STABILITY OF LINEAR SYSTEMS

Dr K P Mohandas , Professor, NIT Calicut

1.INTRODUCTION One of the primary requirements to be satisfied by any system is that it should be stable. The notion of stability of systems can be defined in many ways. But in a general sense, a stable system should have an output which in many ways resembles the desired output of the system. Under the action of disturbances generated inside the system due uncontrollable phenomena or those acting from outside to disturb the output, the output should remain close to the desired value. In the case of regulator system where the output should remain almost constant over a wide range, the output should remain constant in spite of the disturbances. In the case of tracking systems, the output trajectory should remain close to the desired or nominal trajectory. There are several definitions of stability, particularly when dealing with nonlinear systems, but linear systems are defined to be stable or unstable based on the bounded input bounded output (BIBO) criterion. 2. BOUNDED INPUT BOUNDED OUTPUT STABILITY A linear system is said to be BIBO stable, if the output remains bounded for an input that is bounded. Mathematically, we may put If the input r(t) = M , output c(t) = N where M and N are real numbers such that 0 < M < 8 means that 0 < N < 8 3.METHODS OF DETERMINATION OF STABILITY OF LINEAR SYSTEMS Linear systems behave in a well defined fashion. The response of the systems due to initial energy stored in the systems (initial condition response, or transient response or zero input response) can be added to the forced response due to the input (input response, or zero state response) to find the total response of the system due to the validity of the superposition principle. In fact, a linear system which is stable will have the initial condition response decaying to zero as time t increases leaving the output that is due to the input alone to remain. For this to occur, the system should have its response due natural modes decreasing with time. This again can be guaranteed depending on the nature of the roots of the characteristic equation of the transfer function which are the same as the eigen values of the A matrix in the state space model. Let the transfer function of the system be: N ( s) G( s ) = ( s + a)( s + b)( s + c )(..)....( s + p) The impulse response then can be obtained as:
g ( t ) = k1e at + k 2e bt + ....... + k ne pt

------------------------------------------------------------------------------------------------------Stability of linear systems KPM/LSA/

where the ks are the partial fraction coefficients. Now, the nature of the impulse response can be summarized as follows: If any one of the poles of the transfer function (roots of the characteristic equation) at a, -b, .,-p are positive (if real) or have positive real part ( if complex), the corresponding term in the impulse response increases without bound resulting in the output remaining unbounded as time t tends to infinity. On the other hand, if none of the roots have positive real part, and all of them are on the left of the imaginary axis, (the roots are real and negative or complex with negative real part), the corresponding natural response will decay to zero as t tends to infinity. If a pair of roots is on the imaginary axis with zero real part, this will contribute to a sinusoidal oscillation with constant amplitude. This response can be considered as critically stable or marginally stable. In the presence of the input, the response will contain a term corresponding t the input which will be final output if the impulse response decays to zero. For example, with a unit step input r(t) = u(t),
C ( s) = G ( s ) R ( s) =

N (s) 1 . ( s + a )( s + b)( s + c)(..)....( s + p) s

Then c(t) can be obtained as :

Obviously, if the roots of the characteristic equation are all on the left of the imaginary axis, the output will be due to the first term only as t tends to infinity. This is true for other inputs as well. Hence, the determination of stability of linear systems essentially boils down to location of the roots of the characteristic equation. If it can be ascertained that if a system has none of the roots of the characteristic equation on the right half of the imaginary axis and all the roots are on the left half of the s-plane, the BIBO stability is assured. Because of this reason, most of the techniques for determinatio n of stability of linear systems essentially try to find the location of the roots of the characteristic equation. Some of the methods are: 1. 2. 3. 4. Routh Hurwiz Criterion Nyquist Criterion Bode Plot Method Lyapunovs method etc

Among these, the first three can be applied to linear time invariant systems whereas the last one is more general and can be applied to study the stability of linear and nonlinear systems.

------------------------------------------------------------------------------------------------------Stability of linear systems KPM/LSA/

4.ROUTH-HURWITZ CRITERION This criterion is derived from the theory of equations and is an algebraic method to determine the number of roots of a given equation with positive real part. It can be summarized as follows. Given an nth order algebraic equation with constant coefficients: a n s n + an 1 s n1 + a n 2 s n 2 + ......... + a1s + a 0 = 0 Then, the equation has no roots on the right half of the s-plane if the following necessary conditions are fulfilled: Necessary Conditions 1. All coefficients a n , a n 1 , a n 2 ........., a 1 , a 0 should be present , ie none of them should be zero 2. All the coefficients a n , a n1 , an 2 ........., a1 , a 0 should be of the same sign. If these necessary conditions (not sufficient) are satisfied, the sufficiency test can be conducted based on a Rouths table constructed as follows. Rouths tabulation

sn s n-1 s n2 s n2 . . s0

an an 1 b1 c1 . . x1

a n 2 a n 3 b2 c2 . .

a n 4 ................... a n 5 ................... b3....................... c 3....................... .

Here, the first two rows of the table are made from the given equation with the alternate coefficients being filled starting with the highest in the first row and the next highest in the second row. The third and higher rows are computed as follows: b1 = ( an 1 .a n 2 a n .an 3 ) / an 1 b2 = (a n1 .a n 4 an .a n 5 ) / an 1 .and so on Similarly, the coefficients in the next row are computed using the immediately previous two rows, viz

------------------------------------------------------------------------------------------------------Stability of linear systems KPM/LSA/

c1 = (b1.a n3 a n1 .b2 ) / b1 c 2 = (b1.an 5 a n1 .b3) / b1 and so on After completing all the rows designated as s n , s n 1 ,............., s 0 , then, check the sign of the elements in the first column of the table. The sufficiency condition is that the number of roots of the given equation with positive real part is equal to the number of sign changes in the first column of the Rouths table. If there are no sign changes, the n there are no roots of the characteristic equation on the right half of the s-plane. Example: Consider the equation whose roots are known such as ( s + 1)( s + 2)( s + 3)( s + 4) = 0 The given equation is re-arranged as: s 4 + 7s 3 + 17 s 2 + 17 s + 6 = 0 Necessary conditions: 1. All coefficients are present 2. All coefficients are of the same sign Both the necessary conditions are satisfied. Then, form the Rouths table: s4 1 17 6
s3 s
2

7 (7x17 - 1x17) / 7 = 14.57 (14.57x 17 - 7x 6)/ 14.57 = 14.11 6

17 (7x6 - 1x0) / 7 =6 0

s1 s0

It is seen that all the elements of the first column in the table are positive which implies that there are no sign changes. This means, by Rouths criterion that none of the roots of the given equation are with positive real part (which of course, we know at least for this equation) 5. SPECIAL CASES AND THE SOLUTION FOR THEM There are two special cases in which the simple Rouths test breaks down. These and the remedies are given below. Case 1. The first element in a row becomes zero, whereas some of the other elements are nonzero. ------------------------------------------------------------------------------------------------------Stability of linear systems KPM/LSA/ 4

Remedy 1. Replace the zero elements by a small positive number e and continue the tabulation. While determining the sign of the elements of the first column, find the sign as the value of e tends to zero. 2. Replace s in the given equation by s = (1/z) and do the tabulation in the equation in z , rather than in s. The information obtained will be the same as for the equation in s. Example : Consider the equation :
s 5 + s 4 + 2s 3 + 2s 2 + 3s + 5 = 0

Both the necessary conditions are satisfied. Form Rouths table

s5 s4 s s
3 2

1 1 0

2 2 -2 -2

3 5 < - - - test breaks down

The remedy is to replace the first element of the third row by e tabulation as follows s5 s4 s3 s2 s1 s
0

1 1 (2 + 2)/ 5

2 2 -2 -2 5

3 5

{-2. (2 + 2)/ + 2 } /{(2 + 2)/ }

The sign of the elements of the first column, we see that

> 0 ( as per assumption)

> 0 as in the limit

(2 + 2) 2 and

2/

a very large +ve

and {-2. (2 + 2)/ + 2 } /{(2 + 2)/ } tends to a negative quantity as it tends to -2 as >0

------------------------------------------------------------------------------------------------------Stability of linear systems KPM/LSA/

So the elements in the first column are +, +, +, +, - and + with the result that there are two sign changes and hence two roots of the equation on the right half of the s-plane. The second method also can be attempted. But sometimes, the second method will not solve the difficulty .

Case 2 When all the elements in a particular row become zero before completing the tabulation This happens when there are pairs of roots which are equal in magnitude but opposite in sign like when there are two roots on the imaginary axis or when two pairs of roots are symmetrically placed around the origin in a quadruple. Remedy Write down the equation using the coefficients immediately previous to the all zero row. This equation is called auxiliary equation. The roots which are of equal in magnitude and opposite direction can be obtained by solving this equation. To complete the tabulation, differentiate the auxiliary equation with respect to s once and use the coefficients in place of the all zero row. Example.
s 5 + 2 s 4 + 24s 3 + 48s 2 25s 50 = 0

Rouths array
s5 s4 s
3

1 2 0

24 48 0

- 25 - 50 0 < - - - All zero row

The auxiliary equation is : 2 s 4 + 48s 2 50 = 0 Differentiate this with respect to s giving 8s 3 + 96s In the Rouths table, the all zero row is replaced by these coefficients

s5 s s
4 3

1 2 0 8 24 112.7 - 50

24 48 0 96 - 50 0

- 25 - 50 0 < - - - coeffs from dA(s)/ds

s2 s1 s0

------------------------------------------------------------------------------------------------------Stability of linear systems KPM/LSA/

There is one sign change in the elements in the first column and therefore there is one root on the right half of the s-plane. Solving the aux equation A(s) = 2 s 4 + 48s 2 50 = 0 , which can be factorized as: ( s 2 + 25)( s 2 1) = 0 so that the roots are at s = +1, -1 and s = +5j and -5j Obviously both pairs are of equal magnitude, but opposite sign. The roots on the right half plane is identified as +1 . 6. APPLICATION OF ROUTHS CRITERION Rouths criterion can be applied to determine range of certain parameters of a system to ensure stability. For example, it is usually of interest to find the range of the open loop gain K for closed loop stability . Example: Consider the unity feed back control system with open loop K G( s ) = Determine the range of values for closed loop stability. s ( s + 30)( s + 40) Closed loop transfer function is : K G( s ) s ( s + 30)( s + 40) K M ( s) = = = 3 2 K 1 + G( s ) 1 + s + 70s + 1200s + K s ( s + 30)( s + 40) The characteristic equation is: s 3 + 70s 2 + 1200s + K =0 From necessary conditions K > 0 From Rouths table s3 1 1200

gain

s2 s1

70 ( 84000 - K) / 70

K 0

s0 K For all the elements in the first column to be positive (same sign) K < 84000. The range of K for stability is thus 0 < K < 84000
7.LIMITATIONS OF ROUTH-HURWITZ CRITERION 1. It gives only information about absolute stability of the system. The degree of stability of a stable system cannot be obtained 2. The Criterion can be applied only if the characteristic equation has constant coefficients and cannot be applied if they are not real or contain exponential terms as in the case of systems with dead time.

------------------------------------------------------------------------------------------------------Stability of linear systems KPM/LSA/