4 FEM - Lect2

Download as pdf or txt
Download as pdf or txt
You are on page 1of 113
At a glance
Powered by AI
The document discusses the theoretical basis and mathematical underpinnings of the finite element method. It can be understood using structural matrix method, variational method and method of weighted residuals. The steps involve discretization, creating a finite subspace, solving governing equations and obtaining numerical solutions at nodes.

The three approaches to understand the finite element method are: structural matrix method, variational method and method of weighted residuals.

The steps involved in solving problems using the finite element method are: 1) Adopt discretization and flexibility of variation method 2) Transform variation problems into extremum problems of multivariable functions 3) Solve governing equations to obtain numerical solutions at nodes.

Chapter 2

Theoretical Basis of
Finite Element
Contents

1 Introduction

2 The direct method of variation problem

3 The method of weighted residuals

4 The method of finite element

5 Conclusion
I. Introduction
Introduction
 The finite element method is a numerical approximation method of
mathematical physics problems by computer analysis.
It originated from solid mechanics and rapidly developed to other physical
fields such like hydromechanics, heat transfer theory, electromagnetics and
acoustics.

 The fundamental idea of finite element method is to divide the continuous


body into many finite elements and than reassemble them, which means the
analysis of the original continuous body can be transformed into the analysis
of the newly assembled object of these finite elements.

One can understand the finite element method in three approaches:


 Structural matrix method
 Variational method
 Method of weighted residuals

 The whole calculation process of finite element method is conducted


automatically by a computer program. It has very strong generality. All you
need to do is to change the inputs according to the different practical
engineering problems.
This method totally changed the limitation of analytical solution.
Introduction
 The main task of engineering design is to analyze the structural strength, stiffness and
stability using the theory of solid mechanics including structural mechanics, elastic
mechanics and plastic mechanics.
As mentioned before, with the development of science and technology the geometrical
shapes and loading conditions of the engineering structures are becoming increasingly
complicated. This has been exacerbated by the emergency of new materials which
makes it difficult or even impossibile for solving the analytical solutions of structural
analysis.

 So people turned to seek its approximate solutions.


In 1908 Ritz proposed an approximate method using trial functions with unknown
quantities to approximate the energy functional that is to obtain the equations of
unknown quantities by solving their minimum of energy.
The method of Ritz largely promoted the application of elastic mechanics in
engineering. But its limitation is that the trial function must satisfy the boundary
conditions, which can be quite difficult for a structure with complicated geometrical
shape.
Courant extended the method of Ritz in 1943. He divided the entire cross section into
several triangle regions when solving the torsion problem and assumed the
approximate uniform distribution of the warping function in each triangle region to
overcome the difficulty of the Ritz method that demands the entire approximate
functions satisfy all the boundary conditions.
It is quite similar to the idea of finite element method in the early stage. But it
demanded too much numerical calculation for further development.
Introduction
 The line structural system first adopted the computer for numerical calculation since its birth
in 1946.
Its theoretical basis is the matrix displacement method and matrix force method, which can be
called the method of structural matrix, derived from the displacement method and force
method of structural mechanics. It adopts the matrix algebra for simpler expression of
equations and facilitates the computational program.
The mechanics concept of structural matrix is pretty clear and all the theoretical formulas
have very accurate meaning of structural mechanics. The rounding errors occur only in the
process of numerical calculation due to the computer limitation of the storage digits of real
numbers.
 Turner, Clough, Martin and Topp introduced a new calculation method to solve the plane
stress problem evolved from the matrix displacement method in the annual conference of
Aviation Institute in 1956. They divided the structure into several triangle and rectangle
elements and solve the element stiffness matrix between the force and displacement of the
nodes using the approximate displacement function of the element.
 At the same time, Argyris published several papers of energy theory and structure analysis.
He unified the fundamental energy principles of elastic structures, developed the matrix
method and derived the element stiffness matrix of the rectangular grating composed of a
plane stress plate and four edge members.
 Clough for the first time introduced the term “FEM” in the paper “The finite element
method in plane stress analysis” in 1960 since when the finite element method has become an
standard method to discretize the continuous body.
 However in the early stage people only relied on the intuitive concepts without sufficient
theoretical basis. So the application was sometimes not successful due to the limitation of the
practical problems it could solve.
Introduction
 People gradually realized that the finite element method was actually a transformatin
of the Ritz method of variation principle after the work of Besseling, Melosh, Jones
and Gallagher in 1963 and then developed the methods to derive the calculation
formulas of finite element from different variation principles.
 In 1965,Zienkiewicz and Y.K.Cheung discovered that all the field problems can be
solved by the same procedures as the finite element method of solid mechanics if they
can be transformed into the form of variation when solving the Laplace and Poisson
equations and then published the first book of finite element in 1967.
 However the formulae of finite element method are not necessarily based on the
variation principle. In 1969 Szabo and G.C.Lee noticed that one can solve the non-
structural problems in standard procedures of finite element by the weighted residual
method especially the Galerkin one.
 At the same time the mathematical basis of the finite element method attracted great
attentions: the numerical method to solve large-scale linear equations and the method
to solve the matrix eigenvalues were developed, techniques of substructures and
modal synthesis were applied and the convergence of finite element method as well as
the error analysis was being studied.
 The functions of computer is becoming increasingly stronger and the finite element
method is gaining greater attentions. Many universities, research institutes and
software companies have received a lot of financial aids from various industrial
sectors (eg. Aviation, astronavigation, civil engineering, shipbuilding and automobiles)
and develop many general programs of finite element, which has contribute to the
theoretical research and practical applications.
Introduction
 Hundreds of general programs of finite element for large-scale
structures have been developed since the early 1980s, of which the
most well known ones are:
 NASTRAN
 MARC
 SAP-NONSAP The list is in no
 ADINA particular order.
 ANSYS
 ABAQUS
 ……

 Chinese researchers of mechanics also made great contributions to


the early development of finite element method:
 陈伯屏:method of structural matrix
 钱令希:theory of the complementary potential energy
 钱伟长、胡海昌:theory of generalized variation
 冯康:theory of finite element method The list is in no
 …… particular order.

 The concept will be very clear if derived by using the variation


principle in the field of solid mechanics. So we’ll first explain the
methods in a mathematical way and then introduce their
applications in mechanic field.
II. The direct method
of variation problem
2.1 Brief introduction

 We have discussed the extremum problems of functional in detail and


established the variation method of mechanical problems. However we all
stop at the Euler equations of functional.

 We have to solve the Euler equations when dealing with the practical
problems but unfortunately it’s usually very difficult to get the analytical
solutions.
So here comes the direct method of variation problems. People usually
transform the differential problem into a variation one due to its convenience.

 To mechanics problems, establishing the corresponding functional and solve


its extremum problem is only one way to solve it. The following example
shows another way of establishing the proper functional.
2.1 Brief introductions
 From the differential equations to the variation equations
One can use the trial-and-error method to establish the corresponding
functional to solve the normal differential equations. Refer to the relevant
mathematical manuals for the functional forms for the common problems.

For the boundary problems of the following ordinary differential equation (L


is the operator)
 d  du 
 Lu    P ( x )   r ( x)u ( x)  f ( x)
 dx  dx 
 u (a)  u (b)  0

And P( x)  C1[a, b], r ( x)  C[a, b]


P( x)  0, r ( x)  0, r ( x)不恒为零
It equals to the extremum problem of the functional
b
Q[u ( x)]   [( P( x)u( x)) u ( x)  r ( x)u 2 ( x)  2 f ( x)u ( x)]dx
a
Now let’s prove the validity of this equivalency.
2.1 Brief introductions
Prove:
1)If u0(x) is the solution to the differential equation then it must make the
functional Q[u(x)] get its minimum.
That is for any y(x), we have Q[y(x)]> Q[u0 (x)]
Denote the set of all the functions with continuous second-order derivative in
[a,b] and two endpoints of zero as

C02 [a, b]  { f ( x) f ( x)  C 2 [a, b], f (a )  f (b)  0}


For an arbitrary W(x) inside we have

So
V  u0  W  C02 [a, b]
b
Q[V ]   [( PV )V  rV 2  2 fV ]dx
a

  PV V a   [ P V   ]dx   [rV 2  2 fV ]dx


b b 2 b

a a
Because V(a) = V(b) =0, the equation can be written as

Q[V ]   [ P V    rV 2  2 fV ]dx
b 2
a
Consider V(x) = u0(x) + W(x)
2.1 Brief introductions
Prove:We have

Q[V ( x)]  Q[u0 ( x)  W ( x)]   [ P  u0  W    r  u0  W   2 f  u0  W ]dx


b 2 2
a
b b b b
  [ Pu  ru  2 fu0 ]dx  2 [ Pu0 W   ru0W ]dx  2 fWdx   [ PW 2  rW 2 ]dx
0
2 2
0
a a a a
According to the formula of integration by parts, the second term can be written as


 Pu0  Wdx   ru0Wdx
b b b
 [ Pu W   ru W ]dx  Pu W  
b

a 0 0 0 a a a

  [( Pu )  ru ]Wdx   fWdx


b b

a 0 0 a
b b
So Q[V ( x)]   [ Pu02  ru02  2 fu0 ]dx   [ PW 2  rW 2 ]dx
a a
b
 Q[u0 ( x)]   [ PW 2  rW 2 ]dx  Q[u0  W ]
a
The equation above holds apparently for any
W ( x)  C02 [a, b]
2.1 Brief introductions
Prove:
Notice that P(x) and r(x) are not negative. So when W(x) is not identically
equivalent to zero, we have
b
a
[ P( x)W 2 ( x)  r ( x)W 2 ( x)]dx  0
which can be proved by the following proof by contradiction.
If the integral above equals to zero, the integrand equals to zero. That is
P( x)W 2 ( x)  r ( x)W 2 ( x)  0  P( x)W 2 ( x)  0  W 2 ( x)  0
W(x) is a constant, so
W ( x)  0  W ( x)  C
Due to the continuity of the function we have
W (a)  W (b)  0  W ( x)  0
which means once V(x) doesn’t identically equal to u0(x) we have
b
Q[V ( x)]  Q[u0 ( x)]   [ PW 2  rW 2 ]dx  Q[u0 ( x)]
a

So the functional Q[u(x)] gets its minimum at the solution u0(x) of the differential
equation.
2.1 Brief introductions
Prove:
2)If the functional Q[u(x)] reaches its minimum at u0(x) then u0(x) is the
solution of the differential equations.
If the arbitrary W ( x)  C02 [a, b] doesn’t identically equivalent to zero, then for any
real number λ we have
Q[u0 ( x)  W ( x)]  Q[u0 ( x)]   [ P  W    r  W  ]dx
b 2 2
a

2 [ Pu0  W    ru0  W   f (W )]dx  Q[u0 ( x)]


b

When W(x) and u0(x) are fixed, the red part of the equation can be regarded as
the function φ(λ), so
b b
 ( )    [ PW 2  rW 2 ]dx  2  [ Pu0W   ru0W  fW ]dx  0
2
a a

Consider the conclusion of the quadratic trinomial


A 2  B  C  0    B 2  4 AC  0
C=0, so we have B=0. That is
Integrate by parts
b
.
 [ PuW   ru W  fW ]dx  0
a 0 0
2.1 Brief introductions
Prove:
According to the formula of integration by parts we have

 [  Pu   ru
b
0 0  f ]Wdx  0
a
which holds for any W ( x)  C02 [a, b]
Similar to the preliminary theory of variation we have
  Pu0   ru0  f  0
where u0(x) is the solution of the differential equation.

Conclusion:
The necessary and sufficient condition of u0(x) being the solution of the
differential equation is that the functional Q[u(x)] gets its minimum at u0(x),
which means the extremum problem of the functional and the problem of
differential equations are equivalent.

Now let’s discuss the direct method of variation problem.


2.1 Brief introductions
 The direct method of variation problem
Euler adopted the “finite difference method” when first studying the variation
problem, which is a very important direct method. So Euler is the founder of
the direct method of variation problems.
Now let’s first discuss the finite difference method of Euler for further
understanding of other approximate method.
For the differential equations written in the form of operator
Ty( x)  f ( x)
Suppose its corresponding variation method is: solve the function that can
make the following functional get its extremum.
b
Q[ y ( x)]   F ( x, y, y)dx
a

The boundary condition is


y(a)  ya , y(b)  yb
The main procedures of the Euler finite difference method is:
2.1 Brief introductions
The direct method of variation problem
(1)Divide the interval [a,b] into n+1 equal portions and the diversion points are
x0  a, x1 , , xn , xn1  b
Let y=f(x) be the polyline through the points (a,ya), …,(xi,yi) ,…, (b,yb) and yi is
undetermined except the boundary.
(2)Replace the derivative by the difference quotient in the interval (xi,xi+1)
yi 1  yi
y 
x
(3)So the functional Q[y(x)] is actually the function of y1, …,,yn. That is
n
yi 1  yi
F ( x, y, y)dx   F ( xi , yi ,
b
Q[ y ( x)]   )xi   ( y1 , , yn )
a
i 0 x
(4)Select the value of y1, …,,yn to make the function Φ get its extremum.
According to the extremum of the multivariate function we have
  
 0,  0, , 0
y1 y2 yn
So y1, …,,yn can be determined by the equations above.
2.1 Brief introductions
The direct method of variation problem
For any value of interval we can get an approximate curve.
When the number of section increase (n→∞) we can obtain a series of curves or
functions:
 f1 ( x), f 2 ( x), , f m ( x), 
The more the sections are, the closer the curve can approach to the real function.
The real solution makes the functional get its minimum so the functional decreases with
the increase of n. That’s why it’s called the minimizing sequence of the functional.

Definition:The sequences that satisfy the following conditions can be called the
minimizing sequences of the functional.
Q[un ]  min Q[u] n  
Theorem:If the differential operator T is positively definite, then any minimal
sequence of the corresponding functional converges to the solution of its Euler equation.
b
Tu  f  Q[ y ( x)]   F ( x, y, y)dx
a
2.2 The method of Ritz
 Review
From the discussion above we can confirm that the differential problems can
be transformed into the variational ones.
Now let’s explain it in a mathematical way (but not precise).

Lu  f
For a differential equation

L is the differential operator of the function space H. u is a undetermined


function and f is given.

L is defined as
d
Lu    P( x)u ( x)   r ( x)u ( x)
dx
The operator L is linear so
L(c1u1  c2u2 )  c1L(u1 )  c2L(u2 )
Please prove it on your own.
2.2 The method of Ritz

 Review
To simplify the writing we denote the integral as
b
 a
f ( x) g ( x)dx  ( f , g )
and call it the inner product of the function f and g.
The corresponding functional is
b b b
Q[u ( x)]   [( Pu)u  ru  2 fu ]dx   u[( Pu)  ru ]dx  2  fudx
2
a a a
which can be written as
Q[u ( x)]  (Lu, u)  2( f , u )
It’s easy to prove that the inner product defined here has the following
properties ( f ,c g  c g )  c ( f , g )  c ( f , g )
1 1 2 2 1 1 2 2

(c1 f1  c2 f 2 , g )  c1 ( f1 , g )  c2 ( f 2 , g )
Now we’ll discuss the Ritz method of variational problems. That is how
to establish the minimizing sequence by the method of Ritz.
Apparently L is only a special case. So we’ll change it to T for practical
problems.
2.2 The method of Ritz
(1)Transform the problem of differential equations into the variational one
by the variational principle

Tu  f  Q[u ]  (Tu, u )  2( f , u)  min


(2)Transform the variation problem of infinite dimensional function space into the
approximate extremum problem of the multivariable function after establishing the
minimizing sequence. That is to choose the linear independent functions {φ1,φ2, …, φn}
as the basis functions in H to form a nth dimensional subspace

Hn =span{1 ,2 , ,n }


For any arbitrary element inside we have n

when n=1,2,3, …,which means the subspace un is expanding,


i 1

aii un approaches to the
minimizing sequence of the functional.
(3)Substitute un into the expression of the functional

n n n
Q[u ]  Q[un ]  (Tun , un )  2( f , un )  (T aii ,  aii )  2( f ,  aii )
i 1 i 1 i 1
n n
  a a ( T ,  )  2  a ( f ,  )  F ( a , a , , a )
i j i j i i 1
And the functional has become the multivariable function of a ,a , …, a
i , j 1 i 1
2 n
after the
1 2 n
calculation of inner product.
2.2 The method of Ritz
(4)According to the extremum theory of multivariable function we have
the governing equations as follows.
F F
 0, , 0
a1 an
expressed in the form of matrix as
 (T1 , 1 ) (T1 , 2 ) (T1 , n )   a1   ( f , 1 ) 
 ( T ,  ) ( T  ,  ) ( T  ,  )   a  ( f ,  ) 
 2 1 2 2 2 n  2
 2 

    
    
 ( T  ,
n 1 ) ( Tn , 2 ) ( T n , n  n
) a  ( f , n 
)
(5)Solve the equations and get a1,a2, …, an
n
un   aii
i 1
So we have the approximate extremum function of the functional, which is
the approximate solution of a differential equation.
Now let’s look at some examples.
2.2 The method of Ritz

Examples
Solve the function u(x)∈C2[0,1] that satisfy the following equations
u( x)  u ( x)  x

 u (0)  u (1)  0
We can easily solve the exact solution
e
u ( x )  2 (e x  e  x )  x
e 1
Now let’s solve the problem by the method of Ritz.
The equation here is actually the special case of the differential operator L
mentioned before. That is P(x) ≡r(x)≡1,f(x) ≡-x
d d
Lu    P ( x )u ( x )   r ( x )u ( x )   (u )  u  f   x
dx dx
So the corresponding functional is
Q[u ( x)]  (Lu, u)  2( f , u)
The solution space here is C2[0,1].
2.2 The method of Ritz

Example
Let the basis functions be 1 ( x)  x(1  x)
2 ( x)  x 2 (1  x)
3 ( x)  x3 (1  x)
They apparently satisfy
i (0)  0  i (1)  i ( x)  C02 [0,1]
It can be proved that the three functions above are linearly independent.
Substituted into the expression above we can get the following equations.

 0.3667 a1  0.1833a2  0.1763a3  1/12 Attention to


the symmetry

 0.1833a1  0.1333a2  0.8940a3  1/ 20


of the
coefficients

0.1763a  0.8940a  0.0897 a  1/ 30


 1 2 3

Thus a1  0.4425, a2  0.4635, a3  0.0343


.
2.2 The method of Ritz
Example
So the approximate solution is

u3 ( x)  0.4425 x(1  x)  0.4635 x 2 (1  x)  0.0343 x 3 (1  x)


 0.0343x 4  0.4292 x3  0.9060 x 2  0.4425 x
At the point x=0.5 calculate both the exact value and the approximate value,
e
u (0.5)  (e0.5  e0.5 )  0.5  0.0566
e 1
2

u3 (0.5)  0.0343(0.5) 4  0.4292(0.5)3  0.9060(0.5) 2  0.4425(0.5)  0.0505


The error is apparent but it should be noticed that the number of the basis
function is only 3.

 Exercise
Solve the extremum function of the functional
1
Q[ y ]   ( y2  4 y 2 )dx y(0)  0, y(1)  1
0
Solve the problem respectively by the Euler equation and the method of Ritz
(take 3 terms)
2.2 The method of Ritz
 Further discussions
There are still some problems unsolved even one can solve the
variation problem by the method of Ritz. For example whether
the linear equations solved by the Ritz method have solutions
and whether the solution is unique.
Now let’s discuss the problem in detail.
The Ritz method has approximated the functional into the
multivariable function
n n
Q[u ]  F (a1 , a2 , , an )   a a (T ,  )  2 a ( f ,  )
i , j 1
i j i j
i 1
i i

According to the extremum condition we have


F n
 0   a j (Ti ,  j )  ( f , i ) i  1, , n
ai j 1

which are the linear equations of a1,a2, …, an.


We only discuss the differential operator L, whose definition
is
Lu  ( Pu)u  ru
2.2 The method of Ritz
 Further discussion
So the left part of the equationb
(Li ,  j )   [( Pi) j  ri j ]dx
a φj(a)= φj (b)=0
b
  Pi j   [ Pi j  ri j ]dx
b

written as a a
b
(Li ,  j )  kij  B(i ,  j )   [ Pi j  ri j ]dx
a
let ( f , i )  bi
Then the equations can be written as
k11a1  k12 a2   k1n an  b1
k a  k a   k a  b
 21 1 2 2 2n n 2


kn1a1  kn 2 a2   knn an  bn
or expressed with matrix notations
Ka  b
2.2 The method of Ritz
 Further discussion
The coefficient symmetry of the equations mentioned before can be obtained
directly by its calculation formula
b
kij  B(i ,  j )   [ Pi j  ri j ]dx  B( j , i )  k ji
a
which means the coefficient matrix K of the equations is symmetric.
Now let’s prove K is positively definite
cT  (c1 , c2 , , cn )  0 : cT Kc  0
In fact n n

cc k   [ Pc c     rc c   ]dx
b
c Kc 
T
i j ij  i j i j i j i j
a
i , j 1 i , j 1

 n   n   n  n 
  [ P   cii    c j j   r   cii    c j j ]dx
b

a
 i 1   j 1   i 1   j 1 
n n
 B( cii ,  c j j )  B(un ( x), un ( x))
n i 1 n j 1
where un ( x)   cii   c j j
i 1 j 1
2.2 The method of Ritz
Further discussion
n
un ( x)   cii
Now that
i 1
When the vector c≠0, then un(x) doesn’t identically equal to zero. Because the
functions φ1,φ2, …, φn are linearly independent.
Now we only have to prove that there exists δ>0 that will make

c Kc  B(un ( x), un ( x))   [ P  un   r  un  ]dx    un2 dx  0


b 2 2 b
T
a a
According to the property of the definite integral
x
 a
u(t )dt  u ( x)

And use the inequality of Schwarz


b b b
  ( x) ( x)dx    ( x)dx   2 ( x)dx
2
a a a

We have
 u(t )  dt  u(t )  dt
x x x x
u ( x)   u(t ) *1dt    1 dt  xa 
2 2
a a a a
.
2.2 The method of Ritz
 Further discussion 2
b b
 x
 b
  
x
   dx
a  a  a  a   a 
2
So u 2
( x ) dx u (t ) dt dx x a u (t ) dt
 
 b 2 
   x  a    u(t )  dt dx    u(t )  dt   x  a  dx
b b 2 b

a  a  a a

(b  a) 2 (b  a) 2
a  u(t )  dt  2 a  u( x) 
b b

2 2
dx
2
And
2
a  u( x)  dx  (b  a)2
b b

2
u 2 ( x)dx
a

 
Thus
 [ P  un   r  un  ]dx   [ min P( x)  un   r  un  ]dx
b 2 2 b 2 2
a a a  x b

a
b

  [ min P( x)
a  x b
2
(b  a )

2  n
u
2
 r  n  ]dx
u
2

. a
b

  [ min P( x)
a  x b
2
(b  a) 2 
 u n 
2
]dx   a
b
 u n 
2
dx
2.2 The method of Ritz
 Further discussion
In conclusion we have

c Kc  B(un ( x), un ( x))   [ P  un   r  un  ]dx    un2 dx


b 2 2 b
T
a a
where

  min P( x)
a  x b
2
(b  a ) 2
which means the matrix K is positively definite.
So the equations
Ka  b
have the unique solution.
 More explanation
n n

 a a (T ,  )  2 a ( f ,  )
Because
F (a1 , a2 , , an )  i j i j i i
i , j 1 i 1
According to the conclusion of
 2
F
 2F   ai a j  kij ai a j  0
ai a j
That is the second-order variation is greater than zero, which means the solution
we get can make the functional get its minimum.
2.3 Application of the Ritz method
Summary
The discussion before mainly focuses on the mathematical aspect. Now let’s discuss the
application of the Ritz method in mechanics.

The method of Ritz directly use the principle of minimum potential energy
(1)Choose the permissible trial function for the displacement.
ui  ui0  ainuin n  1, 2, ,N
where i is the free index, n is the summation index and ain is the displacement parameter
of the undetermined displacement.
The value range of the free index i depends on the specific problem, for example for the
three-dimensional problem i can be 1, 2, 3.
ui0 is the function of displacement that satisfies the given nonhomogeneous boundary
condition of displacement, which means ui0  ui on Su.
uin is the function of displacement that satisfies the homogeneous boundary condition of
displacement, which means uin =0 on Su.
(2)Write the expression Π(ui) of the total potential energy of the given elastic system ,
substitute the possible trial function of displacement and obtain the expression Π(ain) of
the total potential energy expressed by the undetermined displacement parameters.
(3)Calculate the variation of the total potential energy Π.
Because the form of the displacement function has been determined so only the amplitude
(the undetermined coefficient ain) will change when solving the variation.
2.3 Application of the Ritz method
 Summary
(3)According to the principle of minimum potential energy

   ain  0
ain
Because δain are mutually independent, the coefficients equal to zero. That is

 0 n  1, 2, , N
ain
This is the solving equation of the Ritz method.
It’s actually the approximate equilibrium equation expressed by the displacement
parameters.
For the linear elastic problem total potential energy Π is the second-order functional
of the displacement and its derivative. Substitute the trial function of displacement and
we can get the quadratic function of the undetermined coefficient of displacement.
So the equation above is actually the linear algebraic equations of ain, which can be
easily solved by computer.

(4)Solve the algebraic equations above and we can get the undetermined parameter of
displacement and then substitute it into the equation below.
ui  ui0  ainuin n  1, 2, ,N
We can get the approximate solution of the displacement field.
Stress and strain can also be solved by displacement but generally the obtained stress
field doesn’t satisfy the static equilibrium equations.
2.3 Application of the Ritz method
The beam with uniform load
Let’s consider a beam example. q

The force condition is as shown in the picture.


M0 Ml

Q0 l Ql
y

Now let’s derive the differential equilibrium equations and the static boundary
conditions expressed with the beam deflection according to the principle of
minimum potential energy.
The advantage of variation method is that one can easily obtain the correct
boundary conditions.
Let the beam deflection be w(x). Denote the curvature of the beam axis in flexure
as κ. And the strain energy of the beam is
1 L
U   M  dx
2 0 d 2w
Because M=EIκ. Notice that in the case of small deformation we have   2
2
dx
So the strain energy can be written as U  1
L  d w
2

2 0
EI  2  dx
 dx 
2.3 Application of the Ritz method
The beam with uniform load
The boundary S of this problem is the upper and lower surfaces and the two ends. If the direction
of load is as shown in the picture then the potential energy of the external force is
L  dw   dw 
V    qwdx  Q0 w0  M 0    Q w
L L  M L 
0
 dx 0  dx  L
The total potential energy of2 the beam is
1 L  d 2w  L  dw   dw 
   EI  2  dx   qwdx  Q0 w0  M 0    QL wL  M L  
2 0  dx  0
 dx 0  dx  L
Solve the variation of the equation above
2
1 L  2
d w  L  dw   dw 
   EI   2  dx   q wdx  Q0 w0  M 0    QL wL  M L  
2 0  dx  0
 dx 0  dx  L
2 L
1 L  d 2w    dw  
  EI   2  dx   q wdx  Q w 0   M  
L L

2 0  dx  0
  dx   0
Calculate the first term
2
1 L  d 2w  L d 2w  d 2w  L d 2 w  dw 
2 0
EI   2  dx   EI 2   2  dx   EI 2 d    dx
 dx  0 dx  dx  0 dx  dx 
L
 d 2 w  dw   L d  d 2 w  dw
.   EI 2      0  EI 2   dx
 dx  dx  0 dx  dx  dx
L L
 d 2 w  dw    d  d 2 w  L d
2
 d 2w 
  EI 2       w  EI 2    0 2  EI 2   wdx
 dx  dx   0  dx  dx  
0
dx  dx 
2.3 Application of the Ritz method
The beam with uniform load
Substitute the equation above into the variation of the total potential energy
L L
L  d  d w 
2 2
 dw   d w
2
  d  d w2

    2 
EI 2 
 q   wdx     EI  M    w   EI 2 
 Q 
 
2
0
 dx  dx   dx  dx 0  
dx dx   0

δw is totally arbitrary, so the equation above holds when


Equilibrium equations
d 2  d 2w 
2 
EI 2   q 0  x  L
dx  dx 
 dw   d w
2

    EI 2  M  
 dx   dx 
 x  0, L
 d  d w
2

 w   EI 2   Q   The boundary conditions
 dx  dx  
Apparently the equations above are the differential equations of the beam
curvature and the boundary conditions of the displacement.
Now let’s discuss how to express the static boundary conditions by deflection in
different situations of support.
2.3 Application of the Ritz method
The beam with uniform load
General types of the boundary conditions are
 dw   d w
2

    EI 2  M  
 dx   dx 
 x  0, L
 d  d 2w  
 w   EI 2   Q  
 dx  dx  
dw
1. For the fixed ends because w   0, the equations hold naturally.
dx  dw 
  0
2. For the simply supported ends we have w=0(because δw=0). Also  dx 
So the second term of the equations holds naturally and from the first term we have
d 2w
EI 2   M
dx
If there’s no bending
2
moment of the simply supported end of the beam, then it can be
simplified as EI d w  0
dx 2
 dw 
3. For the free ends because  w  0,     0 , we have
 dx 
d 2w d  d 2w 
EI 2   M ,  EI 2   Q
dx dx  dx 
If one of or both of the moment and shearing force equal to zero, then the boundary
conditions can be simplified accordingly.
2.3 Application of the Ritz method
The beam with uniform load
The expression of the total potential energy of the
simply supported beam shown in the picture is
2
1 L  d w
2
L l
x

   EI  2  dx   qwdx y
2 0
 dx  0
1. The trial function also the deflection is
w  ax(l  x)
It apparently satisfies the boundary condition
w(0)  w(l )  0
Substitute it into the expression of the total
potential energy and we have
EI
  2a 
l l
 dx  qa  x(l  x)dx
2

2 0 0

1
Only a is the variable so  2 EIla 2  ql 3a
6
 1 3 ql 2
Thus the approximate solution of the deflection is.  0  4 EIla  ql  0  a 
a 6 2 24 EI
ql
w x(l  x)
24 EI
2.3 Application of the Ritz method
The beam with uniform load
In engineering practice, people mostly focus on the maximum of deflection, which
occurs in the middle ql 2 l l ql 4
wmax  (l  ) 
24 EI 2 2 96 EI
The exact solution obtained by mechanics4of materials is
5ql
wmax 
384 EI
The error is around 20%, which is quite dissatisfying.
x 
2. Change the trial function to w  a sin  
 l 
which satisfies the boundary conditions apparently. Substitute the equation into
the expression of the total potential energy that
EI 2    l 2   x  x
2
l
 a    sin   dx  qa  sin   dx
2 l  0
 l  0
 l 
. EI  4 2 2ql
 a  a
4l 3

2.3 Application of the Ritz method
The beam with uniform load
So
 EI  4 2ql 4ql 4
0 a 0 a 
a 2l 3
 EI  5
So the approximate solution of deflection is
4ql 4 x 
w sin  
EI  5  l 
Similarly the deflection in the middle is the maximum.
4ql 4   l  4ql
4
wmax  sin  
EI  5
 l 2  EI 
5

The error is 0.38% , which is quite satisfying compared to the solution above.

One can take more terms or even infinite series to obtain more accurate
result.
We can notice the advantage of the direct method of variation problems that we
only need to solve the algebraic equations rather than the differential equations.
Once the trial function is approprivately, it only needs to solve several
(sometimes even one) algebraic equations to get the result accurate enough.
2.4 The method of Galerkin
The method of Galerkin
Galerkin proposed another direct method of variation problem in 1915, which is a special
form of the Ritz method in mechanics.
Now let’s look at his idea.
The variation of the minimum potential energy is
    ( ij , j  fi ) ui dV   ( ij n j  pi ) ui dS  0
V S
, which is the origin of the Ritz method: Establish and substitute the trial function satisfying
the displacement boundary conditions on Su.

However the trial function of the displacement is required to satisfy both the boundary
conditions of the displacement on Su and the external force on Sσ in the method of Galerkin.

So the variation of the potential energy can be written as


    ( ij , j  fi ) ui dV  0
V
Because δui=uin δain. And δain is mutually independent. We have

V
( ij , j  fi )uin dV  0 n  1, 2, ,N
This is the solving equation of the Galerkin method, which is also the linear
equations after integration.
2.4 The method of Galerkin
The method of Galerkin
Take the beam as example. We have obtained its functional form of the total potential
energy l
   F ( x, w, w, w)dx
0
Using the operation notations of variation we can get its first-order variation as
l F d  F  d 2  F 
    w[     2   ]dx
0 
w dx  w  dx  w 
l l
d  dw 
 [ w( Fw   w 
F )]  [ F 
w  ]
dx 0  dx  0
We can know that the terms outside the sign of integration are the variation of
potential energy on the boundary.
When replace the real w by the approximate displacement function wn,there usually
remain only one term of integral because the trial function must satisfy all the
boundary conditions according to the method of Galerkin.
Rewritten as
l F d  F  d 2  F  l
0  wn [ w  dx  w   dx 2  w ]dx  0  wn f ( x, w, w, w, )dx
Substitute the trial function and we can get a set of linear equations.
2.4 The method of Galerkin
The method of Galerkin
Example:Solve the approximate value of the deflection of the beam with two fixed
ends and uniform load.
Solve:Let the trial function of the deflection be
2 x
w  a (1  cos ) l
x

l y

, which apparently satisfies the boundary conditions of displacement.


w(0)  w(l )  0
w(0)  w(l )  0
The equation above can be chosen as the trial function of the Ritz method. So it can be
solved by the Ritz method.
However there are no boundary conditions of force in this problem. So satisfying the
equations above (the displacement and the rotation of two ends equal zero) means
satisfying all the boundary conditions.
So this function can also be the trial function of the Galerkin method and let’s solve the
problem by the method of Galerkin.

The integral equation of Galerkin is


l d 4w
0 ( EI dx 4  q)wdx  0
2.4 The method of Galerkin
The method of Galerkin
Solve:Substitute the trial function of displacement and
 2  2 x 2 x 2 x
4
l l
 EI   a 0 cos (1  cos )dx  q  (1  cos )dx  0
 l  l l 0 l
which leads to
 2  l
4
ql 4
EI   a  ql  0  a  4
 l  2 8 EI
So the approximate curve of the deflection is
ql 4 2 x
w  4 (1  cos )
8 EI l
We can notice an advantage of the Galerkin method that we don’t need to judge
whether the structure is indeterminate.
Because the solution is the same no matter the structure is static determinate or
not.
2.4 The method of Galerkin
 Comparison to the method of Ritz
Example:Solve the deflection of a cantilever beam with uniform load.
Solution 1. The Ritz method
The first-order approximate trial function is
x
w  a(1  cos )
2l x
Substituted into the expression of the l
total potential energy and we have y
1
   EI  w  dx   qwdx
l 2 l

0 2 0

  l
4
1 2
 a 2 EI    qal (1  )
2  2l  2 
 32 2 ql 4
Because a  0 , we have a  4 (1  )
  EI


So the maximum of the deflection at x=l is
322 ql 4
wmax  a(1  cos l )  4 (1  )
2l 4   EI
1 ql
4.5% smaller compared to the accurate result
8 EI
2.4 The method of Galerkin
 Comparison to the method of Ritz
Solution 2. The Galerkin method
The method of Galerkin requires that the trial function must satisfy the boundary
conditions of both the external force and the displacement. So the trial function of the
Ritz method is not appropriate here.
Exercise: What will be the result if adopt the trial function of Ritz method?

It’ll be easier to consider the boundary condition of the external force first when
looking for the trial function of the Galerkin method.
Let d 2w x
2
 a(1  sin )
dx 2l
Apparently it satisfies the condition that the moment and the shearing force equal to
zero at x=l
w(l )  0 w(l )  0
Integrate the equation above twice and
 1 2  2l 2 x 
w  a  x    sin  Ax  B 
 2   2l 
The integral constants A and B are determined by the boundary condition that the
trial function satisfies at x=0. That is w(0)  0 w(0)  0
2.4 The method of Galerkin

 Comparison to the method of Ritz


Solution: 1 2l  
2l
2
x 
The trial function is w  a  x 2  x    sin  B
 2    2l 
Substitute the trial function into the solving equations of the Galerkin method and we have
1 1 8
  3 4
a 6  
ql
3 4 EI

2 
Thus the biggest deflection at x=l is

1 2 4 ql 4
wmax  al (   2 )  0.126
2

2   EI
The result is 0.8% bigger than the accurate solution and is apparently better than the first
approximate solution of the Ritz method.

The trial function of the Galerkin method is also the trial function of the Ritz method. You can try to
substitute the trial function above into the Ritz method to solve the problem.
2.5 Conclusion

 Conclusions
The solving equations of the Galerkin method can be explained by the idea of the weighted
residual method.
For example according to the equilibrium conditions the real solution w of the beam with
uniform load should satisfy:
d 4w l d 4w
EI 4  q  0   ( EI 4  q) wdx  0
dx 0 dx
The trial function wn doesn’t always satisfy the equilibrium equations in the domain
once relax the requirement. So there will be a nonzero residual on the right side after
substituted into the equilibrium equations.
n
d 4 wn
wn   aii  EI 4
 q  Ri  0
i 1 dx
The method of weighted residual is to adjust the undetermined parameters of the trial
functions to make the integral of the product of the residual and some weight
functions equals zero in the entire domain. And we can obtain the approximate
solution. l l d 4w 
0 i 0  dx4  q  i dx  0 i  1, 2,
R dx  EI n
,N

There are different ways to choose the weight function. And when the weight function
is the admissible function φi of the trial function, it becomes the method of Galerkin.
III. The method of weighted residuals
3.1 Brief

We’ve discussed the approximate solutions of the physical problems or the differential
equations.

However, not all the differential equations can be transformed into the variation
problem, which means we should first think about the inverse problem of the variation.

There are two methods to discuss the variation problem:


Take the simple functional of one independent function for example
(1)For the function family with two given parametric variables α and β
y  f ( x,  ,  )
Solve the functional of which the function above is the extreme
(2)For the given differential function

 ( x, y, y, y)  0
Solve the functional of which the function above is the Euler equation.
For the first question, we usually solve its differential equations by eliminating the
parameters and then obtain the coresponding functional, which actually turns out to be
the second question.
3.1 Brief

As stressed before, all the differential equations don’t have the functional of which the
differential equation itself is the Euler equation.
Take the second-order ordinary differential equation for example
 ( x, y, y, y)  0
If it is the Euler equation of the functional below
b
Q[ y ]   F ( x, y, y)dx
a
Then
d
 ( x, y, y, y)  0  Fy  Fy  Fy  Fyx  yFyy  yFyy
dx
So
d d
 
 y   y  ( Fyy  Fyyx  Fyy  y Fyyy  y Fyyy )  Fyy
dx dx
 ( Fyyx  yFyyy  yFyyy )  ( Fyyx  yFyyy  yFyyy )  0

So φ must satisfy d
.  y   y  0
dx
3.1 Brief

We can prove that


   xy  2 xy  x 2 y  0
is the Euler equation of the functional below
1 b 2
Q[ y ]   ( xy  x 2 y2 )dx
2 a
However after eliminate the common factor x in the equation φ, we have
   y  2 y  xy  0
where d d
 y   y  2  x  1  0
dx dx

So there is no functional of which the equation above is the Euler equation.

So we have to adopt the weighted residual method


(WRM) for this kind of problem.
3.1 Brief

 The basic idea of the weighted residual method


We have already mentioned the weighted residual method before when discussed the
method of Galerkin and now let’s explain it from a more mathematical perspective.
Suppose to solve the equation of a differential operator in a function space H.
Tu  f
where T is the operator, u is the undetermined and f is already known in H.

1. Find a set of linearly independent elements φ1,φ2, …, φn in H and denote the formed
M  span{1 , 1 , , n }
subspace as

2. Find a set of linearly independent weight functions ω1, ω 2, …, ω n in H and let


W  span{1 , 1 , , n }
3. In the space M let n
u0   aii
i 1
where a1, a2, …, an are the undetermined coefficients.
3.1 Brief
The basic idea of the weighted residual method
4. Substitute u0 into the equation. Because it cannot perfectly satisfy the equation so let
the residual be
R  Tu 0  f  0
5. The basic idea of the weighted residual method is to make the projection of the residual
R to the weighted function space W equals zero thus we can determine the coefficients a1,
a2, …, an The geometric
( R, i )  0 (i  1, 2, , n) meaning of the
inner product is
the projection
The inner production is written in the integral form as defined before or can be more
precisely expressed as
n
(Tu0  f , i )  (T akk  f , i )  0 (i  1, 2, , n)
k 1
We can obtain the approximate weighting solutions by solving the equation above.
The method of weighted residual can be regarded as the unified model of solving
approximate solutions of the equations. Different basis functions of the subspace
M and the weighted function space W can make up different ways to solve the
approximate solutions.
3.2 Example
Example
Solve the deflection at the middle of the simply supported beam with uniform load.

x
l
y

As discussed before the control differential equation and the boundary conditions of this
problem are d 4w
EI 4  q  0
dx
x  0 : w  0, w  0
x  l : w  0, w  0
Now let’s use the low dimensional approximate method to solve the problem for better
understanding of the essence of the weighted residual method. 。
The low dimensional approximate method means there is only one or several
undetermined parametric variables in the trial function
x
w1  c sin First order approximation
l
x 3 x Second order approximation
w2  c1 sin  c2 sin
l l
We can prove that the equations above have already satisfy the boundary conditions
so there is only internal residual.
3.2 Example
Example
Solve the deflection at the middle of the simply supported beam with uniform load.

x
l

The Galerkin method y


Choose the basis of the trial function as the weighted function and make the product integral
of the internal residual and the weighted function equals zero in the domain. Thus we can
derive the algebraic equations to solve the parametric variables.

Solution
For the first order approximate
l x 
l  
4
x  x
0 R sin l dx  0  EIc  l  sin l  q  sin l dx
 
  l
4
l
 EIc    2q  0
l  2 
4ql 4
So the parametric variable c is c  5
 EI
c is apparently the max deflection at the mid-span and the error is 0.386%
compared to the exact solution.
3.2 Example
Example
Solve the deflection at the middle of the simply supported beam with uniform load.

x
l
y
Moment Method
The moment method is similar to the Galerkin method but the difference is that the
weighted functions is the power series of the coordinates 1,x, x2,… and the number of the
weighted functions is equal to the number of parametric variables. Let the product
integral of the internal residual with every weighted function equals zero thus derive the
equation of the parametric variables.
Solve
For the first order approximate
 
3
l
0 R  1dx  2 EIc    ql  0
l 
So the parametric variable c is
ql 4
c 3
2 EI
c is apparently the max deflection at the mid-span and the error is 23.85%
compared to the exact solution. 。
3.2 Example
Example
Solve the deflection at the middle of the simply supported beam with uniform load.

x
l
y
Least Square Method
Make the square integral of the internal residual get its extremum in the solving domain,
which means the derivative of the integral of every parametric variable equals zero. And we
can get the algebraic equations to solve the parametric variables.

Solve
d l 2 l dR
For the first order approximate
dc 0
R dx  2 0
R
dc
dx

  l
4
l
 EIc    2q  0
l  2 
4
4ql
So the parametric variable c is c 5
 EI
c is apparently the max deflection at the mid-span. The approximate solution
obtained in this way is the same as the Galerkin method.
3.2 Example
Example
Solve the deflection at the middle of the simply supported beam with uniform load.

x
l
y

Collocation Method
Select several collocations in the solving domain at which the internal residual equals zero
thus we can get the equations to solve the variables. The number of the collocations should
be equal to the number of undetermined parametric variables.

Solve
For the first-order approximation there’s only one parametric variable so only one collocation
is needed. Let the collocation be the middle point.
 
4

R x  l  EIc    q  0
2 l 
So the parametric variable c is ql 4
c 4
 EI
C is apparently the max deflection at the mid-span and the error is 21.16%
compared to the exact solution.

3.2 Example
Example
Solve the deflection at the middle of the simply supported beam with uniform load.

x
l
y
Subdomain Method
Divide the solving domain into several parts and make the integral of the internal residual
equals zero in very subdomain thus we can get the algebraic equations to solve the parametric
variables. The number of the subdomains divided should be equal to the number of the
undetermined parametric variables.

Solve
There’s only one parametric variable in the trial function for the first order approximation. So
it only requires the integral in the solving domain equals zero. So now the subdomain method
is the same as the moment method.
 
3
l
0 Rdx  2 EIc
4
   ql  0
l 
ql
So the parametric variable c is c 3
2 EI
C is apparently the max deflection at the mid-span and the error is 23.85%
compared to the exact solution.
3.2 Example
Example
Solve the deflection at the middle of the simply supported beam with uniform load.

Conclusion
We introduced five methods to solve the approximate solution with the same first order
approximate function. We can find out that the errors of the least square method and the
Galerkin method are the smallest.

Exercise: Try to use the five methods above to solve the problem with the second
order approximate function.

Discussion
The methods above can be generally called the weighted residual method and they are
actually very similar to each other, which can all be expressed in a unified formula. That is to
form different methods by choosing the proper weighted functions.

But why these methods are so similar? What’s their inner link? And what’s the essence of the
unified formula of the weighted residual method?

We have to first study the functional analysis and the Sobolev space.
3.3 Further discussion
The functional of variance
The governing differential equations in the theoretical research and the engineering
application can be generally expressed as
Pm  u   f 0
The corresponding boundary conditions are
Qt (u)  ft (t  1, 2, , m)
where Pm is a m-order differential operator, Qt is a (t-1) order differential operator in the
normal direction of the boundary, ft is the given coordinate function and the boundary Γ
of the domain Ω is m
 t
t 1
Usually the problem is supposed to be regular and well posed.
The regularity is the smoothness of the solution, which means if there exists
determined degree of smoothness of the boundary conditions then the solution
must hold certain degree of smoothness.
The “well-posed” means the existence of the solutions, the uniqueness and the
continuity dependence of the definite conditions.
n
Let the trial function be u   cii
i 1 function and ci is the corresponding undetermined
where {φi} is the basis of the trial
parametric variable.
3.3 Further discussion
The functional of variance
The trial space consist of all the trial functions is called the trial function space.
Because the trial function may satisfy neither the equations nor the boundary conditions. So
there will be respectively the internal residual RI and the boundary residual Rt after
substituted into the equations.
Solve the weighted integral of all the residual squares in the corresponding domain and we
have m
J R (u )    I2 RI2 d     t2 Rt2 d 
 t
t 1

The formula above is called the functional of variance, which is actually the functional of the
square residual , where βI is the weighting coefficient in the domain and βt is the
weighting coefficient of the boundary conditions.

The weighting coefficients have the following two effects:


1. Increasing βt to ensure the better satisfaction of the boundary conditions if they are very
important.
2. The weighting coefficients can be used as the dimension to adjust the internal and external
residuals to make the additive operation more meaningful.
Apparently the functional of variance has two important conditions:
1)The functional has the minimum that equals zero.
2)The solution that makes the functional get its minimum is equivalent to that of the
boundary problems of the differential equations.
3.3 Further discussion
The functional of variance
The extremum problems of energy variation can be solved by variation equations in the
principle of energy variation.
Similarly we can use the variation equation to solve the minimum problem of the
functional of variance here.
 J R (u)  0
And there is the corresponding variation principle of the variance functional.

The variation principles of the variance functional and the energy functional are
very similar:
1.They are both the extremum principles of the functional.
2.They are both the second-order positive functional for linear problems so the necessary
and sufficient condition that the functional gets its minimum is that the first-order
variation equals zero.

The differences are:


1. The quadratic term about u in the energy functional forms the certain energy with
general and apparent physical meaning. But its minimum can not be known in advance.
2. The minimum of the functional of variance is determined to be zero. But there’s no
explicit physical meaning for the quadratic term about u. The square root can be directly
defined as the mean square modulus.
3.3 Further discussion
The functional of variance
After the definition of the modulus we can similarly make up the minimizing sequence of
the functional to solve the problem.
The primary task of the numerical analysis is to transform the infinite dimensional
problem to the finite one.
Now let’s explain from the geometric aspect.
Rewrite the solving equation of the weighted residual method in the inner product form as
(Tu0  f , i )  (Tu0  Tu, i )  0
where u0 is the trial solution and u is the real solution so
(Tu0 , i )  (Tu, i )
It can be seen that the projections to the space W of Tu0 and Tu equal to each other.

So the geometric explanation of the weighted residual method is


Find an approximate solution of Tu=f in the trial function space M to make the
approximate value Tu0 and the real Tu have the same projection in the space W.
When the dimension n→∞ of the subspace M, the minimum of the variance functional
will furtherly decrease, which means the approximate solution is approaching to the real
solution.
Theoretically speaking only when the range of the trial function has been expanded to all
the admissible functions, the minimum of the functional is zero.
3.3 Further discussion
The functional of variance
The different kinds of weighted residual methods above are actually the different
methods of forming the minimizing sequence.
We can see that the principle of the weighted residual method is quite simple: The
average value of the residuals must be zero under certain situations. It can also
transform the differential method of the problem to an integral one, which is
applicable for the physical system with or without functional. So it’s a kind of general
mathematical treatment and is widely used in the fields like hydromechanics, solid
mechanics and thermal physics.
It should be noticed that the solving equations derived by the direct Galerkin method
of the variation equation and the Galerkin method of the weighted residual are the
same for the problems with functional.
The coefficient matrixes of the equations obtained are symmetric, which is quite
important for solving the equations.
We’ll call both the two kinds the Galerkin method because we mainly discuss the
finite element method of the elastic mechanics.
Exercise:Derive the differential equation and the boundary conditions of the beam on
elastic foundation by the variation method and select the first order approximation by
p
different weighted residual methods.
x

l/2 l/2
IV. The finite element method
4.1 Brief
We took the one dimensional problem for example when introducing the different
methods. So let’s look at a two-dimensional example for better understanding of the
necessity of the finite element method.

Example
Solve the Poisson equation by the Galerkin method
  2u  2u
 : u  2  2  2
 x y
 :u  0

where Ω:{|x|<a,|y|<a},Γ is the boundary of Ω.

Solution
Choose two linearly independent basis functions as
1  (a 2  x 2 )(a 2  y 2 )
2  (a 2  x 2 )(a 2  y 2 )( x 2  y 2 )
Let u2  c11  c22

The formula above satisfies all the boundary conditions so it can be used as the trial
function of the Galerkin method.
4.1 Brief
The solving equation of the Galerkin method
 1 Rd   0  1  u2  f  d   0
    (T1 , 1 )c1  (T2 , 1 )c2  ( f , 1 )


  2 Rd   0   2  u2  f  d   0 (T1 , 2 )c1  (T2 , 2 )c2  ( f , 2 )

Then
16 16 8 12 16 16 10
(T1 , 1 )   11d    a (T1 , 2 )   2 1d    a
 5 3 3  35  5  3  3
8 128 10 22  32  8  2 12
(T2 , 1 )   12 d   
 35  5  3
a (T2 , 2 ) 
2 2 d   
7  5  5  27
a
16  4 6
32 6
( f , 1 )   21d    a ( f , 2 )   22 d    a
 9  5 9
Thus 1 5  7  37 1 15  35
c1  2 c2 
a 8  277 a 4 16  277
4.1 Brief
So the approximate solution is

u  u2  c11  c22
1 35 2  15 2 2 
 ( a 2
 x 2
)( a 2
 y )  74  ( x  y )
a 16  277
2
 a 2

Apparently the boundary curve is square.
We can see that even for this very simple equation, to form a proper trial function can be
very difficult when the boundary curve is complicated.

Besides the Galerkin method requires a relatively high level of smoothness degree of the
trial function. For example the second order derivative of this trial function must be
integrable. So we usually transform the solving equation to an equivalent integral weak
form to decrease the requirement of the trial function smoothness.

So we can solve the partitioned interpolation of the solving domain after obtaining the
equivalent integral weak form of the equation and then expand the range of the trial
functions.
It leads to the finite element method applicable for a wider domain.
4.1 Brief
Now let’s discuss the main idea of the finite element method.
1. The classical finite element method is to find the variation problem of the
differential equations through the principle of variation and thus to find the
corresponding functional.
One can directly start to establish the functional from the variation method for
the mechanical problems.

2. However the functional is not easy to find for some complicated


differential equations so here we can adopt the weighted residual method.
The basic idea is to directly calculate the inner product of the basis function and
the two ends of the equation rather than find out the corresponding functional
and then the discrete solving equations can be obtained.
The equivalent integral forms obtained from different methods are actually
the same for the equation that has its corresponding functional.

The requirement of the smoothness degree of the trial function is relatively high
when using the equivalent integral method obtained by the Galerkin method.
It has to be transformed to equivalent weak form to decrease the requirement of
the smoothness degree.
The common approach is the integration by parts or to directly solve by the
principle of virtual work for the mechanical problems.
4.1 Brief

It has been discussed before that


d
  P( x)u ( x)   r ( x)u ( x )  f ( x )
dx
the equivalent functional of the equation above (the equivalent integral form) is
Q[u ( x)]      P( x)u ( x)  u ( x)  r ( x)u ( x )u ( x )  2 f ( x )u ( x )  dx
b

a 
 
Integrate the first term by parts
Q[u ( x)]   P( x)u( x)u ( x) a    P( x)u ( x)u ( x)  dx
b b

a  
   r ( x)u ( x)u ( x )  2 f ( x )u ( x)  dx
b

a u(a)= u(b)=0

   Pu2  ru 2  2 fu  dx
b

Apparently it only needs the trial function u(x) to be C0 continuous, which is


the integral weak form.
The expression can be rewritten as follows so as to derive the formulas more
easily.
.
1 b
Q[u ( x)] 
2 a
 Pu  2
 ru 2  2 fu  dx
4.2 The mathematical description of the finite element method

We’ll discuss the finite element method from a mathematical aspect. 。


Take the single-variable function for example.

Suppose the undetermined function u(x) of the operator equation Tu=f is


defined in the interval [a,b]. Arbitrarily select the discrete points.
a  x0  x2  xi  xi 1  xn  b
xi is called the node in the finite element method and the interval [xi, xi+1] is
called the unit.
Establish the linear basis functions for every node xi

 1 x  xi i ( x)

i  线性 x   xi 1,xi 1  linear
1
x0 xi 1 xi xi 1 xn
 0 其他
 others

So there are totally a number of n basis functions according to the n discrete


points in the whole domain.
1 , 2 , , n
4.2 The mathematical description of the finite element method

Apparently φi(i=1,2,…,n) is linearly independent and form the subspace as


follows.
M  span{1 , 2 , , n }
So any element of M can be linearly expressed as follows by the basis functions
above. n
v( x)   aii
i 1
The basic ideal of the finite element method is to use un as the approximation of
the real solution u of the equation in the infinite dimensional function space.
That is to select proper a1 , a2 ,…, an to make v become the best element
approaching u in M. u ( x) u
u1 u2 v( x)
n

x0 x1 x2 xn 1 xn
0 ( x)

1 ( x)
1
n ( x)
1
4.2 The mathematical description of the finite element method

The value of every basis function equals 1 at the according nodes and 0 at
other nodes due to its special construction (usually called the property of
Kronecker).
So the coefficients of the approximate solution a1 , a2 ,…, an are just the
function values ui ( i = 1,… , n) of v (x) at the discrete points x1 , x2 ,…,
xn.

To solve ai(i=1,2,…,n) the classical finite element method is to change the


equation to the equivalent variation problem according to the principle of
variation. Thus the minimum problem of the functional can be transformed
into the extremum problem of multivariable function in finite subspace and be
solved through the linear equations obtained.  1 x  xi
From the definition of basis function we have 
i  线性 x   xi 1,xi 
 0 其他

We can know that the influence domain or the support set of the basis
function is quite small, so the coefficient matrix of the obtained linear
equations is very sparse for easier calculation.
We were actually establishing the minimizing sequence of the functional.
And it can be proved that un is really the approximation of the real
solution u in the finite subspace.
4.3 Solving steps of the finite element method

Now let’s summarize the solving steps of the finite element method
Take the one dimensional equation for example

1. Establish the equivalent integral weak form of the function


For the differential equation
 Pu   ru  f

 
u (a)  ua , u (b)  ub
its equivalent integral weak form is
Q[u ( x)]    Pu 2  ru 2  2 fu  dx
1 b
2 a
2. Discretize the solving domain
Divide the solving domain of the equation into a finite number of sections and
discretize the whole domain into an assembly of finite elements. So the discrete points
of the interval [a, b] are
a  x0  x2  xi  xi 1  xn  b
So the functional in the whole domain can be discretized into the sum of the
n
Q[u ]   Q j
functionals in each of the n elements. e
4.3 Solving steps of the finite element method

 3.Establish the basis function of the slicing difference


Choose the sectioned polynomial φi as the basis function, so the trial function is
n
v( x)   uii ( x)
i 0

where the basis function is as follows


( x1  x) /( x1  x0 ) x  [ x0 , x1 ] 0 ( x)
0  
 0 x  [ x0 , x1 ] x0 1 x1

( x  xi 1 ) /( xi  xi 1 ) x  [ xi 1 , xi ]
 i ( x)
i  ( xi 1  x) /( xi 1  xi ) x  [ xi , xi 1 ] 1
 0 x  [ xi 1 , xi ] [ xi , xi 1 ]
 xi 1 xi xi 1
n ( x)
 0 x  [ xn 1 , xn ]
n   1
( x  xn 1 ) /( xn  xn 1 ) x  [ xn 1 , xn ] xn 1 xn
4.3 Solving steps of the finite element method

 3.Construct the basis function


 Denote the interval (xi-1, xi ) as the element ei and denote the node xi-1,
xi as the node 1 and node 2 with local numbers. And define the shape function
N1(i )  ( xi  x) /( xi  xi 1 ) The superscript of N is
the number of elements,
N 2(i )  ( x  xi 1 ) /( xi  xi 1 ) and the subscript is the
number of local nodes.
So the basis function can also be expressed
0 ( xas
)
 N1(1)
x  e1 1
0   x0 e1 x1
 0 xe 1

 N 2(i ) x  ei i ( x)

i   N1(i 1) x  ei 1 1
ei ei 1
0 x  e e
 i i 1 xi 1 xi xi 1
n ( x)
 0 x  en
n   ( n ) en 1
 N2 x  en
xn 1 xn
4.3 Solving steps of the finite element method
4. Element analysis
We establish the basis function and solve the extremum equations by element when using
the finite element method. n
Calculate the functional by element, v( x)   aii ( x) for example for the normal
element ej i 0

Qj 
1
2 e j
 Pv  2
 rv 2
 2 fv  dx

It should be noticed that only the basis functions φj-1 and φj of the trial function v(x) are
not equal to zero in the element ej(xj-1<x<xj) ( j)
N1( j ) ( x) N 2 ( x)

 j 1 ( x)  j ( x)
specificaly, it is
ej
x j 1 xj
So in the element ej the trial function is
 j 1 ( x)  N1( j ) ( x)  ( x  x j 1 ) /( x j  x j 1 )
 j ( x) n N 2( j ) ( x)  ( x j 1  x) /( x j 1  x j )
v( x)   aii ( x)  u j 1 N1( j ) ( x)  u j N 2( j ) ( x)
i 0
4.3 Solving steps of the finite element method

4. Element analysis


Substitute the equivalent integral weak form and we have
Qj 
1
2 ej
 Pv  2
 rv 2
 2 fv  dx
2
 dN1( j ) dN 2( j ) 
1
  [ P  u j 1 uj   r  u N
j 1 1
( j)
 u N
j 2 
( j) 2
 2 f  u N
j 1 1
( j)
 u j 2 ]dx
N ( j)

2 ej
 dx dx 
2
 dN1( j )   ( j) ( j)
( j) 
  r  N1  ]dx  u j 1u j ej  P
1 2 dN dN
 u j 1  [ P 
2
( j) 1 2
 rN1 N 2  dx
( j)

2 ej
 dx   dx dx 
2
 dN  ( j)
 r  N 2  ]dx  u j 1  fN1( j ) ( x)dx  u j  fN 2( j ) ( x)dx
1 2
 u j  [P  
2 ( j) 2

2 ej
 dx  ej ej

 dN( j ) dN ( j ) 
Let ej
k   P  rN N   dx
( j) ( j)
ej  dx dx 
 
bej   fN( j ) ( x)dx
. ej
4.3 Solving steps of the finite element method
4. Element analysis
So in the interval (xj-1<x<xj) the functional value of the element ej can be
written as 1 1
Q j  u 2j 1k11ej  u j 1u j k12ej  u 2j k22ej  u j 1b1ej  u j b2ej
2 2
So the derivate of the element is
Q j
 k11ej u j 1  k12ej u j  b1ej
u j 1
Q j
 k12ej u j 1  k22ej u j  b2ej
u j

Written in the form of matrix


 Q j 
 
 u j 1 
  11
k ej
k12ej   u j 1   b1 
ej
ej  j 1 
u
 Q   k ej       K    B ej

k22ej  u j   b2ej  u
 j   12  j 
 u 
 j 
Kei is also called the element stiffness matrix of element ej.
4.3 Solving steps of the finite element method
5. Assemble the equations
We discussed a specific element at one time just now. And we can obtain a
number of n derivative formulas of elements from e1 to en.
Add them together according to the public nodes and then get the overall
extremum equations.
First write down the equations
Q n Q j
  0 (i  0,1, , n)
ui j 1 ui
For arbitrary element ej
1 1
Q j  u 2j 1k11ej  u j 1u j k12ej  u 2j k22ej  u j 1b1ej  u j b2ej
2 2
After recurrence we have
1 1
Q j 1  u 2j k11e j 1  u j u j 1k12e j 1  u 2j 1k22
e j 1
 u j b1e j 1  u j 1b2e j 1
2 2
It can be seen that the other terms don’t contain uj except Qj and Qj+1, which
means only the terms Qi and Qi+1 have the partial derivatives of ui .Thus
Q n Q j Qi Qi 1
.   
ui j 1 ui ui ui
4.3 Solving steps of the finite element method
 5. Assemble the equations
So
Q Q1 Q2 
   0
u1 u1 u1

Q Q2 Q3 
   0  Q n Q j
u2 u2 u2    0 (i  1, 2, , n)
 ui j 1 ui

Q Qn 
 0 
un un 
According to the element analysis before we have
Q1
 k12e1u0  k22e1u1  b2e1
u1
Q2
 k11e 2u1  k12e 2u2  b1e 2
u1

Substituted into the equations above.


4.3 Solving steps of the finite element method
 5. Assemble the equations
And we can get the following equations
k e1
22  k11e 2  u1  k12e 2u2  k12e1u0  b2e1  b1e 2



  KU  B
e n 1

e n 1

k21 un  2  k22  k11 un 1  k12 un  b2
en en e n 1
 b1 
en


k21enun 1  k22enun  b2en 
Expressed in the form of matrix where
 b e1
 b e2
  u1 
 k22e1  k11e 2 k12e 2 0  2
 e 2 e3 
1  
  b  b  u 2
 k21 k22  k11 k12
e2 e2 e3 3
0   2 1 
   
 
k21 k22  k11 k12  U  
e3 e3 e4 4
K 
0 0
 B
 0 k21e 4 k22
e4
 k11e5 k12e5     
   e n1 en   un 1 
2b  b1   
 
 n   b2en   un 
n
 0 k21 k22  
We can observe the pattern of the global stiffness assembled by the element
stiffness.
4.3 Solving steps of the finite element method
6. Introduce the boundary conditions
We can naturally introduce the functional when discussing the variation problems without special
treatment for the natural boundary conditions.
For the compulsive (essential) boundary conditions we can directly introduce the equations to reduce
the dimensions of the undetermined equations.
It is usually realized by “put 1 on the diagonal”. For example suppose the equations are
 k11 k12 k1n   u1   b1 
k k2 n  u2  b2 
 21 k22 
    
    
 k1n k2 n knn  un  bn 
Suppose the value of a point i on the boundary is known, then

 k11 k12 i列 k1n   u1   b1 


    
    
i行  0 1 0   ui    ui 
    
    
 kn1 kn 2 knn  un  bn 
Do the same treatment for all the given points on the boundary.
4.3 Solving steps of the finite element method
 7. Solve the equations
Use the calculation method of the linear algebraic equations to solve the
equations after the treatment of boundary conditions. And we can get the
approximate solutions at every node in the whole domain
u1 , u2 , , un 1 u0,un are the known
boundary
The approximate solutions in the whole domain are
n
v( x)   u  ( x)
i i
Exercise i 0
Solve the equations by the finite element method
u( x)  u ( x)   x

 u (0)  u (1)  0
Demand:Divide the interval into four elements by the nodes 0,1/4,1/2,
3/4,1.
Hints:The equivalent integral weak form is
1 1 2
Q[u ( x)]   [u  ( x)  u 2 ( x)  2 xu ( x)]dx
2 0
4.4 Mechanical examples
Take a one dimension problem for example
An one-end-fixed uniform elastic rod with uniform section is in equilibrium under its self-
weight and the tension P at the other end. Solve the deformation and stress distribution of
the rod. It is known that the length is L, the sectional area is A and the density is ρ.
O

Establish the coordinate system. Let u(x) be the displacement


of section x and we have
du
 ( x)   ( x)  E ( x)
dx
Of course we can first establish the equivalent integral form
according to the principle of minimum potential energy and
P
then transform it into the weak form by integration by parts. x

Here we’ll directly establish the equivalent integral weak form according to the principle
of virtual work d
Let the virtual displacement be φ(x) then the according virtual strain is  ( x ) 
In terms of the principle of virtual work: dx
The virtual work of the internal force equals that of the external force
l l l
  ( x) ( x)dx  
0 0
EAu( x) ( x)dx   A g ( x)dx  P ( L)
0
So the equation above is the equivalent integral weak form.
4.4 Mechanical examples
The first step of the finite element method is the discretization.
Usually the differential and virtual work equations are called the continuous form.

The so-called discretization is to approximately transform them into algebraic


equations.
Let’s look at the last example again to explain the basic steps of discretization of the
finite element method. 。

 Element subdivision
Divide the interval [0,l] into a number of n parts
0  x0  x1   xn  l
The interval [xi, xi+1] is called the element ei (i=0,1,…,n-1) and x=xi+1 is the
nodes(i=0,1,…,n).

 Displacement mode
The displacement mode is actually the specific form of determined interpolation
polynomials, which is also the trial function expressed with the function value at the
nodes.
Now we’ll use the linear displacement mode. That is to represent the displacement
using the step-linear-functions. And we only need to write the expression of
displacement trial function in each element.
4.4 Mechanical examples
Displacement mode
Suppose the displacement ui-1 and ui at the two ends xi-1 and xi of the element ei :[xi-1, xi] are given.
Because the displacement of element ei is linear and according to the two-point form of linear equation
of analytic geometry we can get the expression of u(x) in the element ei as
xi  x x  xi 1
u ( x)  ui 1  ui ( xi 1  x  xi )
xi  xi 1 xi  xi 1
Adopt the local number of nodes and
y N1(i ) ( x) N 2(i ) ( x)
xi  x x  xi 1
N (i )
 N (i )
 1
xi  xi 1 xi  xi 1
1 2

Apparently the two functions have the following properties: ei x


1.N1(x) and N2(x) are
2. N1(xi-1)=1, N1(xi)=0, N2(xi-1)=0, N2(xi)=1 xi 1 xi
The two functions are linear basis function of interpolation as shown in the picture.
The effects are apparent. The displacement of the element can be expressed by the linear combination

u ( x)  ui 1 N1(i ) ( x)  ui N 2(i ) ( x) 或 u ( x)   N  ei


of them.

N   N1(i ) N 2( i )   ei  ui 1 ( x) ui ( x)


Where T
4.4 Mechanical examples
 Displacement mode
In element ei :the strain in [xi-1, xi] can be expressed as εi
du dN1(i ) dN 2(i )
i   ui 1  ui  [ B] ei
dx dx dx
where
dN1(i ) dN 2(i ) 1 1
B [ ]  [ ]
dx dx Li Li
The conclusion is the same for the virtual displacement {δ*}ei .

 Element analysis
The equation of the virtual work can be written in the form of discretization
after determining the displacement mode.
Rewrite the global integral of the virtual work equation into a partitioning
integral n 1 n 1

 EA
i 0
ei
u ( x) ( x)dx   A f  ( x)dx  P ( L)
i 0
ei
where f=ρg.
4.4 Mechanical examples
 Element analysis
Substitute the according determined displacement mode into the left part of
the integral of every element ei. And we have
EA u( x) ( x)dx  EA  ( x)  u( x)dx
T
ei ei

 EA ([ B]{ *}ei )T ([ B]{ }ei )dx


ei

 { *}Tei  [ B]T EA[ B] dx{ }ei  { *}Tei [ k ]ei { }ei


ei

 
where [k ]ei   [ B]T EA[ B] dx is called the element stiffness matrix.
ei

Similarly substitute into the right part of the equation


A  ( x) fdx  A  T fdx  A [ N ]{ *}ei  
T
fdx
ei ei ei

 { *}Tei  A[ N ] fdx  { *}Tei {F }ei


ei

where {F }ei  
ei
A[ N ] fdx is called the equivalent force of element nodes.
4.4 Mechanical examples
 The assembling of the global stiffness matrix and the global load vector
Assemble every element of the element stiffness matrix and the element load vector
according to its global number and we can get the global stiffness matrix and the global
load vector.
We can find the following properties of the global stiffness matrix:
1. Symmetric matrix
2. Sparse matrix
3. Nonnegative definite matrix

 Constraint handling
Introduce the boundary condition by “put 1 on the diagonal”.

 Solving equation
The coefficient matrixes of the linear equations obtained by the finite element method
have the properties of symmetry and sparsity but usually very high order. So we
usually use the computer programming to solve
 Further analysis
After solving the displacement we can obtain the parameters such like strain, stress
and the support reverse force by further calculation.
4.5 Example analysis
Question 1 As shown in the picture, apply an axial load P on the
right end of the first-step shaft part.
A1 E1 AE 2 2

p
1 2
L L
6 6
: A  2.0 10 m
1 2
A  1.0 10 m
2 2

E  E  E  2.0  10 Pa
1 2 11

P  100N L  L  1.0m
1 2
4.5 Example analysis
Solve:(1) Structure discretization
Deformation analysis: There is only axial deformation with the same
displacement at every section. Suppose there is one axial displacement at every
node. And express the element with axial deformation by line segment.

u1 u2 u3

1 P1 1 2 P2 2 3 P3

1、2、3 ——node number 1 、2 ——element number


u1 、 u 2 、u 3 ——The number of freedom degree of nodes.

P1 、 P2 、P3 ——The number of load of nodes.

Known quantities: u1 、P2 、P3 Unknown quantities:P1 、u 2 、


u3
4.5 Example analysis

(2) Determine the interpolation function of the displacement.


Arbitrarily select an element of the structure

u1 e
u x  e
u2
x
i j

 system from i to j at the element where


Establish a local reference
i is the original point. ( ij is the positive direction)

i、j —— Number of element endpoints

u1e 、u 2e —— Displacement number of element endpoints.


4.5 Example analysis
Let the axial displacement at an arbitrary point x (section) in the element be u x 
Suppose: ux   a  bx (1)
According to the displacement continuity:

At point i u  x i   u1 e
a  u1e (2)

u x j   
e e
At point j  u2 ae
bL u 2 (3)

Le —— element length
Solve the simultaneous equations(2)(3)
u 2e  u1e
Thus: b  (4)
Le
Substitute(2)and(4)back to(1)and we have

u 2e  u1e e e
u x   u1 
e
e
x Incorporate with identical indicators u1 u 2
L
4.5 Example analysis
 x  e x e
And u x    1  e  u1  e u 2
 L  L
Written in the form of matrix

x  1
e
 x u 
u x   1  e , e   e 
 L L  u 2 

Let u  ux 
 x x 
N   1  e , e  —— Shape function matrix
 L L 

 1
u e

 
u e
  e  —— Displacement vector of element nodes
u 2 
 
So   
u   N  u  
e
4.5 Example analysis
(3)Element analysis
Calculate the elastic strain energy of the element
1 1 e L
U   dv  A 0 dx
e
e

2V e
2
u(x )  
 
x
  N
x  
 u e
    N
x  1
, N 
2
u  
e

1  1
 
 1
e e
 1 u  u 
  e , e   e   
 B   e
 L L  u 2 
 u 2 
 
1 u1   
e e
 1  u 1 
  E   E   e , e   e   E B   e 
e

 L L  u 2 
 u 2 
 
4.5 Example analysis
 
u1
e


 u  B 
1 e Le T
U e
 A e
1
e
,u 2 E e

B 
  e  dx
2 u 2
0
 

1 e e   

e
u
 1
  B 
T
 A L  u1e ,u 2e E e
 B 
   e 
2  
 u 2
 

u
 1
e


1
u  A 
T
 e
1
,u 2 e e e
LE e

B 
 
B 
  e 
2 u 2
 

Let K e   Ae Le E e B 
T
B 
 1
  e  1  1
 
e e
L 1 1 A E
A LE 
e e e
 
1   Le Le   1 1 
, e
 e  L  
 L 
称单元刚度矩阵
4.5 Example analysis
So the elastic strain energy of the element can be expressed as
1
   
T
U e
 ue K e  u e
2  

Where

 1
u e
 —— Displacement vector of the element nodes.
 
u e
  e
u 2 
 
   
T
u e
 u1e ,u 2e

u1 
u 
1
  
u 2 
 
 1  1  4.0  105  4.0  105 
 
K 
1 A1E 1
 1 1    5 
L1     4.0  105
4.0  10 
4.5 Example analysis

As for element 2

u 2 
u 
2
  
u 3 

 1  1  2.0  105  2.0  105 


K 
2

A2 E 2
 1 1    
L2   2.0  10 2.0  105 
5

4.5 Example analysis

U  U U
1 2

1  4.0  105 4.0  105  


u1 
 u1 ,u 2    5   
2  4.0  10 4.0  10  u 2 
5
 
1  2.0  105 2.0  105   u 2 

 u 2 ,u 3    5   
2  2.0  10 2.0  10   u 3 
5

 4.0  105 4.0  105 0  u1 
1   
 2
u1 ,u 2 ,u 3    4.0  105
4.0  10 5
0  u 2 
 0 0 0  u 
   3
0 0 0  u1 
1    
 u1 ,u 2 ,u 3   0 2.0  105 2.0  105  u 2 
2  
0 2.0  105 2.0  105 
u 3 
4.5 Example analysis

 4.0  105 4.0  105 0  u1 


1    
U 
2
u1 ,u 2 ,u 3   4.0  105
6.0  105 2.0  105  u 2 
 0 2.0  105 2.0  105   
 u 3 
1
u  K  u 
T

2

The virtual work of the external force


 P1 
 
Wp  P  u  u1 ,u 2 ,u 3  P2 
 
P3 

P1 —Unknown P2  0 P3  100 N
4.5 Example analysis
The potential energy of the structure is

 4.0  105 4.0  105 0  u1 


1  5  

2
u1 ,u 2 ,u 3    4.0  105 6.0  105 2.0  10  u 2 
 0 2.0  105 2.0  105   
 u 3 
 P1 
 
 u1 ,u 2 ,u 3   0 
100 
 
Substitute the extremum condition of the functional
0 
   
 0 Or  0 
ui  u1 ,u 2 ,u 3  0 
 
4.5 Example analysis
So
 4.0  105 4.0  105 0  u1   P1  0 
 5      
 4.0  105
6.0  105 2.0  10  u 2    0   0 
 0 2.0  105 2.0  105    100  0 
 u
 3    

Transpose the terms and


 4.0  105 4.0  105 0  u1   P1 
 5    
 4.0  105
6.0  105 2.0  10  u 2    0 
 0 2.0  105 2.0  105    100 
 u 3   
简记为
  u  
K  P 
—— The approximate equilibrium equation of the structure
4.5 Example analysis
(5)Constraint handling
Purpose:Introduce the boundary conditions to eliminate the rigid displacement and
ensure the unique solution of the equation.
Solution:Substitute the known displacement
 4.0  105 4.0  105 0  u1 
1    
U 2
u1 ,u 2 ,u 3    4.0  105 6.0  105 2.0  105  u 2 
 0 2.0  105 2.0  105   
 u 3 
Because u1  0
K  The first line and first row respectively multiply zero and
can be eliminated from the equation
 6.0 105  2.0 105 
K  Can be simplified as  5 
  2.0  105
2.0 10 
4.5 Example analysis
Reduce the order of the equation to
 6.0  105 2.0  10  u 2 
5
 0 
 5      
 2.0  105
2.0  10  u 3  100 
(6)Solve the equation”
3 3
u 2  0.25  10 m u 3  0.75  10 m
Eliminating the line and row of the matrix is not quite convenient in practical analysis.
So we usually put 1 in the diagonal and 0 in the other places.
Than the equation can be transformed into
1 0 0  u1   0 
 5    
 0 6.0  10 5
 2.0  10 u
  2   0 
0 2.0  105 2.0  105  u  100 
 
 3
The two forms are equivalent but the latter one is more convenient when solved by
computer.
This step usually requires a lot of large-scale algebraic equations calculation, which can
only be solved by computers.
4.5 Example analysis
(7)Solve the element stress
Element 1
 1 1 
u1   
 1
 
B   
 u
e
  1 , 1 
 L L 
    1,1
 
0
3 
0.25  10 
u 2 
 
 0.25 103
 1  E1 1  2.0 1011  0.25 103
 0.5  108 ( Pa )
Element 2
 1 1  u 2 
 0.25  103 
 2
 
B   
 u
e
  2 , 2     
L L  1,1
  3 
 u 3 
   0.75  10 
 0.5 103
 2  E 2 2  2.0 1011  0.5 103
 1.0 108 ( Pa )
V. Conclusion
5.1 Conclusion
So this is the mathematical basis of the finite element method and the steps to
solve the one dimensional problems.

In terms of mathematics the finite element method can be regarded as the


combination of the classical variation method and the partitioned polynomial
interpolation.
1. It adopts the discretization and the flexibility of variation method and create
the finite subspace on basis of the principle of variation.
2. It then turns the variation problems in the infinite dimensional function
space of the differential equations into the extremum problems of
multivariable functions in the finite element subspace. The unknown
quantities of the multivariable functions are the value of the undetermined
functions at the disperse nodes.
3. Solve the governing equations according to the extreme value theory of the
multivariable functions, which is also called the linear algebraic equations.
The solutions of the algebraic equations are the numerical solution of the
undetermined functions at the disperse nodes.
It should be noticed that the discussion of the theoretical basis is far from
perfect such like proving the convergence and the neglect of the error
estimation. These problems must be solved before widely used.
5.1 Conclusion
In fact since the finite element method developed into the climax of
theoretical research in 1960s, more and more mathematicians have been
studying to free the finite element method from the restraint of the
engineering problems, provide a unified opinion as well as a more precise
mathematical description and finally establish a systematic mathematical
basis after combining with the functional analysis, the variation method, the
difference method, the approximation theory of functions and so on.

The revolution of the finite element method is driven by the emergence of


computer and the application of functional, of which the former is the
calculation tool and the latter is the theoretical basis.

You can refer to the following books


1 有限元法的数学基础 王烈衡 许学军 编著
2 有限元方法及其理论基础 姜礼尚 庞之垣 著
3 有限元方法及其应用 李开泰 黄艾香 黄庆怀 编著

So the part of theoretical basis has come to the end.


We’ll discuss the derivation, basic form, numerical method and the
practical application of the finite element method in solid mechanics.
谢谢!

You might also like