Professional Documents
Culture Documents
Preface
2 | Page
named after the class of problems, such as Legendre equations,
Bessel equations. Recall that a standard solutions has been found
to theses equations.
Although this classification may help to solved similar problem,
no general approach is attempted to solve all types of ODE and
PDE . Recall that all types of ODE and PDE are expressed generally
as partial differential of space dependent variable to be determined
and dependent variables (x,y,z) and time (t) in computational
domain and its boundary. New recent insight of these governing
equation stated that we can extractessental information from the
governing equation by an operator. Operator can be define as a
mathematical procedure to the governing equations such as
differentiation, integration, eigen-value of the matrix coefficients of
equations, etc. Depending on the what is to be known from those
equations the operator maybe different from one to other. Yet, the
basic operator is differentiation and integration.
First we define what is an operator and inversed-operator.
An (any) operator is a single fixed mathematical
procedure applied to a certain mathematical entity (scalar,
vector, tensor, matrix or array of any mathematical
expression or function). The mathematical operation such as
addition, substraion, divission, eigen values, differentioan
and intergration are valid operators. Example of a simple
operator is
L= gdx
L= ( .) dx .
L1=
d
( . ) dx=(.)
dx
where x in this
) and
3 | Page
series form accepted as solution series, altough it may be
truncated to a few terms.
2.
One dimensional
method
ADOMIAN
decomposition
u ( x , y )= u n ( x , y )=u0 +u1 +u
n=0
(2.1)
Where the components are
un ( x , y ) ,n 0
(2.2)
4 | Page
( ( ))
u=f L ( Ru )=f L1 R
un
n =0
(2.4)
Now, the Adomian solution is expressed as
u= un
(2.1)
n=0
(2.5)
u0=f
operator L-1)
uk +1=L1 ( R ( uk ) ) , k 0,
Or equivalently
u0=f ,
u1=L1 ( R ( u 0 ) )
u2=L1 ( R ( u1 ) )
u3=L1 ( R ( u2 ) )
Example 1.
(2.6b)
5 | Page
Consider equation
u n( x)= A+ L un ( x )
1
n=0
n=0
u0 (x )=A
u1 ( x )=L
( uk ( x ) ) , k 0,
which
can
recursion as
u0 (x )=A ,
(2.7a)
u1 ( x )=L
( u0 ( x ) )= Ax ,
(2.7b)
u2 ( x )= L1 ( u1 ( x )) =L1 ( Ax )=A
x
,
2!
(2.7c)
x2
x3
=A ,
2!
3!
( )
u3 ( x ) =L1 ( u2 ( x ) )=L1 A
...
(2.7d))
be
expanded
by
6 | Page
2
u ( x ) =A (1+ x +
x x
+ + )=A e x
2 ! 3!
(2.8)
Example 2.
Using the same linear differential equation, Lu=xu, but now we
d2
L= 2 .
dx
assign
This
example
illutrate
the
Adomian
L=
d2
dx 2
(.) as ????
x
L1 ( . )= (. ) dxdy
0 0
(2.9)
The RHS of eq. (2.9) is called double or folded integral, but their
upper limits are the same x.
Operates the operater L-1 to both sides of Lu=xu, we obtain
(2.10)
un( x )
n=0
or
= A + B+ L (x ( u 0 ))
0
1
7 | Page
u0 ( x ) =A + Bx ,
3
u1 ( x )=L1 ( x u 0 ( x ) )= A
x
x
+B ,
6
12
u2 ( x )= L1 ( x u 1 ( x ) )= A
x
x
+B
,
180
504
(2.11)
) (
x
x
x
x
u ( x )= A 1+ +
+ + B 1+ +
6 180
12 504
(2.12)
Example 3.
Using the same technique we apply Adomian decomposition
procedure to solve linear first order partial differential equation
(PDE):
u x +u y =f ( x , y ) , u ( 0, y ) =g ( y ) , u ( x , 0 ) =h(x )
(2.13)
Lx u+ L y u=f ( x , y )
where operator L x =
(2.14)
,L y =
x
y .
(2.15)
8 | Page
PDE of any order. This implies that no all PDE should be treated in
this manner, since it added burden to your computation! The
modification will be outlined in this sub-chapter.
Consider a any order of PDE presented as,
Lu+ Ru=g
(2.16)
u=f L1 R(u)
(2.17)
Where f represent the terms arising from two conditions: (1) the
given initial condition (t=0), and (2) result integration of source
term g. The Adomian solution of (2.17) is
u= un
(2.18)
n=0
u0=f ,
1
uk +1=L ( R uk ) , k 0
(2.19)
You should be already familiar with (2.19) with is the Adomian
decompositions method. Now, Wazwaz introduces a sligth
variation into this, aiming at faster acceleration to convergence.
For specific cases you may encountering, the function may be
decomposed into
f =f 1+ f 2
u0=f 1 ,
u1=f 2L1 ( R u 0 )
(2.20)
9 | Page
uk +1=L1 ( R uk ) , k 1 ,
(2.21)
Obviously only the first and second components are changed.
However, this change reveals siginificant acceleration to converge!
However, to select appropiate
f1
and
f2
u x +u y =3 x2 y 3 +3 x3 y 2 ,u ( 0, y )=0
(2.23)
Solution
In operator form eq.(2.23) becomes
Lx u=3 x 2 y 2 +3 x 3 y 2u y
(2.24)
L1
x
we obtain
3
u ( x , y )=x 3 y 2+ x 4 y 2L1
x ( uy )
4
(2.25)
f 1 ( x , y )=x 3 y 3 ,
3
f 2 ( x , y )= x 4 2 ,
4
(2.26)
u0 ( x , y )=x 3 y 3
3
u1 ( x , y )= x 4 y 2L1
x (u0)y ,
4
(2.27)
10 | P a g e
uk +1 ( x , y ) =L1
x ( uk ) y , k 1.
Solving eq. (2.27) we obtain
u0 ( x , y )=x 3 y 3
3 4 2 1
3 2
u1 ( x , y )= x y Lx ( 3 x y ) y =0,
4
(2.28)
uk +1 ( x , y ) =0, k 1.
Thus the solution is
u ( x , y )=x y
(2.29)
This example demonstrate that the solution reduces to two to two
iteration for computing the convergence point (X,Y). What is the
value of X,Y?
Example 2. Use the modified decomposition method to to solved
the first oder PDE
u x +u y =u , u ( 0, y )=1+ e y .
(2.30)
y
uu
y
1
u ( x , y )=1+e + Lx
f 1 ( x , y )=e
(2.31)
f =f 1+ f 2
, where
f 2 ( x , y )=1.
(2.32)
The recursive modified can now be applied
u ( x , y )=e y ,
u 0(u0) y,
u1 ( x , y )=1+ L1
x
uk ( uk ) y , k 1.
uk +1 ( x , y ) =L1
x
Expanding eq. (2.33) gives
(2.33)
11 | P a g e
u0 ( x , y )=e y ,
u1 ( x , y )=1,
u2 ( x , y )=x ,
u3 ( x , y )=
x2
2!
uk +1 ( x , y ) =
xk
k ! , and so on.
(2.34).
In close form the exact solution is
u ( x , y )=e x + e y .
(2.35)
Exercises
1. Use the modified decomposition method to solve first order PDE
12 | P a g e
3. Heat Flow
3.1. Introduction
Heat flow is the one of the simple model of flow. It require
media (gas, liquid or solid) to flow from one region of space to
other. Recall that physical phenomena which is related to
engineering are expressed by partial differential equation (PDE)
of space and time which is called spatial independent variables
(x,y,z) and time variable. In general we can focus on fixed
spatial coordinates (x,y,z) and observe the change of
momentum and energy on that point. This is called Eulerian
view. Alternately, we can follow the particle motion and
observe the momentum and energy change. The last option is
called Lagrangian view. Depending on the problem, both are
appropiately use. If the solution based on Euler view point is
found, we can always plot the pattern of solutions at each point
in space, where we deduct the Lagranges solution. The
opposite is also correct.
3.2 One Dimensional (1-D) Heat Flow
The solution of dependent u(x), where x is a single independent
variable. Special solution demand that the solution of PDE
should be valid at the boundary. The problem is called
Boundary Value Problem (BVP). Note that the boudary does
not change with time. Otherwise we are dealing with unsteady
condition. The latter is detemined by initial (t=0) condition. The
latter is called Initial Value Problem (IVP). If both conditions
are prescribed it called Initial-Boundary-Value-Problem
(IBVP).
It should be remember that heat-flow belongs to linear
PDE, and therefore superposition of one solution to other is
valid. Recall that we cannot superimposed special solution with
13 | P a g e
other special solution if PDE is non-linear. The traditional
method you learn in PDE is separation variables of pure spatial
variables and pure time variables. In this chapter we will
focused on Adomian method. Obviously the Adomian
decomposition method should converge with separation
variblles method to exact solution. If you plot the evolution of
the separation method and Adomian decomposition method
both plots will coincide.
The propagation of heat in a rod from one end to the
other end (Figure 3.1) will be our case.
1. The governing Partial Differential Equation (PDE) is
ut
Where
k u xx , 0 x L, t 0
u u x ,t
(3.1)
u xx =
(u)
2
x
u ( 0,t )=0,t 0 .
(3.2)
u ( L, t )=0, t 0.
(3.3)
Eq. (3.2) and (3.3) indicate that the temperature at the end of
the rod are kept constant at zer0 degree Celcius. Recall that
this is Neumann BC (see APPENDIX A).
3. Initial Condition (IC). The initial condition related to
Neumann BC prescribes the t temperatrure over
computational domain (x), either as random distribution or
regular function such as trigonometric, polygonal, or
hypergeometric. Mathetical expression of IC is
14 | P a g e
u ( x , 0 )=f ( x ) , 0 x L
(3.4)
u ( 0,t )=0,t 0,
BC
(3.5)
u ( L, t )=0, t 0,
IC
u ( x , 0 )=f ( x ) , 0 x L
If you write the PDE problem like this, the mathematician say a
well posed problem. In general a well posed problem has one
distinct solution, depicted as particular solution, which is the
central theme of this chapter.
It is of interest to see that eq. (3.5) is a homogeneous
equation as it is. Physically means the the rod is perfectly
thermally insulated, such that zero heat loss. On the contrary
zero thermal insultation do not exist, and some heat loss has to
be accounted for. Note that Adomian decomposition method
treat eiher homogeneous PDE in elegantly straight manner
(using recursion) without any need of transformation of
inhomogeous to homogeneous one as in separation varirables
method. Both will be demonstrated in a moment.
1. Homogeneous Heat Equation: g = 0
The governing equation (3.5) is repeated for convenient
represent a homogeneous heat equation conduction along
perfectly insulated conductor example.
PDE
u ( 0,t )=0,t 0,
BC
(3.6)
u ( L, t )=0, t 0,
u ( x , 0 )=f ( x ) , 0 x L
IC
Mind that
k u xx
that g 0.
15 | P a g e
PDE
u ( 0,t )=0,t 0,
BC
(3.7)
u ( L, t )=0, t 0,
u ( x , 0 )=f ( x ) , 0 x L
IC
PDE
u ( 0,t )=0,t 0,
BC
(3.5)
u ( L, t )=0, t 0,
u ( x , 0 )=f ( x ) , 0 x L
IC
Lt u ( x , t ) =Lx u (x , t)
(3.7)
Where
Lt =
and
Lx =
2
2
x
Lt
and
Lx
1
L1
t ( . ) = ( . ) dt , L x ( . ) = ( . ) dxdx
0
0 0
(3.8)
16 | P a g e
Thus
L1
t Lt ( . )=u ( x ,t )u (x , 0)
Applying
L1
t
(3.9)
and using initial
condition we obtain
1
u ( x , t )=f ( x ) + Lt ( L x u ( x , t ) ) .
(3.10)
u ( x , t ) = un ( x , t ) ,
(3.11)
n=0
u0 ( x , t ) , u1 ( x , t ) , u2 ( x , t ) ,
are to be
x ,t
un ()
n=0
Lx
(3.12)
u n( x , t )=f ( x ) +
n=0
Or equivalently
L x ( u 0+u 1+u 2
(
u0 +u1 +u2 +=f ( x ) +L1
t
###@@@!!!
(3.13)
17 | P a g e
4. Wave Equation: elastic wave, water wave, sound, earthquacke and electro-magnetic wave
These are waves in macroscopic world where the speed of
propagation can be measured directly or indirectly. Macroscopic
means that we can sense with our five senses: hearing, seeing
(e.g. cold and hot), feeling (e.g touching), smelling, and color
sense. Usually the physical wave is transformed to electric wave,
usually voltage with can be mesured by voltage device. Recall that
waves is essentially an oscillation phenomenon. If the oscillator is
periodic constant we called this harmonics mathematically can be
repesented by trigometric series, and therefore Fourier series can
be used as an approximation of any wave-shape, i.e. nontrigonometric such as square waves, etc. Elastic waves emittedn
by hard rgid solid matter of finite size when they are disturbed by
impulsive external force, priodic or non-periodic. Elastic means that
deflection disturbance return elastically to original shape (1-D,2-D,
or 3-D shape), otherwise, i.e. if the deformation due to disturbance
is permanent we call it non-elastic. Elastic deformation of one
object due to collision with other object does not mean that all
energy is return to the collidings objects. In fact some energy is
lost at colli
Periodic forces caused the harmonics forced vibration. Finite size
bodies are wire of finite length and diameter supported at both or
single end, any bulk size of body such as asteroid or man made
satelites, supported Civil Engineering structure such as buildings,
towers and bridges, rocks and sediment (cohesive like clay, or
granular like decomposed rock soil), and membrane (skin)
structure, Machinary and Heavy equipment Engineering, etc. Water
wave such as tide, surge, wind-waves, flash-flood, tsunami.
Sound wave such as explosion wave, gust, Mach-wave in
atmosphere and water. Earth-quacke waves propagate within the
interior of solid rock and sediment cover due to local failure of
strain-stress propagate uniformly in all direction. Electro-magnetic
waves sources in form of sunlights and artificialy as radio-waves
propagate as EM waves with the speed of light and doesnot needs
any kind of propagation medium as other waves. The speed of
wave propagation of non EM waves depend on the medium. The
impulsive force works on finite time on the body caused the body
vibrates with unique natural frequency. However if the body is a
18 | P a g e
system comprising of many discrete components, the body will
vibrates non-linearly where each local vibration will be influenced
by others. The mathematical model of this kind, the non-linear
elastic vibration of structure has been not attemped so far, and
vibration scientist usually hypothesized that a complex structure
may be modelled as multi-degree of freedom (MDOF) linear
structure, and apllied superposition. Still this should be accepted
with caveat, since supperposition means that the
vibration
sampling is the same independent where you take measurement,
while
the frequency is physically localized. Each node point
vibrates as superposition of neighboring nodes. Therefore it
depend on how the structure is assembled from discrete linear or
non-linear buliding block while the building itself may be linear or
non linear.
5. Laplace Equation
6. Non-Linear PDE
7. Linear and Non-linear Physical Problem
8. Soliton and Compacton
9. Solitary Waves: Tsunami
10.Summary