MSE

FTP OrdDiff Problem sheet 7: Linear Systems

Nov. 2016

1. We consider the coupled oscillators

k1 + k2 k2 k2 k2 + k3
ẋ1 = x3 , ẋ2 = x4 , ẋ3 = − x1 + x2 , ẋ4 = x1 − x2
m1 m1 m2 m2

with m1 = 2, m2 = 94 , k1 = 1, k2 = 3, k3 = 15
4 .
a) Find the general solution to this ODE
b) Choose initial conditions, so that the system oscillates in it’s first / second eigenmode.
Sketch the motion of the masses.

2. Consider the system

ẋ1 = −x1 + 4x2
ẋ2 = −x1 + 3x2 + t
a) Find the general solution
 
1
b) Find the solution of the IVP ~x(0) = .
0
3. Consider
ẋ1 = x1 − x2 + t
ẋ2 = −2x1 + 1
a) Find the general solution
 
−1
b) Find the solution of the IVP ~x(0) = .
2
4. Consider two coupled oscillators, with damping proportional to the velocity and external
excitation of the first mass. This leads to the system

 ẋ1 = x3

 ẋ2 = x4



 ẋ3 = − k1m+k1
2 k2
x1 + m 1
x2 − mγ1 x3 + Fm
1 (t)
1
 ẋ = k2 x − k2 +k3 x −
 γ
4 m2 1 m2 2 m2 x 4

Parameters m1 = 2, m2 = 94 , k1 = 1, k2 = 3, k3 = 15
4 and damping constant γ = 0.5.
a) Explain, why all solutions of the homogeneous system (F1 (t) = 0) für t → ∞ converge
to zero.
b) Consider the harmonic excitation

F1 (t)
= ejωt .
m1
Sketch the graphs of ω 7→ |G1 (jω)| und ω 7→ |G2 (jω)|, assuming a particular solution of
the form
x(t) = G(jω)ejωt
.
5. An LTI System ẋ = Ax + bu is given by
      
d x1 −1 1 x1 0
= + u
dt x2 0 −1 x2 1


y = 1 0 x

is excited with the input function u(t) := exp(−t) for t ≥ 0. Find the response function
y(t) for the initial condition x(0) = 0
6. Remark: This is additional, optional Matlab exercise with transfer functions of LTI
systems, not on the exam.

a) Define the electrical oscillator from lecture 7 as a state space system in Matlab using
the matlab command ss
b) simulate the system with an arbitrary input u(t), using the command lsim. Plot the
output y(t)
c) calculate the transfer function using the matlab command tf and use it to find the
output y(t) as follows:
• use fft to calculate the frequency-domain representation û(w) of u
• pointwise multiply û(w) with the transfer function to get ŷ(w)
• use ifft to find y(t) and compare with the results from lsim
Answers for problem sheet 7
1. a) Coefficient matrix  
0 0
1 0
0 0
0 1
A=
−2 3
.
2 0 0
4
3 −3 0 0
Eigenvalues are λ1,2 = ±j, λ3,4 = ±2j, with eigenvectors
   
3 3
 2   −4 
v1,2 =
±3j  ,
 v3,4 =
±6j  ,


±2j ∓8j

the general solution is

       
3 cos(t) 3 sin(t) 3 cos(2t) 3 sin(2t)
 + C2  2 sin(t)  + C3 −4 cos(2t) + C4  −4 sin(2t)  .
 2 cos(t)       
x = C1 
−3 sin(t) 3 cos(t)  −6 sin(2t)   6 cos(2t) 
−2 sin(t) 2 cos(t) 8 sin(2t) −8 cos(2t)

(C1 , C2 , C3 , C4 ∈ R).
b) At t = 0 we get the equations

3C1 +3C3 = y10

2C1 −4C3 = y20
3C2 +6C4 = v10
2C2 −8C4 = v20

for C1 , C2 , C3 , C4 .
• First eigenmode: C3 = C4 = 0, so y10 = 32 y20 and v10 = 32 v20 . A possible choice of
initial conditions is

y10 = 3, y20 = 2, v10 = 3, v20 = 2.

Solution curve for 0 ≤ t ≤ 50:

 • For the second eigenmode: C1 = C2 = 0, so y10 = − 43 y20 und v10 = − 43 v20 . A
possible choice of initial conditions is:

y10 = 3, y20 = −4, v10 = 3, v20 = −4.

Solution for 0 ≤ t ≤ 50:

2. See scan below

3. a) Solve the homogeneous system first: ẋ = Ax: eigenvalues and eigenvectors of A are
   
1 1
λ1 = 2, λ2 = −1, v1 = , v2 = ,
−1 2

also    
1 1
xh = C1 e2t + C2 e−t (C1 , C2 ∈ R)
−1 2
 For the non-homogeneous system ẋ = Ax + b(t) we choose the trial function
 
at + b
xs =
ct + d

Inserting this yields the equations

0 = a−c+1
a = b−d
0 = −2a
c = −2b + 1

with solution a = 0, b = 0, c = 1, d = 0 also

 
0
xs = .
t

The general solution x = xh + xs of the non-homogeneous system is therefore

x1 = C1 e2t + C2 e−t
(C1 , C2 ∈ R).
x2 = −C1 e2t + 2C2 e−t + t

b) The initial condition yields C1 = − 43 , C2 = 31 , therefore

x1 = − 43 e2t + 13 e−t
x2 = 34 e2t + 23 e−t + t

4. a) The system in matrix-form is

0 0 1 0
   
0
 0 0 0 1  0 F1 (t)
ẋ =  x +  
− k1 +k2
m1
k2
m1 − mγ1 0  1 m1
k2 γ
m2 − k2m+k
2
3
0 − m2 0
| {z } | {z }
=:A =:B

Eigenvalues of A are (for the given parameter values)

λ1,2 ≈ −0.1157 ± 1.9966j, λ3,4 ≈ −0.1204 ± 0.9928j

All eigenvallues have negative real part, so the solutions tend to zero (phyisical point of
view: the energy is changed to heat through the friction)
b) As for the electrical networt we find G(jω) to be

G(jω) = −(A − jωE4 )−1 B

und G1 (jω) und G2 (jω) als

Gi (jω) = Ci G(jω) = −Ci (A − jωE4 )−1 B, i = 1, 2

with


C1 = 1 0 0 0 , C2 = 0 1 0 0 .
Inserting A, B, C1 und C2 gives the following graph for G1 (jω) und G2 (jω):
5. Several methods are possible. With the matrix exponential for the state matrix A:

   
k k 1 −k −t 1 t
A = (−1) =⇒ exp(At) = e
0 1 0 1

very similar to the example from the lecture. With this, we find:

Z t
x(t) = exp(At)x(0) + exp(At) exp(−As)bu(s) ds
0

 Z t      Z t 
−t1 t s 1 −s 0 −s −t 1 t −s
= e e e ds = e ds
0 1 0 0 1 1 0 1 0 1
 2   
−t −t /2 t/2
= e = t exp(−t)
t 1

Thus

t2
y(t) = exp(−t)
2

6. See exElectricalNetworkTFSim.m