Scribd
Upload
164 views

Chapter 1 Mathematical Modelling by Differential Equations: Du DX

This document introduces mathematical modeling using differential equations. Physical phenomena can be described by mathematical models involving unknown functions and derivatives, leading to differential equations. These equations are important for engineering, medical, and other applied problems. The goal is to learn how to build models and solve differential equations analytically and numerically. Examples show how displacement, heat transfer, and image processing problems can be modeled as differential equations by relating physical laws to functions, derivatives, and boundary conditions. Solving the examples demonstrates both exact analytic solutions and the need for numerical approximations in practice.

Uploaded by

Kan Samuel
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
164 views

Chapter 1 Mathematical Modelling by Differential Equations: Du DX

This document introduces mathematical modeling using differential equations. Physical phenomena can be described by mathematical models involving unknown functions and derivatives, leading to differential equations. These equations are important for engineering, medical, and other applied problems. The goal is to learn how to build models and solve differential equations analytically and numerically. Examples show how displacement, heat transfer, and image processing problems can be modeled as differential equations by relating physical laws to functions, derivatives, and boundary conditions. Solving the examples demonstrates both exact analytic solutions and the need for numerical approximations in practice.

Uploaded by

Kan Samuel
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Chapter 1 Mathematical modelling by Differential Equations

Many physical phenomenons are goverened by physical rules (such as physical laws, e.g. Force and acceleration in Newton mechanic). These physical rules can be converted into mathematical models. It in turns
leads to Differential equations, which are equations involving unknown functions and derivatives. Besides,
many practical problems in engineering and medical research are related to differential equations and system
of linear or non-linear equations. For example, in image processing, many problems can be solved by Partial
Differential equations (PDEs). Developing methods to solve these equations are therefore very important.

The goal of this course:


1. Learn how to build mathematical models for describing physical phenomenons or solving practical
pronlems (It will be briefly mentioned. Details will be covered in MATH 3290: Mathematical modeling).
2. Solve equations analytically, that is, obtaining exact solution.
3. Solve equations numerically, that is, approximating the solution
In this chapter, we will briefly introduce how we can model problems mathematically by differential equation.

Example 1: Mathematical modeling for elastic bar


Our goal is to model the displacement u(x) of the elastic bar at
each position x under gravity. In order to model the displacement,
we need to understand the physical laws governing the physical phenomenon.
Under gravity, each point x is displaced down to x + u(x), where u(x)
is the displacement. In physics, the derivative of u(x) is called strain. In
other words, du
dx is the strain.
P hysical laws: In physics, it is observed (by experiment) that the internal
force (x) if the stretching is proportional to the strain. That is,
(x) = c(x)
Figure 1:

du(x)
dx

c(x) is determined by the elastic material.

Consider a small portion of the bar:


At equilibrium,




du
du
c(x)
c(x)
+ (4xa) g = 0
dx x+4x
dx x
where: is the density, a is the cross section area
and hence (4xa)is the mass.
Deriving both sides by 4x and let 4x 0, we get


d
du
()
c(x)
= f (x) := ag(x)
dx
dx

Figure 2:
1

() is the differential equation governing the displacement of the bar under gravity.

Boundary condition :
In order to obtain the (unique) solution, only the differential equation () is not enough. We further need
the boundary condition. Thats, some special requirements defined on the boundaries.

There are two types of boundary conditions:

1. (Dirichlet) u(0) = 0 and u(1) = 0 (Fixing two ends of the bar)


2. (Dirichlet and Neumann) u(0) = 0 (Fix the top end) and
(1) = c(x) du
dx |x=1 = 0 (other end is in the air, and hence no internal force)
Consider () under (1)
Question: Solve
d

dx

du
c(x)
dx

= f (x) := x3 ,

u(0) = 1, u(1) = 0, c(x) = 1

Solution

dx

du
c(x)
dx

dx
x
du

dx 0

du
dx
x
du
dx
0 dx
x

u(x)

u(1) = 0 = C0 =
5
x
u(x) = x20 + 20

1
20

=
=
=
=
=

f (x)dx
0

4 x

x
4 0

(boundary condition on x = 0 is needed)

x4
+ C0
4
x
x4
dx + C0 x
4
0

5 x
x
+ C0 x
20 0

(boundary condition on x = 1 is needed)

Consider () under (2)


Question: Solve

d
dx



du
c(x)
= f (x) := x2 + 1,
dx

u(0) = 0, c(x)


du
= 0,
dx x=1

c(x) = 1

Solution:

dx



du
c(x)
dx
dx
1
du
c(x)
dx

u(x)

(x2 + 1)dx
x

du
dx
du
dx

=
=
=

1

x3
+ x
3
x

(boundary condition on x = 1 needed)

4 x3

x
3
3
x
4 x3

xdx (boundary condition on x = 0 needed)


3
0 3
x4
x2
4
x

3
12
2

Remark:
In both case, 2 boundary conditions (2 ends) are needed !
In both cases, exact (analytic) solution can be found.
In practice, analytic solution cannot be obtained need numerical approximation
Example 2: Math modeling for heat conduction
Consider:
Let:
f (x, t)

= strength of heat source at position x at time t

u(x, t)

= temperature at position x at time t

Math modelling:

t
x



u
c(x)
=
x

f (x, t),

0 < x < 1,

t0

where c(x) is the heat conductivity and this equation is called the Heat conduction equation.
Again, we need boundary conditions to solve the equation. We have 3 boundaries:
x = 0,

x = 1 and t = 0

For t = 0, set u(x, 0) = u0 (x) for 0 x 1 (assume the initial temperature is known)
For x = 0, set u(0, t) = 0, t 0 (assume one end (x = 0) is placed at ice basin)
For x = 1, set

u
c(x)
=0
x x=1
3

(assuming other end is insulated, no heat flux pass through it)


Thus, heat transfer can be modeled as:

u
u

t x c(x) x = f (x, t)

u(x, 0) = u0 (x) 0 x 1
u(0, t) = 0 t 0


=0 t0
c(x) u
x

(Partial Differential Equation)

x=1

Example 3: Heat equation


The variation of temperature in space generates heat flux:
~q = ku

(k > 0 = heat conductivity)

Also, heat (a form of energy) related to temperature:


h(~x, t) = cv u(~x, t),

cv = const = specific heat

Consider any subdomain , now, by conservation of energy:


Rate of change of energy
Heat transfer into through
=
in in t
boundary per unit time
Mathematically,
d
dt

h(~v , t)dx

~q (~n)dS

(~n = outward normal of )


Hence, we have

d
cv u(x, t)dx
dt

cv ut (x, t)dx

~q (~n)dS

~qdx

Recall: Divergence theorem


Let ~v =vector field in R3 (or R2 ), is volumn (or 2D domain). Then:

~v d~x =
~v ~ndS

where ~n is the outward and ~v :=


Hence, we have

v1
x1

Omega
v2
x2

v3
x3 .

(cv ut + ~q) dx = 0 for all

Since is arbitrary, we have:


cv ut + ~q = 0
4

Now, ~q = ku implies:
cv ut + (ku) = 0
k
ut = 4u := K4u (heat equation)
cv
where K =

k
cv

=temperature conductivity.

Again, we need boundary conditions for heat equation:


Boundary condition for heat equation
On D,
Dirichlet
u(~x, t) = g(~x) for ~x D
(Domain D is attached to a heat bed with given temperature g(~x))
Neumann
u(~x) ~n = 0 on D
No heat flux on D domain is insulated.
Or
u(~v ) ~n = h(~x) on D
(inject energy into with rate h)
Robin
u(~x) ~n = (g(~x) u(~x, t))

on D,

> 0 is a constant

(Difference of u and its surrounding temperature generate a heat flux, linearly proportional (Newton
Condition law) to the difference).
On t = 0:
u(~x, 0) = u0 (~x),

xD

(Temperature = u0 (~x) at time t = 0)


Remark:
Heat conduction can be formulated as PDE.
Boundary condition is needed to determine solution.
We will learn how to solve it analytically and numerically.

Example 4: Image Processing

In image processing, one important problem is called image denoising. Due to transmission error or other
technical reasons, images that we receive are usually distorted by noises. For example, compare the original
image v and the noisy image v as shown above.
In order to restore the image, we can turn it into a mathematical problem. The basic idea is to look
for a smooth function u : D R approximating v.
In calculus, we know that a function f is not smooth if its derivative |f | is big. Hence, we can solve
the above image processing problem by minimizing the following functional:

E(u) =
|u|2 d + (u v)2 d
()

over all u : R with u| = h.


In order to minimize (), we consider the function G : R R defined by:
G(t) = E(u + tw)
where w : R is an arbitrary function on with w| = 0.
Since u is a minimizer, we see that G(t) attains its minimum at t = 0. Hence, G0 (0) = 0.
Now,
G0 (t)

=
=
=
=
=

d
E(u + tw)
dt 


d
2
2
|(u + tw)| d~x + ((u + tw) v) d~x
dt




d
2
((u + tw) (u + tw))d~x + ((u v) + tw) d~x
dt




d
2
2
(u u + 2tu w + t ww)d~x + ((u v) + tw) d~x
dt

(2u w + 2tw w)d~x +


2(u v + tw)wd~x

0 = G0 (0) =

d
E(u + tw)
t=0
dt

2u wd~x +

2(u v)wd~x

( )

(1)

Recall: Integration by part formula in high-dimension


Let u : R , ~v = (v1 , v2 ) = vector function on R2 . Then:

u ~v =
u(~v ~n)dS
u ~v d~x

where ~n is the outward normal to and ~v is the divergence of ~v .


From (***), we have:

w(u ~n)dS

0=2

Note that w = 0 on .

w (u)d~x + 2

(u v)w

( (u) + 2(u v)) wd~x

0=

for all w

We conclude that u must satisfy the following partial differential equation:


(u) + 2(u v) = 0
with boundary condition: u| = 0
Conclusion:
A lot of physical phenomenons or practical problems can be converted to solving differential equations. It
is therefore important to develop mathematical methods to solve these equation both exactly or at least
approximately.
In this course, we will look at both analytic and numerical methods to solve differential equations.

Exercise
1. Solve

2. Solve

dx

d
dx



du
c(x)
= x3 ,
dx

du
c(x)
dx

=e ,

u(1) = 1, u(2) = 3, c(x) = x


du
u(1) = 1, c(x)
= ex , c(x) = x + 1
dx x=1

You might also like