180 I=1,NPE
ELF(I)=ELF(I)+FXY*SF(I)*CNST
DO 180 I=1,NPEELF(I)=ELF(I)+FXY*SF(I)*CNSTDO 160 J=1,NPES00=SF(I)*SF(J)*CNSTS11=GDSF(1,I)*GDSF(1,J)*CNSTS22=GDSF(2,I)*GDSF(2,J)*CNSTELK(I,J)=ELK(I,J)+A11*S11+A22*S22+A00*S00
C Define the part needed to add to [K] to define [T]IF(ITYPE.GT.1)THEN
S10=GDSF(1,I)*SF(J)*CNSTS20=GDSF(2,I)*SF(J)*CNSTS12=GDSF(1,I)*GDSF(2,J)*CNSTS21=GDSF(2,I)*GDSF(1,J)*CNSTTANG(I,J)=TANG(I,J)
* +UX*(A1U*S10+A1UX*S11+A1UY*S12)* +UY*(A2U*S20+A2UX*S21+A2UY*S22)
ENDIF160 CONTINUE180 CONTINUE200 CONTINUE
Computer Implementation: 19
JN Reddy - 19 Lecture Notes on NONLINEAR FEM
JN Reddy
COMPUTER IMPLEMENTATIONfor the nonlinear analysis (continued)
Logic in the ELMATRCS2D subroutine (continued - end)
C Compute the residual vector and tangent matrix IF(ITYPE.GT.1)THEN
DO 220 I=1,NPEDO 220 J=1,NPE
220 ELF(I)=ELF(I)-ELK(I,J)*ELU(J)DO 240 I=1,NPEDO 240 J=1,NPE
240 ELK(I,J)=ELK(I,J)+TANG(I,J)ENDIF
C RETURNEND
r r r r rK U U F K U UTan[ ( )] [ ( )] eneipe ei
ij ij pe ej jp
KRK K u
u utan
1
Computer Implementation: 20
JN Reddy - 20 Lecture Notes on NONLINEAR FEM
JN Reddy 2-D Problems: 21
(1) Finite element formulation of a two-dimensional model nonlinear equation in a single variable
(2) Computation of tangent matrix coefficients(3) Numerical integration of matrix coefficients(4) Computer implementation of 2D problem
All other items (e.g., iterative methods for the solution of nonlinear equations, error check, and computer logic for nonlinear analysis) remain the same as in the 1-D problems discussed earlier.
SUMMARYJN Reddy - 21 Lecture Notes on NONLINEAR FEM