POTENTIAL ENERGY METHOD

BASED
FINITE ELEMENT FORMULATION
POTENTIAL ENERGY METHOD BASED FINITE ELEMENT FORMULATION
In the PSTP based Rayleigh Ritz method it has been observed that a single composite
trial function solution has been considered for the entire domain.
The essence of PSTP based RR method is: assuming a trial function and matching it
as closely as possible to the exact solution. This is basically a process of “curve
fitting”.
It is a well known fact that curve fitting is best done piece wise, the more the number
of pieces, the better the fit.
We can thus benefit by using piece-wise defined trial functions rather than a set of
functions valid over the entire domain of the problem.
This will lead to the finite element formulation starting from the variational principle.
1D BAR ELEMENT FORMULATED FROM STAIONARITY OF A FUNCTIONAL
let us consider a cantilever bar q(x) = q0
subjected to a distributed force “q0”
throughout the length of the bar.
Let the bar be discretised into “n” L
numbers of sub-domains/finite elements. dx
Let consider the “kth” finite element x=0
kth Sub-domain end points 1 2 3 k n

uk uk+ 1
k
Let us apply the PSTP based RRM to this finite element.
x=0 x=l
STEP-0: To find out the expression for the total potential functional of the bar finite
element. uk uk+ 1
Potential Energy stored in the bar finite
element (Πp) = Strain energy stored in the k k+1
k
bar FE – work done by external forces due x=0 dx x=l
to deformation of the bar FE l
Strain energy stored in an elemental portion of the kth bar finite element is given by:
1 1 1 du
dU k       dv   ( E )    dv   E  ( ) 2  dv
2 2 2 dx
1 du 2 1 du 2
 dU k   E  ( )  ( A  dx)  dU  AE ( ) dx
2 dx 2 dx l l
Strain energy of the entire beam may be given by: U k   dU k   1 AE ( du ) 2 dx
e

Let the work done by external forces 0 2 dx

k 0
on the bar finite element = W = Total uk uk+1
work done by the axial force (q0) in Fk Fk+1
causing a deformation u(x) + Work k
dx
k+1
l
done by the nodal reaction force Fk e
W   q0  dx  u ( x)  Fk uk  Fk 1uk 1 q0
acting at node “k” + work done by 0
the nodal reaction force Fk+1 acting at
node “k+1”.
Potential Energy stored in the bar finite element (Πp) =
l
1 du 2 l
dx
 p   AE ( ) dx  { q0  dx  u ( x)  Fk  uk  Fk 1  uk 1}
e

2 dx
 It is important to note here that, total potential energy (Πp) of the bar finite element is
a functional:
l l
1 du 2 l
 p   AE ( ) dx  { q0  dx  u ( x)  Fk  uk  Fk 1  uk 1}   p  I   F (u, u, x)
e

0 2 dx 0 0

With essential boundary conditions: u(x = 0) = uk; u(x = l) = uk+1

STEP-1: Assume a trial displacement filed (Trial solution) satisfying essential boundary
conditions which causes minimum value of total PE (as natural boundary conditions in
terms of forces and work done have already become a part of the functional itself)
Let the trial solution for the kth finite element be considered to be a piece wise
continuous trial solution defined as a linear interpolation of the nodal degrees of
freedom (uk and uk+1)
 x x
uk ( x)  [1  ]uk  [ ]uk 1
l l
The assumed trial solution satisfies the essential boundary conditions.
STEP-2: Evaluation of total potential functional with respect to the assumed
displacement field
• Substituting the trial displacement field in the potential functional we have:
l
1 duˆ 2 l

 p   AE ( ) dx  { q0  dx  uˆ ( x)  Fk  uk  Fk 1  uk 1}
e

0 2 dx 0
AE (uk 1  uk ) l
2
x x
 p    q0{(1  )uk  uk 1}dx  Fk  uk  Fk 1  uk 1
e

2 l 0 l l
 AE (uk 1  uk ) 2 x2 x2
 p   q0 [( x  )uk  ( )uk 1 ]l0  Fk  uk  Fk 1  uk 1
e

2 l 2l 2l
AE (uk 1  uk ) 2 q0l
 p   (uk 1  uk )  Fk  uk  Fk 1  uk 1
e

2 l 2
STEP-3: Set up and solve the system of equations (Application of PSTP)
• As per PSTP, for equilibrium configuration of the bar finite element, total potential
will be stationary with respect to small variations in the displacement field.
• Variation in displacement field is attained by small variations in the coefficients: uk
and uk+1 
e
AE ql
 p
0 (uk  uk 1 )  0
 Fk
uk l 2
 p
e
AE q0l
 0 (uk 1  uk )   Fk 1
uk 1 l 2
• So the element level PSTP equations for determination of nodal degrees of freedom
( uk and uk+1)/ unknown coefficients can be written in matrix from as below:
 q0 l
 [
AE
l [1
1 1
1
u k

uk 1 = 2
q0 l
2
Fk
+ Fk 1
1D HEAT TRANSFER ELEMENT FORMULATED FROM STAIONARITY OF A
FUNCTIONAL
Let us consider the “kth” sub-domain/finite element from the total “n” numbers of
finite elements considered for a 1D heat conducting structure.
STEP-0: To find out the 1 2 3 k k+1 n n+1
expression for the total 1 2 k n
potential functional of the X
heat conduction element. L
k k+ 1
Considering the X = 0 k X=L
expression for total x=0 x=l
T Tk+1 Nodal Temp.
potential functional for a k

1D structural finite element Qk Qk+1 Nodal Heat flux

we have: uk+ 1  
l
1 duˆ l

0 2 AE( dx ) dx  0 q0  dx  uˆ ( x)  Fk  uk  Fk 1  uk 1
e 2
uk p
k
With essential boundary conditions: u(x = 0) = uk;
x=0 dx x = l u(x = l) = uk+1
l Expression for total potential functional for a 1D
Tk Tk+ 1 heat transfer finite element will be of the similar form:
k 1 dTˆ 2
l l

 p   k ( ) dx   q0Tˆ ( x)dx  Qk  Tk  Qk 1  Tk 1
e

x=0 k 0 2 dx
dx x=l 0

l With essential boundary conditions: T(x = 0) = Tk;

T(x = l) = Tk+1
STEP-1: Assume a trial temperature filed (Trial solution) satisfying essential boundary
conditions which causes minimum value of total PE (as natural boundary conditions in
terms of forces and work done have already become a part of the functional itself)
Let the trial solution for the kth finite element be considered to be a piece wise
continuous trial solution defined as a linear interpolation of the nodal degrees of
freedom (Tk and Tk+1)
ˆ x x
Tk ( x)  [1  ]Tk  [ ]Tk 1
l l
The assumed trial solution satisfies the essential boundary conditions.
STEP-2: Evaluation of total potential functional with respect to the assumed
temperature field
• Substituting the trial displacement field in the potential functional we have:
1 dTˆ 2
l l

 p   k ( ) dx   q0Tˆ ( x)dx  Qk  Tk  Qk 1  Tk 1
e

0 2 dx 0

k (Tk 1  Tk ) 2 l
x x
 p    q0{(1  )Tk  Tk 1}dx  Qk  Tk  Qk 1  Tk 1
e

2 l 0 l l
k (Tk 1  Tk ) 2 x2 x2
 p   q0 [( x  )Tk  ( )Tk 1 ]l0  Qk  Tk  Qk 1  Tk 1
e

2 l 2l 2l
k (Tk 1  Tk ) q0l
2

 p   (Tk 1  Tk )  Qk  Tk  Qk 1  Tk 1
e

2 l 2
STEP-3: Set up and solve the system of equations (Application of PSTP)
• As per PSTP, for equilibrium state of the heat conduction finite element, total
potential will be stationary with respect to small variations in the temperature field.
• Variation in temperature field is attained by small variations in the coefficients: Tk
and Tk+1 
e
k ql
 p
 0  (Tk  Tk 1 )  0  Qk
Tk l 2
 p
e
k q0l
  0  (Tk 1  Tk )   Qk 1
Tk 1 l 2
• So the element level PSTP equations for determination of nodal degrees of freedom
( Tk and Tk+1)/ unknown coefficients can be written in matrix from as below:
q0 l
 [
k
l [1
1 1
1
T̂k

Tˆk 1 = q0 l
2

2
+
Qk

Qk 1