finite element analysis of - university of … · jn reddy modeleqn 1d - 1 finite element analysis...
TRANSCRIPT
ModelEqn 1D - 1JN Reddy
FINITE ELEMENT ANALYSIS OF A 1D MODEL PROBLEM
WITH A SINGLE VARIABLE
Finite element model development of a linear 1D model differential equation involving a singledependent unknown (governing equations, FE model development weak form).
ModelEqn 1D - 2JN Reddy
GOVERNING EQUATIONGOVERNING EQUATION
pointboundary aat)(
)(in)()()(
Puubdxdua
L,xfuxcdxduxa
dxd
=−+
=Ω=+⎟⎠⎞
⎜⎝⎛−
0
0
E, A
b=k L
u(L)
u0
P
x, u
a = EAk, A
L
u(L)u0
P
x, u
a = kA
Elastic deformation of a bar Heat transfer in a bar
uninsulated bar
0=−+
− fcu
dxdua
dxd
ModelEqn 1D - 3JN Reddy
FINITE ELEMENT APPROXIMATIONFINITE ELEMENT APPROXIMATION(NEED FOR SEEKING SOLUTION ON SUB(NEED FOR SEEKING SOLUTION ON SUB--INTERVALS)INTERVALS)
Approximation of the actual solutionover the entire domain requires higher-order approximation1.
Actual solution may be defined by Sub-intervals because of discontiuityof the data a(x).2.
Approximation over sub-intervals subintervals allows lower-order Approximation of the actual solution
ModelEqn 1D - 4JN Reddy
FINITE ELEMENT DISCRETIZATIONFINITE ELEMENT DISCRETIZATION
Approximation over sub-intervals subintervals allows lower-order Approximation of the actual solution
x
x xaQ bQ
axx = bxx =
A typical element (geometry and ‘forces’)
lengthelement heatsor forcesend
b =−= a
ba
xxhQ,Q
ModelEqn 1D - 5JN Reddy
bba
x
x
xb
xa
x
x
x
x
x
x
x
x
QxwQxwdxwfcwudxdu
dxdwa
dxduaxw
dxduaxwdxwfcwu
dxdu
dxdwa
dxduawdxwfcwu
dxdu
dxdwa
dxxfuxcdxduxa
dxdw
b
a
ba
b
a
b
a
b
a
b
a
⋅−−⎥⎦⎤
⎢⎣⎡ −+=
⎟⎠⎞
⎜⎝⎛⋅−⎟
⎠⎞
⎜⎝⎛−⋅−⎥⎦
⎤⎢⎣⎡ −+=
⎥⎦⎤
⎢⎣⎡ ⋅−⎥⎦
⎤⎢⎣⎡ −+=
⎥⎦
⎤⎢⎣
⎡−+⎟
⎠⎞
⎜⎝⎛−=
∫
∫
∫
∫
)()(
)()(
)()()(
a
0
WEAK FORM OVER AN ELEMENTWEAK FORM OVER AN ELEMENT
Heat/Force input
bx
Qdxdua
b
≡⎟⎠⎞
⎜⎝⎛
ax
Qdxdua
a
≡⎟⎠⎞
⎜⎝⎛−
he
21
Heat input
Force out
ModelEqn 1D - 6JN Reddy
LINEAR AND BILINEAR FORMSLINEAR AND BILINEAR FORMSAND THE VARIATIONAL PROBLEMAND THE VARIATIONAL PROBLEM
wwlu,wB allfor holds)()( =
⎥⎦⎤
⎢⎣⎡ ⋅++=⎥⎦
⎤⎢⎣⎡ += ∫∫ bba
x
x
x
xQxwQxwdxwfwl,dxcwu
dxdu
dxdwau,wB
a
b
b
a
)()()()( a
Bilinear Form and Linear Form
Weak Form
)()(
)()(
)()(
a
a
wlu,wB
QxwQxwdxwfdxcwudxdu
dxdwa
QxwQxwdxwfcwudxdu
dxdwa
bba
x
x
x
x
bba
x
x
a
b
b
a
b
a
−=
⎥⎦⎤
⎢⎣⎡ ⋅++−⎥⎦
⎤⎢⎣⎡ +=
⋅−−⎥⎦⎤
⎢⎣⎡ −+=
∫∫
∫0
Variational Problem: Find u such that
ModelEqn 1D - 7JN Reddy
EQUIVALENCE BETWEEN MINIMUM OF A EQUIVALENCE BETWEEN MINIMUM OF A QUADRATIC FUNCTIONAL AND WEAK FORMQUADRATIC FUNCTIONAL AND WEAK FORM
δuulu,δuBδI allfor )()( 00 =δ−⇒=
which is the same as the weak form or the variationalproblem with δu = w
Quadratic Functional:
Variational Problem: Find u such that I(u) is a minimum:
⎥⎦⎤
⎢⎣⎡ ⋅++−
⎥⎥⎦
⎤
⎢⎢⎣
⎡+⎟
⎠⎞
⎜⎝⎛=
≡
∫∫ bba
x
x
x
xQxuQxudxufdxcu
dxdua
ulu,uBuI
a
b
b
a
)()(
)(-)()(
a2
2
21
21
Strain energyWork done by applied forces
ModelEqn 1D - 8JN Reddy
FINITE ELEMENT MODELFINITE ELEMENT MODEL
)()()(1
xψuxUxu ej
n
j
ej
e ∑=
=≈
Finite element model
bbiaai
x
xi
ei
x
xji
jieij
eee
QxψQxψdxfψF
,dxψcψdx
dψdx
dψaK
FuK
b
a
b
a
)( )(
][
++=
⎟⎟⎠
⎞⎜⎜⎝
⎛+=
=
∫
∫
Finite element approximation (to be derived later)
ModelEqn 1D - 9JN Reddy
APPROXIMATION FUNCTIONS FORAPPROXIMATION FUNCTIONS FORLINEAR ELEMENTLINEAR ELEMENT
he
21
xccxU 21 +=)(
1uuxU aa =≡)(2uuxU bb =≡)(
xccxUxu 21 +=≈ )()(
2u
bb
aa
xccuxUxccuxU
212
211
+==+==
)()(
abab
ab
xxuuc,
xxxuxu
c−−
=−−
=→ 122
211
he
21
ab
b
xxxx
xψ−−
≡)(1ab
a
xxxx
xψ−−
≡)(2
xaxx = bxx =
2211
1221
21
uxψuxψ
xxxuu
xxxuxu
xccxUxu
abab
ab
)()(
)()(
+=−−
+−−
=
+=≈
1u
ModelEqn 1D - 10JN Reddy
ALTERNATE DERIVATION OF ALTERNATE DERIVATION OF APPROXIMATION FUNCTIONSAPPROXIMATION FUNCTIONS
(Linear Element)(Linear Element)
2u
he
21
ab
b
xxxx
xψ−−
≡)(1ab
a
xxxx
xψ−−
≡)(2
xaxx = bxx =
1u
2211 uxψuxψxU )()()( +=
⎩⎨⎧
=≠
=≡ji,ji,
δxψ ijji ifif
)(10
ba xx,xx ≡≡ 21
1-
1-
)(1)(and
)()()(1)(and
)()(
abb
a
aba
b
xxαxψ
xxαxψxxαxψ
xxαxψ
−=→=
−=−=→=
−=
22
22
11
11
(interpolation functions)
Alternate derivation usingthe interpolation property
ModelEqn 1D - 11JN Reddy
ALTERNATE DERIVATION OF ALTERNATE DERIVATION OF APPROXIMATION FUNCTIONSAPPROXIMATION FUNCTIONS
(Quadratic Element)(Quadratic Element)
he
21
xaxx = ea hxx +=
1u
332211 uxψuxψuxψxU )()()()( ++=
233
33
222
22
211
11
250
4
250
hαhxψ
xh.xxxαxψh
αhxψ
xhxxxαxψh
αxψ
xh.xxhxαxψ
a
aa
a
aa
a
aa
−=→=+
−+−=
=→=+
−+−=
=→=
−+−+=
1)(and
)()()(
1)0.5(and
)()()(
1)(and
))(()(
Alternate derivation usingthe interpolation property
ea h.xx 50+=
3
)(xψ1 )(xψ3)(xψ2
3u
1.0 1.01.0(b)
1 2 3
)(1 xeψ )(2 x
eψ )(3 xeψ
1 2 3xhe
(a)
ex3ex2ex1
x
)(xeiψ
x
ModelEqn 1D - 12JN Reddy
NUMERICAL EVALUATION OF COEFFICIENTSNUMERICAL EVALUATION OF COEFFICIENTSfor elementfor element--wise constant datawise constant data
i
x
xi
ei
x
xji
jieij
eee
QdxfψF,dxψcψdx
dψdx
dψaK
ff,cc,aa:data
b
a
b
a
constantFor
+=⎟⎟⎠
⎞⎜⎜⎝
⎛+=
===
∫∫
⎭⎬⎫
⎩⎨⎧
+⎭⎬⎫
⎩⎨⎧
=⎥⎦
⎤⎢⎣
⎡+⎥
⎦
⎤⎢⎣
⎡−
−=
2
1
11
22112
61111
QQhf
F,hc
ha
K eeeee
e
ee ][
⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧+
⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧=
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−
−+
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−−−
−=
3
2
1
141
24212162124
307818168187
3QQQ
hfF,
hcha
K eeeee
e
ee ][
Linear element:
Quadratic element:
eh
eh
1 2 3
1 2
ModelEqn 1D - 13JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE -- 11
elementeach in0=⎟⎠⎞
⎜⎝⎛−
dxduEA
dxd
lb3000
,psi1010psi,1014
psi,1030 in., 0.1
in., 0.875 in., 0.5
ft.,6ft,4 ft.,8
66
6
321
=
×=×=
×==
==
===
P
EE
Ed
dd
hhh
ab
sa
bs
Rigid member (constrained to move horizontally)
h22P
Steel (Es , As)
Aluminum (Ea , Aa)
Brass (Eb , Ab)
h3
h1
Problem: Wish to determine the deformation and stresses in the three members of the structure.
Solution: (1) Note that the governing equation is
ModelEqn 1D - 14JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 1 (continued)1 (continued)
(2) We use linear elements to represent the members of thestructure and label the elements, element nodes and globalnodes.
(3) The element equations for a typical element are
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧⎥⎦
⎤⎢⎣
⎡−
−e
e
e
e
e
ee
uu
hAE
2
1
2
1
1111
)( 00 === eeeee f,c,AEa
Rigid member (constrained to move horizontally)
h22P
Steel (Es , As)
Aluminum (Ea , Aa)
Brass (Eb , Ab)
1
2
3
1
3
4Global node numberS
1 2
21
1 2h33 3
h1
Element node numbers
2
ModelEqn 1D - 15JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 1 (continued)1 (continued)
(4) The displacement boundary conditions expressed in terms ofthe global displacements are
000 421 === U,U,U
(5) The force equilibrium conditions are obtained by looking at the equilibrium of forces on the rigid bar. We have
3)(1
1Q )(12Q1
2)(21Q )(2
2Q)(3
1Q )(32Q
)(12Q
)(22Q
)(31QP2
PQQQQQPQ 202 31
22
12
31
22
12 =++→=−−+− )()()()()()(
ModelEqn 1D - 16JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 1 (continued)1 (continued)
(6) The equilibrium conditions suggest that we must add thesecond equation of element 1, the second equation of element2, and the first equation of element 3 so that we can replacethe sum of the Q’s with 2P. The element equations in termsof the global displacements are
Using the fact that U1=U2=U4=0, we obtain
PQQQUh
AEh
AEhAE 23
122
123
3
33
2
22
1
11 =++=⎟⎟⎠
⎞⎜⎜⎝
⎛++
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
==
⎥⎦
⎤⎢⎣
⎡−
−2
2
32
22
2
22
2
1
2
1
1111
UuUu
hAE
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
==
⎥⎦
⎤⎢⎣
⎡−
−1
1
31
11
1
11
2
1
2
1
1111
UuUu
hAE
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
==
⎥⎦
⎤⎢⎣
⎡−
−3
3
43
33
3
33
2
1
2
1
1111
UuUu
hAE
0 0
0
ModelEqn 1D - 17JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 1 (continued)1 (continued)
Thus we can compute U3 , which represents the elongation inelements 1 and 2 and compression in element 3.
This completes Example 1 (one can substitute the given data to obtain numerical values of the displacements, forces, and stresses in the members).
)(eQ2
33
33323
2
22223
1
1112 U
hAE
Q,Uh
AEQ,UhAEQ −=== )()()(
(7) We can then determine the member forces using
33
3
3
323
32
2
2
222
31
1
1
121 U
hE
AQσ,U
hE
AQσ,U
hE
AQσ −=====
)()(
)()(
)()(
The member stresses are then can be computed from
ModelEqn 1D - 18JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 22
Problem: Wish to determine the numerical solution of the differential equation
0100
1022
2
==
<<−=−−
)()(
in
u,u
xxudx
ud
Solution: We have the following correspondence compared to the model equation:
dxψxf,xf,c,a i
x
x
ei
b
a∫−=−=−== 2211
(1) We wish to use a mesh of linear elements to solve theproblem. The equations of a typical element are
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
+⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
−+−−−−−
−=⎭⎬⎫
⎩⎨⎧⎟⎟⎠
⎞⎜⎜⎝
⎛⎥⎦
⎤⎢⎣
⎡−⎥
⎦
⎤⎢⎣
⎡−
−e
e
ababx
ababx
ee
ee
e QQ
xxxxxxxx
huuh
h a
b
2
14433
3
44333
2
1 12112
611111
)()()()(
41
41
ModelEqn 1D - 19JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 2 (continued)2 (continued)(2) We consider a mesh of 4 linear elements (h = 0.25).
The element equations are
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
+⎭⎬⎫
⎩⎨⎧
−=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧⎥⎦
⎤⎢⎣
⎡−
−1
1
1
1
241
2
1
2
1
003910001300
94979794
.
.uu
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
+⎭⎬⎫
⎩⎨⎧
−=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧⎥⎦
⎤⎢⎣
⎡−
−2
2
2
2
241
2
1
2
1
022320014320
94979794
.
.uu
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
+⎭⎬⎫
⎩⎨⎧
−=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧⎥⎦
⎤⎢⎣
⎡−
−3
3
3
3
241
2
1
2
1
055990042970
94979794
.
.uu
⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧
+⎭⎬⎫
⎩⎨⎧
−=⎪⎭
⎪⎬⎫
⎪⎩
⎪⎨⎧⎥⎦
⎤⎢⎣
⎡−
−4
4
4
4
241
2
1
2
1
105470087240
94979794
.
.uu
1 2 3 4 5
1 2 3 4
1 21
1 22
1 23
1 24
)(11Q )(1
2Q
)(21Q )(2
2Q
)(31Q )(3
2Q
)(41Q )(4
2Q
ModelEqn 1D - 20JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 2 (continued)2 (continued)
(3) The boundary conditions are
1 2 3 4 5
1 2 3 4
1 2
1)(1
1Q )(12Q )(2
1Q )(22Q
2 1
2
0000 41
42
31
32
31
22
21
12 =+=+=+=+ )()()()()()()()( QQ,QQ,QQ,QQ
00 51 == U,U
The equilibrium conditions are
ModelEqn 1D - 21JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 2 (continued)2 (continued)(4) The assembled equations are
⎪⎪⎪
⎭
⎪⎪⎪
⎬
⎫
⎪⎪⎪
⎩
⎪⎪⎪
⎨
⎧
+++
+
⎪⎪⎪
⎭
⎪⎪⎪
⎬
⎫
⎪⎪⎪
⎩
⎪⎪⎪
⎨
⎧
−=
⎪⎪⎪
⎭
⎪⎪⎪
⎬
⎫
⎪⎪⎪
⎩
⎪⎪⎪
⎨
⎧
⎥⎥⎥⎥⎥⎥
⎦
⎤
⎢⎢⎢⎢⎢⎢
⎣
⎡
−−−
−−−−
−
42
41
32
31
22
21
12
11
5
4
3
2
1
94970009718897000971889700097188970009794
241
QQQQQQQ
Q
UUUUU
0.1054700.1432300.0651040.0182290.001302
⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧−=
⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧
⎥⎥⎥
⎦
⎤
⎢⎢⎢
⎣
⎡
−−−
−
0.143230.065100.01823
7.83334.041704.0417 7.83334.0417
4.04177.8333
4
3
20
UUU
(5) The condensed equations for the unknown U’s and Q’s are
0
⎭⎬⎫
⎩⎨⎧
=⎭⎬⎫
⎩⎨⎧
+⎪⎭
⎪⎬
⎫
⎪⎩
⎪⎨
⎧
⎥⎦
⎤⎢⎣
⎡−−
=⎭⎬⎫
⎩⎨⎧
0.26386 0.09520
0.105470.001302
3.9167 4.0417004.0417 3.9167
4
3
2
42
11
UUU
00
ModelEqn 1D - 22JN Reddy
NUMERICAL EXAMPLE NUMERICAL EXAMPLE –– 2 (continued)2 (continued)
0.0 0.2 0.4 0.6 0.8 1.0
Coordinate, x
-0.05
-0.04
-0.03
-0.02
-0.01
0.00
Sol
uti
on, u(x)
Analytical
4L2Q4Q
FEM
Plot of the solution
0.0 0.2 0.4 0.6 0.8 1.0
Coordinate, x
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0.25
0.30
Sol
uti
on, du/dx
Analytical
4L2Q4Q
FEM
& "
"
"
"
"
*
**
*
*
&
& &
& &
& &
Plot of the derivative of the solution