MEEN 673

Nonlinear Finite Element Analysis

Nonlinear Bending of Strait Beams
Read: Chapter 5

JN Reddy

CONTENTS
The Euler-Bernoulli beam theory
The Timoshenko beam theory
Governing Equations
Weak Forms
Finite element models
Computer Implementation:
calculation of element
matrices
Numerical examples
Beams 1

THE EULER-BERNOULLI BEAM THEORY

(development of governing equations)
q( x )
z, w

z
x

f (x )

dw

dx

Deformed Beam

JN Reddy

Undeformed Beam

dw
dx

Euler-Bernoulli
Beam Theory (EBT)
Straightness,
inextensibility, and
normality

Kinematics of Deformation in the

Euler-Bernoulli Beam Theory (EBT)

dw
dx

Displacement field

u = (u + z qx ) e 1 + w e 3 ,

u
w

dw
qx = dx

zz
z

y
x

zy

zx

yx

dw
dx

u1 ( x , z ) = u - z
xz
xy
xx

yy yz

dw
dx

u2 = 0,
u3 ( x , z ) = w( x )

Notation for stress components

JN Reddy

Von Krmn NONLINEAR STRAINS

Green-Lagrange Strain Tensor Components
1 ui u j 1 um um
Eij =
+
+ 2
2 x j xi
x i x j
u
u
u1 1 u1

+ 12 2 + 12 3
Exx =
+ 2
x1
x1
x1
x1
2

Order-of-magnitude assumption
u3
O( e )
x1

u1
O( e ),
x1

u1 1 u3

+ 2
Exx exx =
x
x

JN Reddy

Beams 4

NONLINEAR ANALYSIS OF
EULER-BERNOULLI BEAMS
z, w

Beam cross section

q(x) f(x)
z
M
M
M
y
x

N
N

cf w
L
V V
Displacements and strain-displacement relations
dw

u = (u + z qx ) e1 + w e3 , qx = dx
dw
, u2 = 0, u3 = w( x )
u1 ( x , z ) = u - z
dx
q(x)

F0

u1 1 u3
du 1 dw
d 2w
,
exx =
+ 2
+ 2
=
- z
2

dx
dx
dx
x
x
2

Nonlinear Problems (1-D) : 5

NONLINEAR ANALYSIS OF
EULER-BERNOULLI BEAMS
Equilibrium equations
dN
dM
dV
d dw
+ f = 0,
- V = 0,
+
+ q = 0
N
dx
dx
dx dx dx
dN
+ f = 0,
dx

d2 M
d dw
+
+ q = 0
N
2
dx
dx dx

Stress resultants in terms of deflection

2
2
2
du

du

d
w
dw
dw

1
1
dA = EA
dA = E
+ 2
+ 2
Ez

2
dx
dx
dx
dx
dx

2
du

d2w
d2w
dw

1
z dA = EI
z dA = E
+ 2
Ez

2
2
dx
dx
dx
dx

N = xx
A

M = xx
A

dM
d
d2w
V =
=
EI

dx
dx
dx 2

JN Reddy

Beams 6

NONLINEAR ANALYSIS OF
EULER-BERNOULLI BEAMS
Equilibrium equations in terms of displacements
(u,w)
2

d
du 1 dw

+ 2
- EA
- f = 0

dx
dx
dx

2
du

d 2 d 2w
d dw
dw
1

EI
EA
+

- q = 0

2
2
2

dx
dx
dx dx dx
dx

z,w

x,u

F
-w( L )

-u( L )

Clearly, transverse load induces both axial

displacement u and transverse displacement w.
JN Reddy

Beams 7

EULER-BERNOULLI BEAM THEORY

(continued)

Weak forms
0=

xb
xa

xb
dN

v1 - f dx =
dx
xa

dv1

N - v1 f dx - v1 ( x a )Q1 - v1 ( xb )Q4
dx

xb
xa

0=

xb
xa

xb
xa

dv1

N - v1 f dx - v1 ( x a )[-N ( x a )] - v1 ( xb ) N ( xb )
dx

2
du

dw

N = EA
+ 12

dx
dx

d2 M

d dw

v2 - q dx
N
dx 2
dx dx

d 2v d 2w dv dw

dv2
dv2
2
2

EI

Q3 - v2 ( xb )Q5 - Q6
+
- v2 q dx - v2 ( x a )Q2 - N
2
2

dx
dx
dx
dx
dx
dx

xa
xb

Q2e =-V(xa )

Q3e =-M(xa )
e
1

Q =-N(xa )

Q5e =V(xb )
2

Q6e = M(xb )
Q4e = N(xb )

he
JN Reddy

Beams 8

BEAM ELEMENT DEGREES OF

FREEDOM
Generalized displacements
2 =w(xa )

5 =w(xb )

4 =u(xb )

1 =u(xa )
3 =q(xa ) 1

he

6 =q(xb )

Generalized forces
Q2 =-V(xa )

Q5 =V(xb )
Q4 = N(xb )

Q1 =-N(xa )
Q3 =-M(xa )

JN Reddy

he

Q6 = M(xb )

FINITE ELEMENT APPROXIMATION

Primary variables (serve as the nodal variables that must be
dw
continuous across elements)
u, w, =
dx
4

w( x ) jfj ( x ),
j =1

JN Reddy

u( x ) u j y j ( x ),
j =1

Hermite cubic polynomials



2
3
x xa
x xa
e1 = 1 3
+2
he
he
2

x xa
e2 = (x xa ) 1
he
2
3


x xa
x xa
e3 = 3
2
he
he
&
%
2
x xa
x xa
e4 = (x xa )

he
he

10

HERMITE CUBIC INTERPOLATION

FUNCTIONS fi ( x )
f1 ( x )
1

slope = 0

f2 ( x )
xx

he

f3 ( x )

slope = 0

JN Reddy

slope = 0
he

f4 ( x )

he

slope = 1

xx

slope = 1

slope = 0

he

xx
11

FINITE ELEMENT MODEL

Finite Element Equations
2

Q1e

Q5e
Q3e

e2

j=1

[ K 11 ] [ K 12 ]
{ F 1 }

3e
{
u
}

1
[ K 21 ] [ K 22 ] { } { F 2 }

xb
xb
dyi dy j
dw dyi df j
11
12
1
K ij = EA
dx , K ij = 2 EA
dx ,
xa
x
dx dx
dx dx dx
a
xb
xb
dw dfi dy j
21
1
K ij = EA
dx , Fi = f yi dx + yi ( x a )Q1 + yi ( xb )Q4
xa
xa
dx dx dx
2
2
2
xb
xb
f
d

d
f
dfi df j
dw
j
22
i

K ij = EI
dx + EA
dx ,

2
2

xa
xa
dx dx dx
dx dx
xb
df i
df i
2

Fi = qyi dx + fi ( x a )Q2 + fi ( xb )Q5 + Q3 + Q

dx
dx 6
xa
xa

JN Reddy

Q6e

Q4e

u( x ) u j y j ( x ), w( x ) jf j ( x )
j=1

Q2e

xb

5e
2

e6

e4

12

MEMBRANE LOCKING
Beam on roller supports

Membrane strain

du 1 dw
e =
+ 2

dx
dx

0
xx

q (x )

du 1 dw
0
exx =
+ 2
= 0

dx
dx
2

du
1 dw
=-2

dx
dx

Remedy
dw
make
to behave like a constant

dx
2

JN Reddy

13

SOLUTION OF NONLINEAR EQUATIONS

Direct Iteration

Non-Linear Finite Element Model

[K e (De )] {De } = {F e } assembled [K (U )] {U } = {F }
Direct Iteration Method

Solution {U }r at r th iteration is known and solve for{U }r +1

[K ({U }r )]{U }r +1 = {F }
F

FC

K(U1)

K(U0)

JN Reddy

K(U)U F(U)

K(U2)
UC - Converged
solution
U0 - Initial guess
solution

U0 U1 U2 U3 UC

Nonlinear Problems: (1-D) - 14

SOLUTION OF NONLINEAR EQUATIONS

(continued)

Direct Iteration Method

Solution {U }r at r th iteration is known and solve for{U }r +1
[K ({U }r )]{U }r +1 = {F }

Convergence Criterion
NEQ

Possible convergence
U r -U r +1

e=

I =1
NEQ

r +1 2
UI

specified tolerance

I =1

JN Reddy

15

SOLUTION OF NONLINEAR EQUATIONS

Newtons Iteration Method

Taylors series
Residual,

{R} [K ({U }r )]{U }r +1 - {F }r

r

r
2

1
R
+ (U r +1 -U r )2
+
2

2!
U

R
{R(U r +1 )} = {R(U r )} + (U r +1 -U r )

U
r

{R(U r )} + (U r +1 -U r )
+ O(dU )2 ,
U

dU = U r +1 -U r

Requiring the residual {R}r +1 to be zero at the r + 1st iteration, we have

[K tan ({U }r )]{dU } = -{R}r = {F }r - [K (U r )]r {U }r
The tangent matrix at the element level is
ab tan
K ij

(
JN Reddy

Ria

=
=
Dbj
Dbj

g =1

K ipal Dgp - Fi a

p =1
n

16

SOLUTION OF NONLINEAR EQUATIONS

Newtons Iteration (continued)

Tijab

nb
2

2
n K ag

ip
ag g
a
ab
g
ab
=
=
=
+

K
F
K
T
D
D

ip
p
i
ij
p
ij
b
D
Dbj
Dbj g =1 p =1

g =1 p =1
j
[T ({D}r )]{dD} = {F }r - [K (Dr )]r {D}r , {D}r +1 = {D}r + {dD}

Ria

FC

T(2)

T(0)
T(1)

1
2

K() F R()

C - Converged
solution

0 - Initial guess
solution

0
1 = 1 + 0

C = 3
2 = 2 + 0

Nonlinear Problems: (1-D) - 17

Summary of the N-R Method

d r = {R({}(r1) )}
[T ({}(r1) ]{}

{}r = {}(r1) + {}

Computation of tangent stiffness matrix

Ri =

2 [
[

=1 p=1

Tij

Ri
j

Kip
p Fi =

n
11
[
Kip
n
[
p=1

JN Reddy

p=1

j
p=1

p=1

11
= Kij
+

1
Kip
up +

n
[

+
Kij

11
Tij11 = Kij
+

n
[

4
[

P =1

2
KiP
P Fi

4
[
 1 
Kip up +

j
P =1

 2 
KiP P

4
12
[
KiP
up +
P
uj
uj

0 up +

P =1

4
[

P =1

0 P
Beams 18

THE TIMOSHENKO BEAM THEORY

q( x )
z, w

z
x, u

f (x )

Undeformed Beam

Deformed Beams

dw
dx

dw
dx

u
JN Reddy

dw
dx

Euler-Bernoulli
Beam Theory (EBT)
Straightness,
inextensibility, and
normality

Timoshenko Beam
Theory (TBT)
Straightness and
inextensibility
19

KINEMATICS OF THE TIMOSHENKO

BEAM THEORY
Displacement field

u = (u + z fx ) e 1 + w e 3

z
u

u1 ( x , z ) = u( x ) + zf( x ),
u2 = 0,
Exx

u3 ( x , z ) = w( x )

w
z

2
u1 1 u3
exx =
+ 2

x1
x
2
dfx
du 1 dw
=
+ 2
+ z

dx
dx
dx

2Exz 2exz = g xz

JN Reddy

u1 u3
=
+
x3 x1

dw
= fx +
dx

dw
dx

x
z

Constitutive
Equations

sxx = E exx , sxz = G g xz

20

TIMOSHENKO BEAM THEORY (continued)

Equilibrium Equations
dN
dM
dV
d dw
+ f = 0,
- V = 0,
+
+ q = 0
N
dx
dx
dx
dx dx

Beam Constitutive Equations

2
2
du

du

f
d
d
w
dw

+ 12
N = sxx dA = E
+ 12
+ z x dA = EA
dx
dx
dx
dx
dx

A
A

2
du

f
d
dw

z dA = EI dfx
x
+ z
M = sxx z dA =
E
+ 12
dx
dx
dx
dx

A
A

dw
dw

V = K s sxz dA = GK s fx +
dA = GAK s fx +

dx
dx

JN Reddy

21

WEAK FORMS OF TBT

Weak Form of Eq. (1)

0=

xb
xa

v1 u

xb
dN

v1 - f dx =
dx
xa

dv1

N - v1 f dx
dx

- v1 ( x a )[-N ( x a )] - v1 ( xb )N ( xb )
=

xb
xa

dv1

N - v1 f dx - v1 ( x a )Q1 - v1 ( xb )Q4

dx

du

dv
dw

1
1

0 = EA
dx
v
f
+

2
1

xa

dx
dx dx

- v1 ( x a )Q1 - v1 ( xb )Q4
xb

JN Reddy

Beams 22

WEAK FORMS OF TBT

(continued)
Weak Form of Eq. (2)

v2 w

d dw
dw

N
GAK s fx +
v2 0=
- q dx

dx

xa
dx dx dx

xb

dv2
dv2
dw
dw

dx

f
N
GAK
v
q
=
+
+

2
s
x

xa
dx dx
dx
dx

xb

dw
dw
- v2 GAK s fx +
+ N

dx
dx

xa

xb

dv2
dv2
dw
dw

GAK s fx +
N
- v2q dx
0=
+

xa
dx dx
dx
dx

- v2 (x a ) Q2 - v2 (xb ) Q5

xb

JN Reddy

2
du

d
w

N = EA
+ 12
dx
dx

23

WEAK FORMS OF TBT

(continued)
Weak Form of Eq. (3)

v3 fx

d df

dw
x

v3 0=
+ GAK s fx +
dx
EI

xa
dx
dx
dx
xb

x b dv

3 EI dfx + GAK sv3 fx + dw dx - v3 EI dfx

=

dx

xa

dx
dx
dx x

a
x b dv

d
f
dw
x
3

0=
+ GAK sv3 fx +
dx
EI

xa
dx
dx
dx
- v3 (x a ) Q3 - v3 (xb ) Q6

xb

JN Reddy

24

FINITE ELEMENT MODELS OF

TIMOSHENKO BEAMS
Finite Element Approximation

u j y(1)
j (x ), w

j=1

w1

w1
1
JN Reddy

w j y(2)
j (x ), f

j=1

K 11

K 21

K 31

K 12

K 22

K 32

F 1
K 13

{ }

{
}

{w}
= F 2
K 23
{ }

{S }

3
K 33

{F }

s1

he

w2
2

he

w3

S j y(3)
j (x )

j=1

w2
2

m=n=2

s1
1

m=n=3

s2
he

s2
2

he

s3
3
25

SHEAR LOCKING IN TIMOSHENKO BEAMS

dw
(1) Thick beam experiences shear deformation, fx dx
dw
(2) Shear deformation is negligible in thin beams, fx = dx
Linear interpolation of both w, fx
w( x ) w1y1 ( x ) + w2 y2 ( x ),

fx ( x ) S1y1 ( x ) + S2 y2 ( x )

w2

w1

S2

S1

1
he

he

In the thin beam limit it is not possible for the element to realize
the requirement
dw

fx = -

JN Reddy

dx

26

SHEAR LOCKING - REMEDY

In the thin beam limit, should become constant so that it
matches dw/dx. However, if is a constant then the bending
energy becomes zero. If we can mimic the two states (constant
and linear) in the formulation, we can overcome the problem.
Numerical integration of the coefficients allows us to evaluate
both and d/dx as constants. The terms highlighted should
be evaluated using reduced integration.

K ij22 =
K ij23
K i33
j

JN Reddy

=
=

xb
xa
xb
xa
xb

xa

(2)

dyi(2) dy j

+ ... dx
GAK s
dx dx

dyi(2) (3)
y j dx = K 32
GAK s
ji
dx
dy(3) dy(3)

j
(3)
(3
)
i
EI
+ GAK s yi y j dx

dx dx

27

GENERAL LOGIC IN A COMPUTER PROGRAM

for the nonlinear analysis
NL=1,NLS

Logic in the
MAIN program

Iter = 0
F=F+F

Iter = Iter + 1

NLS = no. of load steps

Initialize global Kij, fi
DO 1 to N
Transfer global information
(material properties, geometry and solution)
to element
CALL ELKF to calculate Kij(N)
and fi(n), and assemble to form
global Kij and Fi
Impose boundary conditions
and solve the equations
Error <
no

Yes
JN Reddy

yes

Iter < Itmax

Print Solution

No Write a

message

STOP

Beams 28

CALCULATION OF BEAM PARAMETERS

AND INITIALIZATIONS
IF(MODEL.GE.2)THEN

C
C Define the beam stiffness coefficients, EA, EI, GAKs, from the
C geometric and material parameters read in the main program
C (should be passed to this subroutine)
C
C Initialize arrays
C
DO 20 I=1,NPE
ELF1(I)=0.0
ELF2(I)=0.0
ELF3(I)=0.0
DO 20 J=1,NPE
ELK11(I,J)=0.0
ELK12(I,J)=0.0
ELK13(I,J)=0.0
ELK21(I,J)=0.0
ELK22(I,J)=0.0
ELK23(I,J)=0.0
ELK31(I,J)=0.0
ELK32(I,J)=0.0
ELK33(I,J)=0.0
Beams 29

CALCULATION OF BEAM PARAMETERS

AND INITIALIZATIONS

C
C
C

C
C

IF(NONLIN.GT.1)THEN
TAN12(I,J)=0.0
TAN13(I,J)=0.0
TAN22(I,J)=0.0
TAN23(I,J)=0.0
TAN32(I,J)=0.0
TAN33(I,J)=0.0
ENDIF
20
CONTINUE
ENDIF
Full integration of the coefficients
DO 100 NI=1,NGP
XI=GAUSPT(NI,NGP)
CALL INTERPLN1D(ELX,GJ,IEL,MODEL,NPE,XI)
X=ELX(1)+0.5*(1.0+XI)*EL
CNST=GJ*GAUSWT(NI,NGP)
DEFINE AXX, BXX, CXX, DXX, FX, and so on as needed to define
the element force and stiffness coefficients
Beams 30

CALCULATION OF ELEMENT MATRICES

C
C

The EULER-BERNOULLI beam element (MODEL=2) - LINEAR

MODEL = Type of physical problem
=1, 2nd order eqn. in 1 variable
=2, EBT
>2, TBT

IF(MODEL.EQ.2)THEN
fi
yi
DO 50 I=1,NPE
I0=2*I-1
ELF1(I)=ELF1(I)+F0*FX*SFL(I)*CNST
ELF2(I)=ELF2(I)+F0*QX*SFH(I0)*CNST
ELF3(I)=ELF3(I)+F0*QX*SFH(I0+1)*CNST
dyi
DO 50 J=1,NPE
d 2fi
J0=2*J-1
2
dx
dx
S11=GDSFL(I)*GDSFL(J)*CNST
H22=GDDSFH(I0)*GDDSFH(J0)*CNST
H23=GDDSFH(I0)*GDDSFH(J0+1)*CNST
H32=GDDSFH(I0+1)*GDDSFH(J0)*CNST
H33=GDDSFH(I0+1)*GDDSFH(J0+1)*CNST
ELK11(I,J)=ELK11(I,J)+AXX*S11
K ij11 =
ELK22(I,J)=ELK22(I,J)+DXX*H22
ELK23(I,J)=ELK23(I,J)+DXX*H23
K ij22 =
ELK32(I,J)=ELK32(I,J)+DXX*H32
ELK33(I,J)=ELK33(I,J)+DXX*H33
50
CONTINUE
ENDIF

xb

Fi = f (x ) yi dx
1

xa

xb

Fi = q (x ) fi dx
2

xa

d yi d yj
xa Axx dx dx dx ,
2
xb
d 2fi d fj
xa Dxx dx 2 dx 2 dx
xb

Beams 31

CALCULATION OF ELEMENT MATRICES

C
C
C

60

The TIMOSHENKO beam element (MODEL=3) - LINEAR

IF(MODEL.GT.2)THEN
DO 60 I=1,NPE
ELF1(I)=ELF1(I)+F0*FX*SFL(I)*CNST
ELF2(I)=ELF2(I)+F0*QX*SFL(I)*CNST
DO 60 J=1,NPE
S11=GDSFL(I)*GDSFL(J)*CNST
ELK11(I,J)=ELK11(I,J)+AXX*S11
ELK33(I,J)=ELK33(I,J)+DXX*S11
CONTINUE
ENDIF

100 CONTINUE ! (loop on NI =1, NGP ends here)

C
C
C
C

Define shear and nonlinear coefficients for the two beam theories as
appropriate in the reduced integration do-loop; define ELK and TAN coefficients

Beams 32

REARRANGE ELEMENT COEFFICIENTS

IF(MODEL.GT.1)THEN
II=1
DO 220 I=1,NPE
ELF(II) =ELF1(I)
ELF(II+1)=ELF2(I)
ELF(II+2)=ELF3(I)
JJ=1
DO 210 J=1,NPE
ELK(II,JJ)
= ELK11(I,J)
ELK(II,JJ+1) = ELK12(I,J)
ELK(II,JJ+2) = ELK13(I,J)
ELK(II+1,JJ) = ELK21(I,J)
ELK(II+2,JJ) = ELK31(I,J)
ELK(II+1,JJ+1) = ELK22(I,J)
ELK(II+1,JJ+2) = ELK23(I,J)
ELK(II+2,JJ+1) = ELK32(I,J)
ELK(II+2,JJ+2) = ELK33(I,J)
210
JJ=NDF*J+1
220 II=NDF*I+1
ENDIF
Beams 33

COMPUTATION OF

RESIDUAL VECTOR AND TANGENT MATRIX

C
C

Compute the residual vector and tangent coefficient matrix for

the Newton iteration method (only)

230

250
260
ENDIF

IF(NONLIN.GT.1)THEN
DO 230 I=1,NET
DO 230 J=1,NET
ELF(I)=ELF(I)-ELK(I,J)*ELU(J)
II=1
DO 260 I=1,NPE
JJ=1
DO 250 J=1,NPE
ELK(II,JJ+1)
= ELK(II,JJ+1) +TAN12(I,J)
ELK(II+1,JJ+1) = ELK(II+1,JJ+1)+TAN22(I,J)
IF(MODEL.EQ.2)THEN
ELK(II,JJ+2)
= ELK(II,JJ+2) +TAN13(I,J)
ELK(II+1,JJ+2) = ELK(II+1,JJ+2)+TAN23(I,J)
ELK(II+2,JJ+1) = ELK(II+2,JJ+1)+TAN32(I,J)
ELK(II+2,JJ+2) = ELK(II+2,JJ+2)+TAN33(I,J)
ENDIF
JJ=NDF*J+1
II=NDF*I+1
Beams 34

NUMERICAL EXAMPLES
Pinned-pinned beam (EBT)
1.20
1.10

Deflection,

w0

1.00

q0

Pinned-pinned

0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20

Clamped-clamped

q0

0.10
0.00
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0

Load,

q0

Nonlinear Problems: (1-D) - 35

Pinned-pinned beam (TBT)

w (in.)

L = Length,
H = Height of the beam
L
= 100
H

q0

L
= 50
H
L
= 10
H

q(psi.)
JN Reddy

Nonlinear Problems: (1-D) - 36

Pinned-pinned beam (EBT, TBT)

1.0
L / H = 10(TBT)
L / H = 10(EBT)

Deflection w

0.8

L / H = 50
(TBT,EBT)

q0

0.6

L / H = 80
(TBT,EBT)
L / H = 100
(TBT,EBT)

0.4
0.2
w = w(0.5L)

0.0
0

EH
L4

H = beam height
L = beam length

10

Load (lb/in), q0
JN Reddy

Beams 37

Hinged-Hinged beam (EBT and TBT)

EBT = EulerBernoulli beam theory
TBT = Timoshenko beam theory
TBT

EBT

L/H=10

w = wEH 3/qL4

Nondimensional deflection

0.3

TBT

0.2

L/H=100
q0

0.1

Load, q0 (psi)

0.0
JN Reddy

10

Nonlinear Problems: (1-D) - 38

SUMMARY
In this lecture we have covered the following
topics:
Derived the governing equations of the
Euler-Bernoulli beam theory
Derived the governing equations of the
Timoshenko beam theory
Developed Weak forms of EBT and TBT
Developed Finite element models of EBT
and TBT
Discussed membrane locking (due to the
geometric nonlinearity)
Discussed shear locking in Timoshenko beam
finite element
Discussed examples
JN Reddy

39