state space representation part-1
TRANSCRIPT
-
8/22/2019 State space representation Part-1
1/47
Lecture -6: State-space Representation -2
hours
By Mr S Wijewardana
PhD student QMUL 21-04-
2013
Learning Objectives:1. Reduction of higher order equations to first-
order.
2. State Equations and Stare Variables.3. State-space Analysis & State-space Model.
4. From State-space to Transfer Functions.
-
8/22/2019 State space representation Part-1
2/47
Solution of higher order differential equations are
much more difficult to handle than first order
differential equations.
It is easy to handle mathematically when the order
of the differential equation is of first order.
include
Consider the differential equation given below:
The easiest way to solve the
above equation is to do a
substitution:
Now we can write the equation:
xyy ...
wy .
xww .
This equation is much
easier to handle than the
second derivative equation -1
--------eq1
-
8/22/2019 State space representation Part-1
3/47
From our Maths knowledge we can see that we have to use an integral factor to
solve the equation now:
Where p is the coefficient in front of the independent variable x:Nothing will happen to the equation if we multiply each term by an
integrating factor:
xdxpdx eee
xewewexxx
.
21
2
1
1
1
.2
1
)1(1
)(
)(
cecxx
dxecxwyecxw
cexe
dxexe
dxxewe
xewedx
d
x
x
x
xx
xx
xx
xx
-
8/22/2019 State space representation Part-1
4/47
Consider two differential equations as shown below:
Reduce them to first order differential equations:
033
0343
.......
.......
vvvuu
vvvuuu -----------eq- 1
------------eq-2
Assume:
54
.
43
.
2
.
1
.
..
5
.
43
.
21
:
;;;;
xx
xx
xux
Hence
vxvxvxuxux
Eq-1 &2 can now be
written:
033
0343
345
.
22
.
345
.
12
.
2
xxxxx
xxxxxx
-
8/22/2019 State space representation Part-1
5/47
By adding and subtracting the last two equations:
402
302223
4315
.
34122
.
eqxxxx
eqxxxxx
We can write all equations as shown below:
431
.
5
.
4
.
4321
.
.
25
4
3
22322
21
xxxx
xx
xx
xxxxx
xx
We can write the above
equations in Matrix form:
-
8/22/2019 State space representation Part-1
6/47
5
4
3
2
1
5
.4
. 3
.2
.1
.
01102
10000
01000
02232
00010
x
x
x
x
x
x
xx
x
x
State Equations and State Variables
)()()()(
.)(
11
1
1 tftyadt
tdya
dt
tyda
dt
tydon
n
nn
n
______________________________________________________________________
_
Consider a differential equation as shown
below:
---Eq-1
-
8/22/2019 State space representation Part-1
7/47
If we substitute:
1
1
2
1
)()(
.
.
)()(
)()(
n
n
ndt
tydtx
dt
tdytx
tytx
Then we can get:
)()(
)()(
32
21
txdt
tdx
txdt
tdx
Hence we can write:
)()()()()()( 1122110 tftxatxatxatxadt
tdx nnnnn
)()()()(
.)(
11
1
1 tftyadt
tdya
dt
tyda
dt
tydon
n
nn
n
Eq-2:
State
equations
-
8/22/2019 State space representation Part-1
8/47
Eq-2: as you can see clearly now a first order
differential equation with different set of variables.
(However, original equation is the nth order
equation.)The variables x1, x2, x3 are called statevariables
-
8/22/2019 State space representation Part-1
9/47
State-space method is a generalized mathematical modelling technique
which can be used to analyse control systems.(also called time-domainapproach)
This has become more popular as it provides more flexibility of modelling
simple to complex nonlinear systems.
State space model can also be applied to other areas like economics,
agriculture, weather etc.
The model is simply a set of first order differential equations written in
Matrix form.
The normal classicalfrequency domain method is limited only to linear,time invariant systems(LTI systems)(But, this method gives the stability and
transient response information very rapidly)
However, state-space method can be applied to time varying systems like
missiles etc.
Other advantage of the state-space model is the flexibility of choosingstate variables.
-
8/22/2019 State space representation Part-1
10/47
Consider a system described by an nth orderdifferential equation:
radtda
dtda
dtd
n
n
nn
n
121
1
.
We are free to choose our own state variable in order to bring the differential
equation into first order that we can handle easily.
Convenient way to select state variables is to choose the output, (t), and itsn-1 derivatives as state variables: this choice is normally called the Phase variable
choice.
Let us now select the state variables:
))(1(...
4
..
3
.
2
1
,
,
dotsn
nxx
xx
x
------------------eq1
------------------eq2
A dot above omega signifiesdifferentiation w.r.t. time.
-
8/22/2019 State space representation Part-1
11/47
From equations 1 to 3 we can get
nnxxxx
xx
1
.
32
.
2
.
,
1
Substituting the above relationship to our original nth order
differential equation: ( for convenience look at from left to right)
))(1(...
4
..
3
.
2
1
,
,
dotsn
nxx
xx
x
nnnnn
nnnnn
xaxaxaxarx
rxaxaxaxax
112211
.
112211
.
...
-
8/22/2019 State space representation Part-1
12/47
Last equation in the previous slide is a set of first order differential
equations what we want:
Whole set of equations can now be represented in Matrix form as
shown below:
n
nn
x
x
x
r
x
x
x
aaax
x
x
x
2
1
2
1
221
.
3
.
2
.
1
.
001
1
.
0
.
.....
.....
0100
0010
These two equations represent the state-space model for the dynamicsystem of nth order differential equations. However, the standardnotation used in control systems is shown in the next slide:
-
8/22/2019 State space representation Part-1
13/47
DuCxy
BuAxx
.
State-space Model
-----------State equation
-----------Output equation
X is the state vector:
):1(,
)(
)(
)(
)(2
1
vectormatrixn
tx
tx
tx
tx
n
-
8/22/2019 State space representation Part-1
14/47
):1(,
)(
)(
)(
)( 2
1
vectormatrixq
ty
ty
ty
ty
q
Yis the output vector:
U is the input control vector:
):1(,
)(
)(
)(
)(2
1
vectormatrixp
tu
tu
tu
tu
p
-
8/22/2019 State space representation Part-1
15/47
A is the system matrix:
):(,
21
22221
11211
vectormatrixnn
aaa
aaa
aaa
A
nnnn
n
n
B is the input matrix:
):(,
21
22221
11211
vectormatrixpn
bbb
bbb
bbb
B
npnn
p
p
-
8/22/2019 State space representation Part-1
16/47
C = output matrix:
D = feed forward matrix:
):(,
21
22221
11211
vectormatrixnq
ccc
ccc
ccc
C
qnqq
n
n
):(,
21
22221
11211
vectormatrixpq
ddd
ddd
ddd
D
qpqq
p
p
-
8/22/2019 State space representation Part-1
17/47
State Variables:A set of user defined minimum number of linearly independent variables which
are used to define the system variables such that the knowledge of these
variables at any time t 0 , and information on the input excitation subsequently
applied, are sufficient to determine the state of the system at any time t t 0 .
State Vector:
A vector whose elements are the state variables
State Space:
The n-dimensional space whose axes are the state variables.
State Equation:
A set of n simultaneous, first order differential equations with n variables, where
the n variables to be solved are the state variables.
Output Equation:
Output equation is an algebraic equation, which signifies the output variables of
a system as a linear combination of the state variables and the inputs.
Note here, the number of state variables we define is equal to the order of the
differential equation.
We cannot select a variable which is linearly combined with another state
-
8/22/2019 State space representation Part-1
18/47
State Transition
Matrix
The solution to the homogeneous(In the matrix equation
AX=B if B =0 then we call homogeneous) state equationwhat we described earlier:
Axx .
Can be represented in the form of )0().()( xttx
)(tWhere is an n x n matrix and is the unique solution of .)0(),(.
ItA
)(t is called the state transition matrix.
Also note Atet )(
-
8/22/2019 State space representation Part-1
19/47
Example:
In a normal LRC circuit if VR(t) is chosen as state variable, then i(t)
cannot be written as a state variable because VR(t)= R.i(t)( one variable
is a linear combination of the other)
Example:
Find the state-space representation in Phase-variable form for the
transfer function
65
5
)3)(2(
5
)(
)(:
)3)(2(
5
)(
)(
2
sssssR
sC
Solution
sssR
sC
State space model in phase-variable presentation is:
equationOutputDuCxy
EquationSBuAxx
.. -----eq1
------eq2
-
8/22/2019 State space representation Part-1
20/47
Equation-1 and 2 can also be written in the form:
n
nn
x
x
x
Cy
Br
x
x
x
A
x
x
x
2
1
2
1
.
2
.
.
1
Cross multiplication of the TF yields:
(s2+5s+6) C(s) = 5 R(s)
Taking the s operator as: d/dt = s; d2/dt2 =
s2
rccc565
...
-
8/22/2019 State space representation Part-1
21/47
rccc 565...
Now we have to select the state variables:
..
2
.
3
..
3
.
1
.
2
.
2
111
cxxcx
cxxcx
xxcx
Now I can write:
rxxx
xx
565 122.
21
.
In matrix form:
2
1
2
1
2
.1
.
01
50
5610
x
xy
rxx
xx
-
8/22/2019 State space representation Part-1
22/47
5 x2
1.
x
2
.
x
-5
rxxx
xx
565 122.
21
.
x1
-6
r
Y(t) or
c(t)
Block diagram for the
computer simulation
-
8/22/2019 State space representation Part-1
23/47
Example:
Find the state-space representation in phase-variable form for the transfer
function:
65
)1(5
)(
)(2
ss
s
sR
sC
Solution:
Cross multiplication of the TF yields the derivative of R(which is
the input variable).C(s)[s2+5s+6] = (5s+5)R(s)
For convenience let us eliminate R(s).
To eliminate R(s) let us introduce a dummy variable called U:
Then:
)65(
)1(5
. 2
ss
s
R
U
U
C
Now we can
write: )1(5)(
)( s
sU
sCand
)65(
1
)(
)(
2
sssR
sU
-
8/22/2019 State space representation Part-1
24/47
Defining the state variables as before:
C(s) = 5s.U(s) +5u(s)
R(s) = s2.U(s) +5s.U(s) +U(s).6
uuc
ruuu
55
65
.
...
..
2
.
2
.
1
.
1;
ux
xuxux
12
122
.
21
.
55
5
xxcy
rxxx
xx
2
155
1
0
2
1
56
10
2
1.
.
x
xy
rx
x
x
x
-
8/22/2019 State space representation Part-1
25/47
Y(t) or
c(t)
r(t)1
x2
1.
x
2
.
x
-5
x1
-6
5
5
Block diagram for the state space model in
phase variable form(Computer simulation
diagram)
Question:
-
8/22/2019 State space representation Part-1
26/47
22187
23
)(
)(23
2
ssS
ss
sR
sC
Question:
-
8/22/2019 State space representation Part-1
27/47
Stability
we say system x =Ax is stable if etA 0 as t
meaning:
state x(t) converges to 0, as t , no matter what x(0) is
all trajectories of x =Ax converge to 0 as t fact: x =Ax is stable if and only if all eigenvalues of A have negative real
part:
i < 0, i = 1, . . . , n
-
8/22/2019 State space representation Part-1
28/47
From State Space to Transfer Function
When analysing control systems more often we haveto learn how to transfer the systems from statespace to TF and vise-versa.
Assume the standard state-space model:
DuCxy
BuAxx
.
2)()()(
1)()()(
eqsDUsCXsY
eqsBUsAXssX
In Laplace-
domain:
From equation-1:
[sI-A] X(s) = B U(s)
X(s) = [sI-A]-1 BU(s)
-
8/22/2019 State space representation Part-1
29/47
Substituting in equation 2 we get:
Y(s) = C[sIA]-1 BU(s) + DU(s)
Y(s) = [C(sIA)-1
B +D] .U(s)
The matrix [C(sI-A)-1B +D] is called the transfer
function matrix because it relates the output
vector Y(s) to the input vector U(s). {if U(s) andY(s) are scalars} Then we can write:
DBAsICsU
sYsG 1)(
)(
)()(
2]([
)(
.)(
)(
eqAsI
DAsIBAsIadjCsG
DBAsI
AsIadjCsG
-
8/22/2019 State space representation Part-1
30/47
From equation-2, we can see that the dynamic characteristics and stability are
determined by poles of the closed-loop transfer function, that is , the roots of sI-A=0
This equation is called the characteristic equation in state space model.
0 AsIIf the coefficients of matrix A are real
then are also
real.0 AsI
Example: A control system is defined by the state space model by the followingequations. Find the transfer function of the system.
2
1
2
1
2
.1
.
01
0
1
32
10
x
x
y
ux
x
x
x
By comparison with the state space model:
]01[,0
1,
32
10
CBA
-
8/22/2019 State space representation Part-1
31/47
We have already derived the transfer function
matrix as shown below:
DBAsICsU
sYsG 1)(
)(
)()(
21
3
)(
0
1
2
13
21
1)()(
2
13
23
1)(
1
2
1
ss
s
sG
s
s
ss
oBAsICsG
s
s
ssAsI
Using Matlab:
A=[0 1;-2 -3];
>> B=[1;0];
>> C=[1 0];>> D=0;
>> [num,den]=ss2tf(A,B,C,D)
num =
0 1 3
den =
1 3 2
-
8/22/2019 State space representation Part-1
32/47
Use Matlab to find the transfer function of the state-space model:
01,2
0,
32
10
CBA
Answer:num =
0 0 2.0000
den =
1 3 2
-
8/22/2019 State space representation Part-1
33/47
Solution of State Variables
How to obtain a general solution to the linear time-invariant state equation.
(space vehicle: mass will change after some time: time variant systems)
Solution of scalar differential equations.
Consider a system with only initial conditions without any forcing function:
)()(.
taxtx
Assume a solution of x(t) is in the form of power series:
x(t)= b0
+b1t + b
2t + b
3t + ---------eq2
...32)( 2321.
tbtbbtx
By substituting the assumed solution to the first equation:
b1 + 2b2t + 3b3t2 + = a [b0 + b1t + b2t
2 +]
Equating the coefficients of equal powers of t:
b1 = a b0
b2 = a b1/2 = a2b0= (1/2!) a
2b0
b3
= (1/3!)a3.b0
--------------------eq1
-
8/22/2019 State space representation Part-1
34/47
Substituting in equation 2 : When t = 0,
x(0) = b0 .
Therefore the solution x(t) can be written as:
)0()(
)0(.. .!3!2
1)(3322
xetx
xtata
attx
at
...)0(!2
1)0()0()(
2
xaaxxtx
x(0) is the initial condition of state variable
Combining equation 2
Laplace Transform Method to the solution of Homogeneous State Equations
-
8/22/2019 State space representation Part-1
35/47
p g q
Our system is:
)()(.
tAxtx ---------eq1
Taking Laplace transform of equation-1:Laplace transform of x: (x)= X(s)
Hence s X(s)X(0) = AX(s): Note the theorem )0()(
fssF
dt
df
s X(s)- A X(s) = X(0)
(s I - A) X(s) = X(0)
X(s)= (sIA)-1 X(0)
Hence: x(t) = -1 (s IA)-1x(0)
...3
2
2
1
s
A
s
A
s
IAsI
-
8/22/2019 State space representation Part-1
36/47
-1 (s IA)-1
)0()(
)(
!3!21
3322
xetx
eAsI
etAtA
AtI
At
At
At
we can write
Hence the
solution is:
)()(
3210
2
1
2
.1
.
txtx
x
x
Example:
Find x1(t) and x2(t) given initial conditions:
2
1
)0(
)0(
2
1
x
xUsing Laplace
transformmethod.
-
8/22/2019 State space representation Part-1
37/47
We have shown that given the initial conditions the solution to the above
state equation can be represented in the form of x(t) = eAt x(0).
This solution is for the state equation that can be represented as follows:
)()(.
tAxtx
Taking Laplace transform of both side of the equation we can write:
sX(s) -X(0) = A X(s)
(sIA)X(s) =X(0)
X(s) = (sI -A)-1 X(0)
2
1
)0(,32
10
XAWe were giventhat
Hence we can
write:
32
1
32
10
0
0
s
s
s
sAsI
-
8/22/2019 State space representation Part-1
38/47
tt
tt
ee
ee
tx
tx
ss
sss
s
ss
s
s
XAsI
2
2
2
1
2
1
64
34
)(
)(
)2)(1(
22)2)(1(
5
23
2
1
2
13
)0(
We can check the solution whether it is correct by applying the initialconditions: x1(0) =1, x2(0) =2, 1 = 1 and 2 =2: which implies as correct.
32
10A
In the same example if initial conditions are not given:
The state transition matrix is defined as: Atet)( -1 (s IA)-1
-
8/22/2019 State space representation Part-1
39/47
32
1
32
10
0
0][
s
s
s
sAsI
The inverse of (sIA) is given by:
)2)(1()2)(1(2
)2)(1(
1
)2)(1(
3
2
13
)2)(1(
1][ 1
sss
ss
ssss
s
s
s
ssAsI
Atet)( -1 (s IA)-1Since:
)(tWe can get thestate transition
matrix:
tttt
ttt
eeee
eeee22
221
222
2
-
8/22/2019 State space representation Part-1
40/47
We have learned how to construct signal flow graphs (SFG) for simultaneous
equations in lecture -5: The signal flow graphs which represent the state
equations or differential equations are called state diagrams. State diagram is anextension of SFG to portray state equations and differential equations.
Given the state diagram it is much easier to construct state equations or TF than
from SFG. Construction procedures and rules are same for SFG and State
diagrams but, state diagram is in Laplace domain.
State Diagram
State Diagram for State Equations:
Consider the control system:
4954
3853
26237
12352
321
3213
.
3212
.
3211
.
eqxxxy
eqrxxxx
eqrxxxx
eqrxxxx
-
8/22/2019 State space representation Part-1
41/47
R(s) sX3(s) X3(s) sX2(s) X2(s) sX1(s) X1(s)
Y(s)
Initially let us define three state variable x1, x2, x3 and let these three state
variables are represented as nodes and to the left of each node, respective
derivatives of them are also shown as below:
Let us define the input as r and the output as y.
In Laplace domain we can show the derivatives and nodes as shown below:
In the next step, we interconnect the state variables and their derivatives
with the definition of integration (1/s):
sX3(s) X3(s)
1/s
Integratio
n
-
8/22/2019 State space representation Part-1
42/47
12352 3211.
eqrxxxx
State diagram for equation -1:
R(s) sX3(s) X3(s) sX2(s) X2(s) sX1(s) X1(s) Y(s)-51/s
2
1/s1/s
2
3
-
8/22/2019 State space representation Part-1
43/47
R(s) sX3(s) X3(s) sX2(s) X2(s) sX1(s) X1(s) Y(s)
R(s) sX3(s) X3(s) sX2(s) X2(s) sX1(s) X1(s) Y(s)
6
1/s
21/s 1/s
-7
-3
262373212
.
eqrxxxx
1/s8
1
1/s-5
-3
1/s
3853 3213.
eqrxxxx
-
8/22/2019 State space representation Part-1
44/47
R(s) sX3(s) X3(s) sX2(s) X2(s) sX1(s) X1(s) Y(s)
321 954 xxxy
State diagram for equation -1:
5
1/s
9
1/s 1/s
-4
-
8/22/2019 State space representation Part-1
45/47
R(s) sX3(s) X3(s) sX2(s) X2(s) sX1(s) X1(s) Y(s)
Now we have to combine all figures to get one state diagram for all equations:
3
1
8
5
1/s
9
1/s 1/s
-4
6
2
-3
-5
2-5
2
-7-3
Homework: check the numbers and the branches
-
8/22/2019 State space representation Part-1
46/47
End
Characteristic Equation From State-Space Representation:
-
8/22/2019 State space representation Part-1
47/47
The roots of the characteristic equation are the same as the eigenvalues of
matrix A. The characteristic equation of matrix A can simply be computed
with Matlab by using the command :
p = poly(A);
or p = poly(A,v); %retun as a variable v