ENGG2013 Lecture 24 - Chinese University of Hong Kong

Download Report

Transcript ENGG2013 Lecture 24 - Chinese University of Hong Kong

ENGG2013 Unit 24
Linear DE and Applications
Apr, 2011.
Outline
• Method of separating variable
• Method of integrating factor
• System of linear and first-order differential
equations
– Graphical method using phase plane
kshum
2
Nomenclatures
• “First-order”: only the first derivative is involved.
• “Autonomous”: the independent variable does
not appear in the DE
• “Linear”:
– “Homogeneous”
– “Non-homogeneous”
c(t) not identically zero
kshum
3
Separable DE
• “Separable”: A first-order DE is called separable if
it can be written in the following form
• Examples
–
–
–
–
–
x’ = cos(t)
x’ = x+1
x’ = t2sin(x)
t x’ = x2–1
All linear homogeneous DE
kshum
4
SEPARABLE DE AND
METHOD OF SEPARATING
VARIABLES
kshum
5
How to solve separable DE
• Write x’= f(x) g(t) as
.
• Separate variable x and t (move all “x” to the
LHS and all “t” to the RHS)
• Integrate both sides
kshum
6
Example
Solve
(1) Write the DE as
(2) Separate the variables
(3) Integrate both sides
General solution to x’=t/x
kshum
7
Solution curves
• The solutions are hyperbolae
Sample solutions
x ' = t/x
Some constant
4
3
2
x
1
0
-1
-2
-3
-4
-4
-3
-2
-1
0
1
2
3
4
t
kshum
8
Example: Newton’s law of cooling
• Suppose that the room temperature is Tr = 24
degree Celsius. The temperature of a can of
coffee is 15 oC at T=0 and rises to 16 oC after
one minute.
– T(0) = 15, T(1) = 16.
• Find the temperature after 10 minutes
Proportionality constant
kshum
9
LINEAR NON-HOMOGENEOUS DE
METHOD OF INTEGRATING FACTOR
kshum
10
Example: RC in series
• Physical laws
– Voltage drop across resistor = VR(t) = R I(t)
– Voltage drop across inductor = C VC(t) = Q(t)
Charge
Vc
C
From Kirchoff voltage law
VC(t) + VR(t) = sin(wt)
sin(wt)
R
Linear non-homogeneous
kshum
11
Linear DE in standard form
• Linear equation has the following form
• By dividing both sides by p(t), we can write
the differential equation in standard form
kshum
12
Product rule of differentiation
• Idea: Given a DE in standard form
Multiply both sides by some function u(t)
so that the product rule can be applied.
kshum
13
Illustrations
1. Solve the initial value problem
2. Find the general solution to
kshum
14
Example: Mixing problem
Water tank
1000 L
Initial Caesium concentration
= 1 Bq/L
• In-flow of water: 10 L per minute
• Out-flow of water: 10 L per minute
• In-flowing water contains Caesium with concentration
5 Bq/L
• Describe the concentration of Ce in the water tank as a
function of time.
kshum
15
http://en.wikipedia.org/wiki/Henri_Becquerel
Henri Becquerel
• French physicist
• Dec 1852 ~ Aug 1908
• Nobel prize laureate of Physics
in 1903 (together with Marie
Curie and Pierre Curie) for the
discovery of radioactivity.
• Bq is the SI unit for radioactivity
– Defined as the number of nucleus
decays per second.
kshum
16
Back to the RC example
• Write it in standard form
• Multiply by an unknown function u(t)
kshum
17
Integrating factor
• Is there any function u(t) such that
u’(t) = u(t)/RC ?
•  Choose u(t) = exp(t/RC)
kshum
18
Now we can integrate
Use a standard fact from calculus
kshum
19
Solution to RC in series
• General solution
• If it is known that Q(0) = 0, then
Steady-state solution
kshum
approaches zero
as t  
20
Sample solution curves
• Take R=C = 1, w=10 for example.
1
0.8
0.6
0.4
Q
0.2
0
-0.2
-0.4
-0.6
Different solutions
correspond to different initial
values.
-0.8
-1
0
1
2
3
4
5
6
7
8
t
Transient state
kshum
Steady state
21
SYSTEM OF DIFFERENTIAL
EQUATIONS
kshum
22
Interaction between components
• If we have two or more objects, each and they
interact with each other, we need a system of
differential equations.
• Metronomes synchronization
– http://www.youtube.com/watch?v=yysnkY4WHyM
• Double pendulum
– http://www.youtube.com/watch?v=pYPRnxS6uAw
kshum
23
General form of a system of linear
differential equation
• System variables: x1(t), x2(t), …, xn(t).
• A system of DE
Some functions
kshum
24
System of linear constant-coeff.
differential equations
• System variables: x1(t), x2(t), x3(t).
• Constant-coefficient linear DE
– aij are constants,
– g1(t), g2(t) and g3(t) are some function of t.
• Matrix form:
kshum
25
Application 1: Mixing
Water tank 1
f1
f12
Water tank 2
f2
Volume = V2 m3
Volume = V1 m3
Concentration = C1(t)
Concentration = C2(t)
f21
• C1(t) and C2(t) are concentrations of a substance, e.g. salt, in tank 1 and 2.
• Given
–
–
–
–
–
Initial concentrations C1(0) = a, C1(0) = b.
In-low to tank 1 = f1 m3/s, with concentration c.
Flow from tank 1 to tank 2 = f12 m3/s
Flow from tank 2 to tank 1 = f21 m3/s
Out-flow from tank 2 = f2 m3/s
• Objective: Find C1(t) and C2(t).
kshum
26
Modeling
•
•
•
•
Consider a short time interval [t, t+t]
C1 = C1(t+t)–C1(t) = cf1t + f21C2t – f12C1t
C2 = C2(t+t)–C2(t) = f12C1t – f21C2t – f2C2t
Take t  0, we have
C1’ = – f12C1 + f21C2+ cf1
C2’ = f12C1 – (f21+ f2) C2
kshum
27
Graphical method
• For autonomous system,
• we can plot the phase plane (aka phase portrait) to
understand the system qualitatively.
• Select a grid of points, and draw an arrow for each
point. The direction of each arrow is
kshum
28
Phase Plane
C1’ = – 6C1 + C2+ 10
C2’ = 6C1 – 6 C2
5
4.5
4
• Suppose
3.5
3
C2
– f1 = 5
– f2 = 5
– f12 = 6
– f21 = 1
–c=2
– Initial concentrations
are zero
2.5
2
1.5
1
0.5
0
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
C1
Converges to (2,2)
kshum
29
Convergence
C1’ = – 6C1 + C2+ 10
C2’ = 6C1 – 6 C2
• (C1,C2)=(2,2) is a critical point.
– C1’ and C2’ are both zero when C1= C2=2.
• The analyze the stability of critical point, we
usually make a change of coordinates and
move the critical point to the origin.
• Let x1 = C1–2, x2 = C2–2.
x1’ = – 6x1 + x2
x2’ = 6x1 – 6 x2
kshum
30
Phase plane of a system with
stable node
All arrows points towards
the origin
4
3
2
x2
1
0
-1
-2
-3
-4
-4
-2
0
x1
2
kshum
4
31
Sample solution curves
The origin is a stable node
4
3
2
x2
1
0
-1
-2
-3
-4
-4
-2
0
2
4
x
1
kshum
32
Theoretical explanation for
convergence
• The eigenvalues of the coefficient matrix
are negative. Indeed, they are equal to –3.5505
and –8.4495.
• The corresponding eigenvectors are
[0.3780 0.9258] and [–0.3780 0.9258]
kshum
33
Eigen-direction
•
If we start on any point in
the direction of the
eigenvectors, the system
converges to the critical
4
point in a straight line.
This is another geometric 3
interpretation of the
2
eigenvectors.
1
x2
•
0
-1
-2
-3
-4
-4
-2
0
2
4
x1
kshum
34
Application 2: RLC mesh circuit
• Suppose that the initial charge
at the capacity is Q0.
• Describe the currents in the
two loops after the switch is
closed.
i1(t)
i2(t)
Physical Laws
• Resistor: V=R i
• Inductor: V=L i’
• Capacitor: V=Q/C
• KVL, KCL
Homework exercise
kshum
35
An expanding system
• Both eigenvalues are positive.
kshum
36
Phase Plane of a system with
unstable node
The origin is an
unstable node.
The red arrows
indicate the
eigenvectors
5
4
3
2
y
1
0
-1
-2
-3
-4
-5
-5
-4
-3
-2
-1
0
1
2
3
4
5
x
kshum
37
A system with saddle point
• One eigenvalue is positive, and another
eigenvalue is negative
kshum
38
Phase Plane of a system with
origin is a saddle point.
saddle node The
The thick red arrows indicate
the eigenvectors
5
4
3
2
y
1
0
-1
-2
-3
-4
-5
-5
-4
-3
-2
-1
0
x
kshum
1
2
3
4
5
39
Conclusion
The convergence and stability of a system of
linear equations is intimately related to the
signs of eigenvalues.
kshum
40