ELEC-C8201 Control and Automation

Lecture 2

Dynamic models, state-space representation,

system responses
 Physical modelling

 Consider physical modelling of involved parameters in electrical circuits,

mechanical systems (both linear and rotary motion) and flow systems.
 This review focuses on simple linear components, leaving, for example,
thermal and energy considerations outside the scope of the review.
 Basic components of electrical circuits
i(t)
 Resistor (resistance) R

v(t )  Ri(t )
v(t)

 Coil (inductance) i(t)

di (t ) L
v(t )  L
dt
v(t)
 Capacitor (capacitance)

dv(t ) C i(t)
i (t )  C
dt
v(t)
 Example. Electrical circuit
Making a model for the electrical i1(t) R2 i2(t)

R1
circuit
 The input, or impulse, is v0 (t) and
vR1(t) vR2(t)
the output quantities, i.e., the
v0(t) v1(t) C1 v2(t) C2
voltages V1 (t) and V2 (t).
i3(t)
 Electric currents and resistors can
be modelled as

dv1 (t ) dv (t ) vR1 (t )  R1i1 (t ), vR 2 (t )  R2i2 (t )

C1  i3 (t ), C2 2  i2 (t )
dt dt
 Kirchoff's First Law
 Second Kirchoff law i1 (t )  i2 (t )  i3 (t )
 v0 (t )  v1 (t )
 i1 (t ) 
 0
v ( t )  v R1 (t )  v1 ( t )  0
v ( t )  R i
11 (t )  v1 (t )  R1
    
1v (t )  v R2 ( t )  v 2 (t ) 1v (t )  R i
2 2 ( t )  v 2 (t ) i (t )  v1 (t )  v2 (t )
 2 R2
 Example. Electrical circuit
 This model has voltages as states (memory elements), so it is advisable
to eliminate the electrical currents as unnecessary variables from the
developed equations
 dv1 (t ) 1 1 v0 (t )  v1 (t ) v1 (t )  v2 (t )
 dt  C i3 (t )  C  i1 (t )  i2 (t )   R1C1

R2C1
 1 1

 dv2 (t )  1 i (t )  v1 (t )  v2 (t )
 dt C2
2
R2C2

 dv1 (t )  1 1   1   1 
     v1 (t )    v2 (t )    v0 (t )
 dt  R1C1 R2C1   R2C1   R1C1 
 
 dv2 (t )   1  v (t )   1  v (t )
 dt   1   2
 R C
 2 2 R
 2 2C
 Basic components of mechanical systems
Linear Motion: m
 Mass (inertia)

d 2 x(t )
Fm (t )  m x(t)
dt 2
k
 Spring
Fk (t )  k x (t )  k ( x1 (t )  x2 (t ))
x1(t) x2(t)

 Damper B
d x(t )  dx1 (t ) dx2 (t ) 
Fb (t )  B  B  
dt  dt dt 
x1(t) x2(t)
 Example. Mechanical system
 Make a model for a mechanical k
system in which two mass pieces F(t)
are connected together with a m2 m1
spring and a damper
 The input is an external force F B
(T) and the position of the second
mass x2(t) is the output x2(t) x1(t)
 First mass force equation:
d 2 x1 (t )  dx1 (t ) dx2 (t ) 
m1 (t ) 2
 B     k  x1 (t )  x2 (t )   F (t )
dt  dt dt 
 Second mass force equation:
d 2 x2 (t )  dx1 (t ) dx2 (t ) 
m2 (t ) 2
 B    k  x1 (t )  x2 (t ) 
dt  dt dt 
 Basic components of mechanical systems
Rotational motion:
 Moment of inertia

d 2 (t ) (t) J
TJ (t )  J
dt 2

 Torque Spring
k
Tk (t )  k1 (t )
1(t) 2(t)

 Torque Damper B
d1 (t ) 1(t) 2(t)
Tb (t )  B
dt
 Example. Mechanical system
 Make a model for the rotating
B
system shown in the figure. The 1(t) k 2 (t)
excitation is torque T (t) and the T(t) J
response is angles 1(t) and2(t)

 d 21 (t )
 J dt 2  k 1 (t )   2 (t )   T (t )

 B d 2 (t )  k  (t )   (t ) 
 dt
1 2
 Basic components of flow systems
F1(t)
 Flow tank
dV (t )
 F1 (t )  F2 (t )
dt V(t) F2(t)
 Ideal Mixer F1(t)
dV (t )C2 (t ) C1(t)
 F1 (t )C1 (t )  F2 (t )C2 (t )
dt
V(t)
 Pipe flow delay C2(t) F2(t)
 V  C2(t)
C2 (t )  C1 (t  Td (t ))  C1  t  
 F (t ) 
F(t)
 Flow through an orifice C1(t) V C2(t)
F(t)
F (t )  A(t ) R p(t )  A(t ) R p1 (t )  p2 (t ) p1(t) p2(t)
R
A(t)
 Example. Flow System
 The flow system is shown in the diagram. F2 F1
There is the input flow concentration C1 (t) C2(t) C1(t)
and the concentration of the output flow is
C3(t). Flow and volumes are constants
V F3 C3(t)
 The flow branching point can be F1
F3  F1  F2 F2 C3(t)
 Considering ideal sensor and pipe flow Vp C3(t)
dVC3 (t )  Vp 
 F1C1 (t )  F2C2 (t )  F3C3 (t ), C2 (t )  C3  t  
dt  F2 
 Eliminating the variable C2(t) and F3 :

dC3 (t ) 1   Vp  
  F1C1 (t )  F2C3  t     F1  F2  C3 (t ) 
dt V  F2  
 Example. Flow System Simulation
 Configuring simulation model for flow system with Simullink

V= 0.2 Impulse
Vp= 2 response
Step response F1= 1
F2= 2
 State Space Representation
 The state space representation is a compact way of representing high-
order differential equations/systems.
 The instantaneous state of the system is a complete description of the
system. If the initial state (state quantities at the beginning) and all the
input quantities are known from the beginning, then the system state and
the output quantities can be determined at an arbitrary time. It follows
that the state space representation is very suitable for simulation.
 The control of state variables allows for better system control compared
with the control of the system's output quantities.
 The state space representation is a standard-format representation, so
the management mechanisms can be standardized independently of the
system (equations are independent of the individual system number and
parameters)
 The state space representation is suitable for modeling and managing
multivariate systems
 State Space Representation
 In a state space representation, an arbitrary order differential equation/system is
represented as a group of first-order differential equations.
 The selection of spaces can be made in infinitely different ways = > The state
space representation is not unique but many different state variables can
describe the same input/output model.
 The general state space representation is of the format x RS
 (t )  f (x(t ), u (t ))
Ty(t )  g(x(t ), u(t ))
 x(t) is the state, u(t) the control input y(t) output - All these quantities can be
vectors or scalars.
 f(x(t),u(t)) is the system equation (describing system dynamics) and g(x(t),u(t)) is
the starting description (explains how the output quantities depend on the input
and the state)
 If u(t) is a scalar u(t) and y(t) is a scalar y(t), then this is a SISO system-
regardless of the vector x(t)’s dimension.
 Example. Flow System in State Space representation

 The flow process is mixed with antifreeze C1 F1 (t) C2 F2 (t)

(dilute solution with a chemical
concentration of C1 in a spring solution with
a concentration of C2).
 The objective is to obtain the desired
production volume (flow F) of the product
(with concentration C) that meets the h(t) C(t)
specifications, using the flow rate (F1 and
F2).
 The mixing tank is connected to a discharge A
valve open to atmospheric pressure = > The
removal flow is proportional to the square C(t) F(t)
root of the surface:

F ( t )  k h(t )
 Example. Flow System in State Space representation

 A balance equation is formed (simplified volume balance) and

component balance
dV (t ) R|
A
dh(t )
 F1 (t )  F2 (t )  F (t )
dt
S|
dC (t )V (t )
dt

dt |T  C1 F1 ( t )  C2 F2 (t )  C (t ) F ( t )

dC (t )V (t ) dC (t )
dt

dt
V (t )  C (t )
dV (t ) dC (t )
dt

dt
b
V (t )  C (t ) F1 (t )  F2 (t )  F (t ) g
R| A dh(t )  F (t )  F (t )  k h(t )
1 2
 S| dC(t )
dt

|T dt Ah(t )  bC  C(t )gF (t )  bC  C(t )gF (t )

1 1 2 2
 Example. Flow System in State Space representation

 Obtain a simple equation group of first-order differential equations

R| dh(t )  1 d F (t )  F (t )  k h(t ) i
1 2
 S| dC(t ) 1
dt A

|T dt Ah(t ) cb C  C ( t ) g F ( t )  bC  C ( t ) g F ( t ) h
1 1 2 2

 Choose H and C for state , F1 and F2 for input and output quantities as F
and C
L
x( t )  M
x (t ) O L h( t ) O
1
P M P ,
L
u( t )  M
u (t ) O L F (t ) O
1
P M
1
P ,
L
y( t )  M
y (t ) O L F (t ) O
1
P M P
N x ( t ) Q NC ( t ) Q
2 Nu ( t ) Q N F ( t ) Q
2 2 N Q N Q
y (
2 t ) C ( t )
 These variable selections can be used to write directly in the standard
format: RS
x (t )  f (x(t ), u (t ))
Ty(t )  g(x(t ), u(t ))
 Example. Flow System in State Space representation

 Writing the system equation:

L 1
d i O
L x (t ) O L f (x(t ), u(t )) O M u ( t )  u ( t )  k x ( t ) P
M 1 P
1 2 1
x (t )  M P  f (x(t ), u( t ))  M P A
1 1

N Q
x ( t ) N f ( x ( t ), u ( t )) Q MM Ax (t ) cbC  x (t )gu (t )  bC  x (t )gu (t )hPP
N Q
2 2
1 2 1 2 2 2
1

 For the initial description, the dependency of the output variables on state
variables
L y ( t ) O L F ( t ) O L k h( t ) O L k x ( t ) O
y(t )  M P M P M P M P
1 1

N Q N Q N Q N
y ( t
2 ) C ( t ) C ( t ) x ( t ) Q 2

 Initial description:

L y (t ) O L g ( x(t ), u( t )) O L k x (t ) O
y( t )  M P  g( x(t ), u( t ))  M P M P
1 1 1

N Q
y2( t ) N g
2 ( x ( t ), u ( t )) Q N x ( t )
2 Q
 Example. Flow System in State Space representation

 The flow process state space representation can be

R| L 1
d i O
M
||x (t )  M 1 A
1 u ( t )2 u ( t )  k x
1 ( t ) P
P
M
S| MN Ax (t ) cb1 C  x (t ) gu (t )  bC  x (t ) gu (t )hP
2 1 2 2 2
PQ
1

||y(t )  LMk x (t ) OP
1

T N x (t ) Q
2

 In this example, the selection of modes was easy because the

appropriate state variables were obtained directly from the system
model. Examining other methods for selecting variables in connection
with linear mode representation.
 Linear mode representation
 The previous example described a non-linear state space
representation. If the system being viewed is linear, its variables and
parameters can be assembled into separate vectors and matrix to obtain
a standard linear mode representation.

RSx (t )  f (x(t ), u(t ))  Ax(t )  Bu(t )

Ty(t )  g(x(t ), u(t ))  Cx(t )  Du(t )
 The parameter array A is called system matrix, B control matrix, C output
matrix, and D as a direct effect matrix. Often D = 0, in which case the
entire direct effect term disappears from the state representation. (This is
done with a strictly proper system).
 Linear mode representation
 Differential equations  x1  a11 x1    a1n xn  b11u1    b1mum
 x  a x    a x  b u    b u
 2 21 1 2n n 21 1 2m m


 xn  an1 x1    ann xn  bn1u1    bnmum

 Can be presented as an array equation

 x1   a11  a1n   x1   b11  b1m   u1 

                 
  
 xn   an1  ann   xn  bn1  bnm  um 
eli: x  Ax  Bu
 Electrical Circuit State Space representation
 Examine the electrical circuits of previous example and develop the state

 dv1 (t )  1 1   1   1 
     v1 (t )    v2 (t )    v0 (t )
 dt  R1C1 R2C1   R2C1   R1C1 

 dv2 (t )   1  v (t )   1  v (t )
 dt   1   2
 R C
 2 2 R
 2 2C

 The natural selection of state variables is the voltages of the capacitors

(because they are represented with first order differential equations).
Only the voltage of the second capacitor is now selected as the output
v2.
x(t ) 
LM OP
v1 (t )
, u( t )  v0 (t ), y (t )  v2 (t )
N Q
v2 ( t )

 These choices provide

 State-space representation of an electrical circuit

   1 1  1 1 
     x1 (t )  x2 (t )  u (t ) 
x (t )    R1C1 R2C1  R2C1 R1C1   f ( x(t ), u (t ))
  
 1 1
  x1 (t )  x2 (t ) 
  R2C2 R2C2 
 y (t )  x2 (t )  g (x(t ), u (t ))
 And thus:
   1 1  1 
       1 
x (t )    R1C1 R2C1  R2C1  x(t )   R C  u (t )  Ax(t )  Bu (t )
    1 1

 1 1
 0 
   
 R2C2 R2C2 

 y (t )   0 1 x(t )  0 u (t )  Cx(t )  Du (t )
 Forming a State Space representation
 How to generate a state space representation systematically?
 Choosing physical and rational state variables from model equations (as in
previous examples)
 Using a derivative operator p
 Using canonical forms

 Physically correct state variables

 Often the easiest way
 Using the "derivatives" operator
 can be formed e.g. Canonical shape, or in some cases a diagonal shape,
should be controlled or observed
 Forming a State Space representation
 Examining the mechanical System F(t)

mx( t )  Bx ( t )  kx (t )  F ( t ) x(t)

m
 Physically relevant state variables
 Select position x and speed v = dx/dt RSx (t )  x(t )
1

y (t )  x ( t )  x1 (t ), u(t )  F (t ) Tx (t )  x(t )
2 k B

 Determining the derivatives of the selected state variables

R|x (t )  x(t )  x (t )
1 2

S|x (t )  x(t )   B x(t )  k x(t )  1 F (t )   B x  k x (t )  1 u(t )

T 2
m m m m
2
m
1
m

R|x (t )  L 0 1 O x(t )  L 0O u(t )

S| MN  PQ MN PQ
k
m
B
m
1
m

T y ( t )  1 0 x( t )
 Forming a State Space representation
 Second method: Use the p operator p  d / dt
 You write the original equation using the p operator whenever there is a
differentiation: x 
2 (t )


B

k 1
x1 ( t )
8
B k 1
x (t )  x (t )   x (t )  F (t )  p{ p{x (t )}  x (t )}   x (t )  F (t )
m m m m m m
R|x (t )  x(t ) R|x (t )   B x (t )  x (t )
|| 1

B B ||
1
m
1 2

S| x ( t )  p{ x ( t )}  x ( t )  
x ( t )  x (t ) k 1
2
m
1 1
m
|| p{x (t )}  x (t )   k x(t )  1 F (t )   k x (t )  1 u(t ) S|
2 
x ( t )  
m
1x ( t ) 
m
u(t )

T 2 2
m m
1
m m || y(t )  x(t )  x (t )
1

T
R|x (t )  L 1O x(t )  L 0O u(t )
B

S| MN 0PQ MN PQ
m
k 1
m m

T y ( t )  1 0 x( t )
 Differential equation to transfer function

The differential equation can also be used as a shortcut to the

transfer function. Common linear differential equation
y ( n ) (t )  a1 y ( n 1) (t ) an 1 y (1) ( t )  an y (t )  b1u ( n 1) ( t ) bn 1u (1) (t )  bn u( t )

is Laplace-transformed (and assuming initial values as zeros)

 1
s n
 a s n 1
   an 1 s  an Y ( s )   1    bn1s  bn U ( s)
b s n 1

This makes it easy to create a transfer function

Y ( s) b1s n 1    bn 1s  bn
G ( s)   n
U ( s) s  a1s n 1    an 1s  an
Similarly, the transfer function can be re-transformed to
differential equations
 Example: Mass block

Specify the transfer function and weighting function in the

previous example RS
y (t )  x ( t )
x(t )  2 x (t )  5 x(t )  F (t ) T
u( t )  F ( t )

y (t )  2 y (t )  5 y (t )  u (t )  s 2Y ( s)  2 sY ( s)  5Y ( s )  U ( s)

c h
s  2 s  5 Y ( s)  U ( s)  G ( s) 
2 Y ( s)
 2
1
U ( s) s  2 s  5

g (t )  L1
lG(s)q  L 1 RS 1 UV 1 1
 L
RS 2 UV
1 t
 e sin( 2t )
Ts 2
W
 2s  5 2 T
( s  1)  2
2 2
2 W
The weighting function is now the same as the unit impulse
response
 Determination of response

When the transfer function is known, the response (forced

response) is calculated as follows

 Laplace conversion of external control u(t) l q

U ( s)  L u(t )
 Solve for output Y(s) in Laplace domain Y ( s)  G ( s)  U ( s)
 Take inverse Laplace transform of output y (t )  L 1
lY (s)q
Or: y (t )  L1
mG( s)  Llu(t )qr
 Example: Mass block

Determine the unit step and ramp responses for the mass
position. A transfer function was previously assigned to the
system as: 1
G ( s) 
s2  2 s  5
1
Unit step: U ( s) 
s
Response: 1 1 1
Y ( s)  G ( s )  U ( s)   
s 2  2 s  5 s s( s 2  2 s  5)
Taking partial fractions:
1 A Bs  C A( s 2  2 s  5)  s( Bs  C ) ( A  B ) s 2  (2 A  C) s  5 A
  2  
s( s  2 s  5) s s  2 s  5
2
s( s  2 s  5)
2
s( s2  2 s  5)

R| A  B  0 R| A  1
5
 ( A  B ) s 2  (2 A  C ) s  5 A  1  S|2 A  C  0  S|B   1
5

T5 A  1 TC   2
5
 Example: Mass block
1 1
5s 5
1 2
11  s 1 1 2 
Y ( s)    2
5
    
s ( s  2s  5) s s  2 s  5 5  s  ( s  1)  2 2 ( s  1) 2  22  
2 2 2

The response can be:

l q e1  ce
y (t )  L1 Y ( s)  1
5
t
cos( 2t )  21 e  t sin(2t ) hj
Ramp input:
1 1 1 1
U ( s)  Y ( s)  G ( s)  U ( s)   
s2 s2  2 s  5 s2 s 2 ( s 2  2 s  5)

As before, taking partial fractions and solving:

l q ct 
y (t )  L1 Y ( s)  1
5
2
5  25 e  t cos(2t )  103 e  t sin(2t ) h
 Steady State Response
Steady state response tells you how much the signal is
strengthened or dampened after passing through the system
 With the unit step response, steady state amplification indicates to which level
the response will remain (asymptotically stable system)
 With the ramp response, the amplification indicates the continuity of the slope
of the response (Asymptotically stable system)
Steady state response can be calculated from the transfer
function using the limit value theorem. Steady state
response can also be set for unstable system transfer
functions, but it does not have a physical interpretation that is
linked to the end value of the response.

Steady state response (static gain) is: k  lim G ( s)

s 0
l q
 Example: Mass block

Evaluating the mass block system for Steady State response:

1
G ( s)  2
s  2s  5


R
k  limlG ( s)q  limS
1

1UV
s 0
Tss 0 2
 2s  5 5 W
Also seen in time domain as t tends to ꚙ
 Unit Step Response y (t )  1
5 e1  ce cos(2t )  e sin(2t )hj
t 1
2
t

 Ramp response y (t )  1
5 ct   e cos(2t )  e sin(2t )h
2
5
2
5
t 3
10
t
 MATLAB: Models, Inputs and Responses

 Enter a template function in the workspace:

 One can write in transfer function or state space tf, zpk & ss –commands.
 For example, consider the mechanical system in example 2:
 The system depicts a differential equation:
mx(t )  Bx (t )  kx(t )  F (t )  x(t )  2 x (t )  5 x(t )  u (t )
 Entering in workspace as: sys=tf(1,[1 2 5])
Transfer function: 1
-------------
s^2 + 2 s + 5
 Examining Impulse and step responses
[imp1,t1]=impulse(sys);
(Tai: [imp1,t1]=impulse(tf(1,[1 2 5])); )
[ste2,t2]=step(sys);
plot(t1,imp1)
plot(t2,ste2)
 MATLAB: Models, Inputs and Responses
 Output responses:

 For confirmation:
 k=dcgain(sys)
k =0.2000
 Responses can also be examined in
ltiview-LTI window
 ltiview
 file -> import -> sys

 edit -> plot configurations

 MATLAB: Models, Inputs and Responses
 In addition to the basic responses, the MATLAB's
workspace allows you to calculate responses to
arbitrary control inputs using the lsim command.
 Ramp response
 ram=lsim(sys,t2,t2);

 plot(t2,ram)

 Sine response
 osc=lsim(sys,sin(10*t2),t2);

 plot(t2,osc)

 lsim–For a command, you can use some other

command to take a calculated time vector, or generate
one yourself, either by using a generic Matlab
command (colons) or by using a special linspace
command
 t3=(0:0.1:10)';

 t3=linspace(0,10,101)';
 MATLAB: Models, Inputs and Responses
 In earlier lectures, it was told that the basic responses
were derived from each other by either derivation or
integration. Now, we see how well this works
numerically
 Take the base response as step function and
find out impluse response by derivation and
slope (ramp) response by integration:
 The numerical derivative of a step response:
 imp2=diff(ste2)./diff(t2);
 plot(t2,[0;imp2],t1,imp1)

 Step Response Numerical Integral:

 delta=mean(diff(t2))
 ram2=cumsum(ste2)*delta;

 plot(t2,[ram2 ram])

 Numerical derivatives and integration work reasonably

well in this (undisturbed) case.
From State Space to Transfer Function form
The solution of convolution integration is laborious and often easier to
achieve if the state equation is left on Laplace form, with the time domain
function just written as Laplace inverse form (just as in the overall
response)
X( s )  ( sI  A ) 1 x(0)  ( sI  A) 1 BU( s )
x(t )  L1 ( sI  A ) 1 x(0)  L1 ( sI  A) 1 BU ( s)
  (t )x(0)  L1 ( sI  A) 1 BU( s )  x0 (t )  xu (t )
The output is obtained accordingly:
Y ( s )  CX( s)  DU( s)  C  ( sI  A )1 x(0)  ( sI  A) 1 BU( s )   DU( s )
 C( sI  A) 1 x(0)   C( sI  A) 1 B  D  U( s )  C ( s )x(0)  G ( s )U( s )
 Y0 ( s )  Yu ( s)
y (t )  L1 Y( s )
From State Space to Transfer Function form
Given below is a formula to define a transfer function from a state space
representation (the transfer function was defined as dependent on the
input quantities and outputs in the Laplace plane – when the initial values
are not taken into consideration)

G ( s )  C( sI  A) 1 B  D
Often the feedforward term D is zero, in which case the formula is even
simpler to form:
G ( s)  C( sI  A) 1 B
 Example: Mass block
 Solving a mass block Unit step response (forced response) using the
transfer function
1
 s 1  0 
G ( s )  C( sI  A) B  1 0 
1
 1 
 5 s  2   
 s  2 1 0 
1 0   1 
  5 s    1
  2
s  2s  5
2
s  2s  5
1
U (s) 
 The response s
 1 
y (t )  L Y ( s )  L G ( s )U ( s )  L  2
1 1 1

 s ( s  2 s  5) 

 15 1   e  t cos(2t )  12 e t sin(2t )  
Models and conversions between them:
Tilojen
DE to SSvalinta
Epälineaarinen
Non Linear Epälineaarinen
Non Linear
Differential Equations State Space
differentiaaliyhtälö tilaesitys
Tilojen
SS to DEeliminointi
Linearisointi
Linearize Linearisointi
Tilojen
DE to SS valinta
Lineaarinen
Linear Differential Lineaarinen
Linear State  =eA t Transfer Function
Equations Space
Tilansiirtomatriisi
matrix
differentiaaliyhtälö tilaesitys
Tilojen eliminointi
SS to DE
L{ } L-1 { }
Canonical forms
Kanoniset muodot
Transfer Functions
Siirtofunktio
(Laplace domain

G=C(sI-A)-1 B+D
Todellinen
Real L-1 { } L{ }
systeemi
System
Painofunktio
Time domain eli Tilaesityksen
General State
impulssivaste
impulse response yleinen ratkaisu
Space solution

d/dt

Step
Askelvaste
response

d/dt

Pengervaste
Ramp
response