Automatic Calculus

Lec 5
Principles of minimum
potential energy and Rayleigh-
Ritz-Galerkin
Prof. Gian Paolo Cimellaro
&
Prof. Giuseppe Marano
Dipartimento d’ingegneria strutturale e geotecnica,
Politecnico di Torino
Summary:
•Exact solution
•Potential energy of a system
•Elastic bar
•Principle of Minimum Potential Energy
•Rayleigh-Ritz-Galerkin Principle
A generic problem
d nx d n1x dx
cn ( x)  n  cn1 ( x)  n1  ...  c1 ( x)   f ( x)
dt dt dt

A generic 1D problem
d 2u
E  A( x) 2  b( x)
dx

Solution strategy:
1. Exact solution (Closed-form solution)
2. Approximate solutions
2.1 Variational Method (Minimum of Total Potential Energy)
2.2 Weigthed Residual Methods (Galerkin method)
EXACT SOLUTION
 Axially loaded elastic bar (truss element)
y
A(x) = cross section at x
b(x) = body force distribution
F
x (force per unit length)
x E(x) = Young’s modulus
x=L u(x) = displacement of the bar at x
x=0
Axial du Axial   Eε  E du
ε Tension T   A  EA du
strain dx stress dx dx

dT d 2u  u (0)  0
Equilibrium  b  EA 2  b  0  bc : 
dx dx  EAdu / dx( L)  F

1  x2 
u  b   (bL  F )  x 
EA  2 
 Axially loaded elastic bar (truss element)
y
A(x) = cross section at x
b(x) = body force distribution
F
x (force per unit length)
x E(x) = Young’s modulus
x=L u(x) = displacement of the bar at x
x=0

Equilibrium
du T ( x)
T(x)  R  b  x  T(x)  EA  du  dx
dx EA
Integrating

u
x
T ( x)
dx  

x
bL  F  bx 
bc : u(0)  0  
1 
 
x2
  

dx u  b (bL F ) x 
0
EA 0
EA EA  2 
 Variational Method
(Minimum of Total Potential Energy)
(Total) Potential energy
The potential energy of an elastic body is defined as

  Strain energy (U)  potential energy of loading W

dU
F

u u+du
Differential strain energy of the spring for a small change in
displacement (du) of the spring
dU  Fdu
 Strain energy of a linear spring
F
x
k k
1 F
u k u

F = Force in the spring

Hooke’s Law
u = deflection of the spring
F = ku
k = “stiffness” of the spring

For a linear spring dU  kudu

u 1
The total strain energy of the spring U 

0
k u du  k u 2
2
 Strain energy of a nonlinear spring

dU
The total strain energy of the spring
F
dU  Fdu

u u+du

u
U   F du  Area under the force  dispalcement curve
0

u u
U   F du   k (u)  u du
0 0
Potential energy of the loading (for a single spring as in the figure)

x
W  Fu
k
F Potential energy of a
k concentrated external
u loading

Potential energy of a linear spring

  Strain energy (U)  potential energy of loading W

1 2
Π  ku  Fu How to calculate u ?
2
Principle of Minimum Potential Energy
Among all admissible displacements that a body can have, the one
that minimizes the total potential energy of the body satisfies the
strong formulation

Admissible displacements: these are any reasonable displacement

that you can think of that satisfy the displacement boundary
conditions of the original problem (and of course certain minimum
continuity requirements). Example:
Any other “admissible”
displacement field w(x)

Exact solution for the

displacement field uexact(x)
x
0 L
Generically, the following formulation is given:
  (u1, u2 ,..., un )  n : DOFs

Mathematically, the minimum of potential energy can be derived as:


 0 (i  1,2,..., n)
ui

Example
k1 k2
F
x DOFs  d2 x d3 x 
d1x d 2x d 3x
 
  d  0
  ( d 2 x , d 3 x )   0 (i  1,2)   2x
ui 
   0
 d 3 x
 Principle of minimum potential energy for a system of springs
k1 k2
F
x
d1x d 2x d 3x
For this system of spring, first write down the total potential
energy of the system as:
1 1 
 system   k1 (d 2 x )  k 2 (d 3 x  d 2 x ) 2   Fd3x
2

2 2 
Obtain the equilibrium equations by minimizing the potential energy
 system
 k1d 2 x  k 2 (d 3 x  d 2 x )  0 Equation (1)
d 2 x
 system
 k 2 (d 3 x  d 2 x )  F  0 Equation (2)
d 3 x
Principle of minimum potential energy for a system of springs

In matrix form, equations 1 and 2 look like

k1  k 2  k 2  d 2 x   0 
 k     
 2 k 2  d 3 x   F 

Does this equation look familiar?

Also look at example problem worked out in class

 Axially loaded elastic bar (truss element)
y A(x) = cross section at x
b(x) = body force distribution
F (force per unit length)
x E(x) = Young’s modulus
x u(x) = displacement of the bar
x=0 x=L at x

du
Axial strain ε 
dx
du
Axial stress   Eε  E
dx 2
1 1  du 
Strain energy per unit volume of the bar dU  σε  E 
2 2  dx 
Strain energy of the bar
1 L 1
U   dU   σε dV   σε Adx since dV=Adx
2 x  0 2
Axially loaded elastic bar

Strain energy of the bar

2
L 1 1 L  du 
U σεA dx   EA  dx
0 2 2 0  dx 

Potential energy of the loading

L
W   bu dx  Fu(x  L)
0

Potential energy of the axially loaded bar

2
1 L  du  L
   EA  dx   bu dx  Fu(x  L)
2 0  dx  0
 Lets see what this means for an axially loaded elastic bar
y A(x) = cross section at x
b(x) = body force distribution
F (force per unit length)
x E(x) = Young’s modulus
x
x=0 x=L

Potential energy of the axially loaded bar corresponding to the

exact solution uexact(x)

2
1 L  du exact  L
(u exact )   EA  dx   bu exact dx  Fuexact (x  L)
2 0  dx  0
Potential energy of the axially loaded bar corresponding to the
“admissible” displacement w(x)
2
1  dw 
L L
(w)   EA  dx   bw dx  Fw(x  L)
2 0  dx  0

Any other “admissible”

displacement field w(x)

Exact solution for the

displacement field uexact(x)
x
0 L
Example:

d 2u
AE 2  b  0; 0 xL
dx
u  0 at x  0
du
EA  F at x  L
dx

Assume EA=1; b=1; L=1; F=1

Analytical solution is
x2
uexact  2x 
2
Potential energy corresponding to this analytical solution
2
1  du exact 
1 1 7
(u exact )     dx  0 u exact dx  u exact (x  1)  
2  dx 
0 6
 Now assume an admissible displacement
w x
Why is this an “admissible” displacement? This displacement is
quite arbitrary. But, it satisfies the given displacement boundary
condition w(x=0)=0. Also, its first derivate does not blow up.

Potential energy corresponding to this admissible displacement

2
1  dw 
1 1
(w)     dx  0 w dx  w(x  1)  1
2  dx 
0

Notice
7
since   1
6
 (u exact )   (w)
Principle of Minimum Potential Energy
Among all admissible displacements that a body can have, the one
that minimizes the total potential energy of the body satisfies the
strong formulation

Mathematical statement: If ‘uexact’ is the exact solution (which

satisfies the differential equation together with the boundary
conditions), and ‘w’ is an admissible displacement (that is quite
arbitrary except for the fact that it satisfies the displacement
boundary conditions and its first derivative does not blow up),
then
(u exact )  (w)
unless w=uexact (i.e. the exact solution minimizes the potential
energy)
The Principle of Minimum Potential Energy and the strong
formulation are exactly equivalent statements of the same
problem.

The exact solution (uexact) that satisfies the strong form, renders
the potential energy of the system a minimum.

So, why use the Principle of Minimum Potential Energy?

The short answer is that it is much less demanding than the strong
formulation. The long answer is, it
1. requires only the first derivative to be finite
2. incorporates the force boundary condition automatically. The
admissible displacement (which is the function that you need to
choose) needs to satisfy only the displacement boundary condition
Finite element formulation, takes as its starting point, not the
strong formulation, but the Principle of Minimum Potential
Energy.

Task is to find the function ‘w’ that minimizes the potential energy
of the system
2
1  dw 
L L
(w)   EA  dx  0 bw dx  Fw(x  L)
2 0
 dx 
From the Principle of Minimum Potential Energy, that function ‘w’
is the exact solution.
WEIGTHED RESIDUAL METHOD
(Rayleigh-Ritz-Galerkin Method)
Sometimes the variational formulation may be difficult to obtain from
differential equations:
An approximate function is chosen to approximate the independent
variable.

Approximate solution strategy:

Guess u( x)  a0j o ( x)  a1j1 ( x)  a2j 2 ( x)  ...
Where jo(x), j1(x),… are “known” functions and ao, a1, etc are
constants chosen such that the approximate solution
1. Satisfies the boundary conditions
2. Satisfies the differential equation

ji: Admissible function that generates residual

ai: Constant parameter: weight of residual
Rayleigh-Ritz-Galerkin Principle

The minimization of the potential energy is difficult to perform

exactly.
The Rayleigh-Ritz-Galerkin principle is an approximate way of
doing this.

Step 1. Assume a solution

w( x)  a0j o ( x)  a1j1 ( x)  a2j 2 ( x)  ...

Where jo(x), j1(x),… are “admissible” functions and ao, a1,

etc are constants to be determined from the solution.
 Rayleigh-Ritz-Galerkin Principle

Step 2. Plug the approximate solution into the potential energy

2
1  dw 
L L
(w)   EA  dx  0 bw dx  Fw(x  L)
2 0
 dx 

 dj 0 dj1
2
1
L 
  (a 0 , a 1 ,...)   EA  a0  a1  ... dx
2 0
 dx dx 
  b a 0j 0  a1j1  ... dx
L

 F a0j 0 ( x  L)  a1j1 ( x  L)  ...

Rayleigh-Ritz-Galerkin Principle

Step 3. Obtain the coefficients ao, a1, etc by setting

Principle of Minimum of
(w) Residuals
 0, i  0,1,2,... Searching weights that minimize
ai the residuals of total potential
energy of the system

Then, the approximate solution is

u( x)  a0j o ( x)  a1j1 ( x)  a2j 2 ( x)  ...

Where the coefficients have been obtained from step 3

 Example of application of Rayleigh-Ritz-Galerkin Principle

E=A=1
F x F=2
x=1
x=0 x=2

The potential energy of this bar (of length 2) is

2
EA  du 
2
 (u)    
2  dx 
0
dx  Fu(x  1)
Potential Energy
Strain Energy of load F applied
at x 1

Let us assume a polynomial “admissible” displacement field

u  a0  a1 x  a2 x 2
Note that this is NOT the analytical solution for this problem.
Example of application of Rayleigh Ritz Principle

For this “admissible” displacement to satisfy the displacement

boundary conditions the following conditions must be satisfied:
u(x  0)  a0  0
u(x  2)  a0  2a1  4a2  0
Hence, we obtain
a0  0
a1  2a2

Hence, the “admissible” displacement simplifies to

u  a0  a1 x  a2 x 2

 a2  2 x  x 2 
Now we apply Rayleigh Ritz principle, which says that if I plug
this approximation into the expression for the potential energy , I
can obtain the unknown (in this case a2) by minimizing 

2
1 2  du 
 (u)     dx  Fu(x  1)
2 0  dx 

     
2
1  d2 
   a2 2 x  x  dx  F a2 2 x  x 2
2

2 0  dx 
evaluated
at x 1

4 2
 a2  2a2
3
 8 3
 0  a2  2  0  a2  
a2 3 4
Hence the approximate solution to this problem, using the
Rayleigh-Ritz-Galerkin principle is

u  a0  a1 x  a2 x 2

 a2  2 x  x 2 
3
   2x  x2
4
 

Notice that the exact answer to this problem (can you prove this?) is

 x for 0  x  1
u exact 
2  x for 1  x  2
 The displacement solution :

Exact solution
1

0.8 Approximate
solution

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x

How can you improve the approximation?

The stress within the bar:

1.5
Exact Stress

0.5
Approximate
stress
Stress

-0.5

-1

-1.5

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

x