Modeling with High-Order Differential Equations
National Chiao Tung University Chun-Jen Tsai 10/16/2019
Linear Models: IVP
Many linear dynamic systems can be represented using a 2nd order DE with constant coefficients:
In this formulation, g(t) is the input or forcing function of the system, the output of the system is a solution y(t) of the DE that satisfies the initial conditions
y(t0) = y0, y(t0) = y1 on an interval containing t0. )
0 (
2 1 2
2 a y g t
dt a dy dt
y
a d
Free Undamped Motion
Hooke’s law describes the restoring force:
F = ks
Newton’s 2nd law (F = ma) describes the motion:
m(d2x/dt2) = –k(s + x) + mg
= –kx + (mg – ks) = –kx
DE of free undamped motion:
Solution of the motion:
x(t) = c1 cos
t + c2 sin
t.1 0
2 2
2
) 0 ( ' , )
0 ( ,
0 x x x x
dt x x
d
equilibrium position
s l
m
mg - ks = 0 m
motion x l + s
Alternative Form of Solution
By applying trigonometric formula, we have:
2 2 1
2 2
1 , tan
), sin(
)
( c
c c c
A t
A t
x
x negative
x positive x = 0
5
1 x
0 x
20
17
x
0 x
20
17 x
Aging Spring
In real world, the spring constant k usually varies as the spring gets old. Replace k with k(t) = ke–t, k > 0,
> 0, we have a more realistic system model:mx + ke–tx = 0
Non-constant coefficient 2nd-order linear DE!
Free Damped Motion
DE of free damped motion:
The roots of the auxiliary eq.:
dt kx dx
dt x
m d22
0
2 2
2
2 x
dt dx dt
x
d
2
2
m m
m
Three Cases of Damped Motion
Case I: Over-damped
Case II: Critically damped
Case III: Under-damped
) (
)
(t e t c1e 2 2t c2e 2 2t x
) (
)
(t e c1 c2t x t
) sin
cos (
) (
2 2
2
2 2
1
t c
t c
e t
x t
t x
x
t
x
t undamped
underdamped
Driven Motion
Now, consider the effect of external
force f(t) on the damped motion system:
)
2 (
2
t dt f
kx dx dt
x
m d
) (
2 2
2 2
t F dt x
dx dt
x
d
m
Transient and Steady-State Terms
When F(t) is a periodic function and
> 0, the solution is the sum of a non-periodic function xc(t) and aperiodic function xp(t). Moreover limt xc(t) = 0.
1
-1
transient
steady-state x
t
1
-1
x( t) = transient + steady-state x
t
/2 /2
xp(t)
Example: Transient/Steady State
The solution of
is x(t) = (x1 – 2)e–t sin t + 2 sin t,
2 1 2
) 0 ( ,
0 )
0 ( ,
sin 2 cos
4 2
2 x t t x x x
dt dx dt
x
d
transient steady-state 4
- 4
x1= 7 x
t - 2
2
x1= 3 x1= 0
x1= -3
2
Undamped Forced Motion
The solution of
is
where c1 = 0, c2 = –γF0 /ω(ω2 – γ2).
There is no transient term.
0 )
0 ( ,
0 )
0 ( ,
0 sin
2 2
2 x F t x x
dt x
d
F t t
c t
c t
x
sin sincos )
( 1 2 2 0 2
( sin sin ) ,
) ) (
( 2F0 2 t t
t x
Pure Resonance
In the previous example, when
, thedisplacement of the system become large as t .
) (
) sin
sin (
lim
) (
sin lim sin
) (
2 0 3
2 0 2
d d
t d t
d F
t F t
t x
t F t
F t
sin 2 cos 2
0 2
0
t x
Tacoma Narrow Bridge, WA, USA
Opened in July 1, 1940, collapsed in Nov. 7, 1940.
The wind-blow frequency matched the natural frequency of the bridge, which caused a pure resonance effect that destroyed the bridge.
Damping System of Taipei 101
Taipei 101 uses a 730-ton damping ball† to stabilize the building under wind-blow effect
Linear Models: BVP
The deflection of a flexible beam can be modelled by a 4th-order differential equation:
load per unit length flexural rigidity
A straight flexible beam The deflection curve of the beam
)
4 (
4
x dx w
y EI d
Flexible Beam Applications
For precise robot arm control, we must take into account the bending effect of the robot links:
Base position (x0, y0)
True end-point position (x4, y4)
Estimated end-point position (𝑥 , 𝑦 ) without considering bending effect
Boundary Conditions
Boundary conditions of a flexible beam:
End of beam Boundary conditions
embedded y = 0 y = 0
free y = 0 y = 0
supported y = 0 y = 0
x = 0 x = L x = 0 x = L x = 0 x = L
Embedded at both ends Free at the right end Supported at both ends
Eigenvalue Problems
An eigenvalue problem in DE is a homogeneous BVP such that the boundary conditions evaluate to 0 and there is a parameter
at the coefficient of y:y" + p(x)y' +
q(x)y = 0, y(a) = 0, y(b) = 0.The eigenvalue problem tries to find a
(eigenvalue) such that the BVP has a nontrivial solution. The non-trivial solution that corresponding to an eigenvalue
is then called an eigenfunction.Example: y" + y = 0, y(0) = y(L) = 0 (1/2)
The problem can be solved by enumerating different cases when
= 0,
< 0, and
> 0.(1)
= 0, we have y = 0, the general solution is y(x) = Ax + B.
y = 0 is the only solution for the BVP
= 0 is not an eigenvalue of the BVP(2)
< 0, let
= –
2,
> 0, we have y –
2y = 0, the general solution is y(x) = c1ex + c2e–x.
y = 0 is the only solution for the BVP
< 0 do not have eigenvalues of the BVPExample: y" + y = 0, y(0) = y(L) = 0 (2/2)
(3)
> 0, let
=
2,
> 0, we have y +
2y = 0, the solution is y(x) = c1cos(
x) + c2 sin(
x). y(0) = 0 implies c1 = 0
y(L) = 0 implies sin(
L) = 0, or
L = n
, n Z The BVP has infinitely many eigenvalues:
and the corresponding eigenfunctions are:
Nonlinear Spring Models (1/2)
The general mathematical model of an undamped spring has the form:
for a linear spring model, F(x) = kx. However, spring are quite often nonlinear, e.g. F(x) = kx + k1x3.
x F
soft spring hard spring
linear spring
0 )
2 (
2 F x
dt x m d
Nonlinear Spring Models (2/2)
Damping force of a spring system can be nonlinear as well:
Restoring force F(x) is usually an odd function such as kx + k1x3. The reason is that we want F(–x) = –F(x).
0 )
2 (
2 F x
dt dx dt
dx dt
x
m d
Nonlinear Pendulum
The pendulum system can be modeled as
Using Maclaurin series of sin
, we haveO
l
mg sin
mg cos
P W = mg
. 0
2 sin
2
l g dt
d
! , 5
! sin 3
5 3
Linearization of Nonlinear Systems
Assuming that sin
–
3/6, we have:System can be linearized by assuming sin
: Impact of initial values:
(0) = 12,
. 6 0
3 2
2
l g l
g dt
d
.
2 0
2
l g dt
d
A nonlinear model
similar to the spring systems!
𝜃 0 = , 𝜃 0 = 2
𝜃 0 = , 𝜃 0 =