Control Systems m i l e r p r a in r e v y n is o Frequency domain analysis L. Lanari Monday, November 3, 2014 outline • • • • introduce the Laplace unilateral transform define its properties show its advantages in turning ODEs to algebraic equations define an Input/Output representation of the system through the transfer function • • exploring the structure of the transfer function solve a realization problem Lanari: CS - Frequency domain analysis Monday, November 3, 2014 2 Laplace transform L f (t) F (s) t∈R complex valued function in the complex variable s∈C F (s) = for f with no impulse at t = 0 � ∞ 0 f (t)e−st dt Laplace (unilateral) transform F (s) = L[f (t)] Region of convergence Lanari: CS - Frequency domain analysis Monday, November 3, 2014 3 Laplace transform Laplace transform of the impulse F (s) = L[δ(t)] = � ∞ � ∞ f (t)e−st dt 0− δ(t)e−st dt = 0− � ∞ δ(t)e−st dt = e−s0 = 1 −∞ Linearity L[αf (t) + βg(t)] = αL[f (t)] + βL[g(t)] Lanari: CS - Frequency domain analysis Monday, November 3, 2014 4 Inverse Laplace transform L f (t) F (s) L−1 1 −1 f (t) = L [F (s)] = 2πj � α−j∞ F (s)est ds α−j∞ f (t) defined for t ≥ 0 unique for one-sided functions (one-to-one) or f (t) = 0 Lanari: CS - Frequency domain analysis Monday, November 3, 2014 for t<0 5 Laplace transform Derivative property � � df (t) = sL[f (t)] − f (0) L dt can be applied iteratively � � L f¨(t) = s2 L[f (t)] − sf (0) − f˙(0) very useful in model derivation Lanari: CS - Frequency domain analysis Monday, November 3, 2014 6 LTI systems also useful here x(t) ˙ = Ax(t) + Bu(t) (proof): linearity + derivative property algebraic solution X(s) = (sI − A)−1 x0 + (sI − A)−1 B U (s) and therefore Y (s) = C(sI − A) −1 Lanari: CS - Frequency domain analysis Monday, November 3, 2014 � x0 + C(sI − A) −1 � B + D U (s) 7 LTI systems state ZIR transform � �� state ZSR transform � � �� � X(s) = (sI − A)−1 x0 + (sI − A)−1 B U (s) x(t) = eAt x0 + � t eA(t−τ ) Bu(τ )dτ 0 and therefore, comparing � L e � L e at � 1 = s−a At � = (sI − A)−1 1 L [δ−1 (t)] = s Heaviside step function δ−1 (t) 1 for t ≥ 0 0 for t < 0 1 t Lanari: CS - Frequency domain analysis Monday, November 3, 2014 8 LTI systems y(t) = CeAt x0 + = = Convolution theorem L 0 w(t − τ )u(τ )dτ W (s) = C(sI − A)−1 B + D transform �� t w(t) = CeAt B + D δ(t) with Y (s) � C(sI − A) � x0 + C(sI − A) C(sI − A)−1 x0 + W (s) U (s) t 0 −1 −1 � B + D U (s) � w(t − τ )u(τ )dτ = W (s) U (s) being L[δ(t)] = 1 the output response transform corresponding to u(t) = ±(t) is indeed W(s) same for H(s) Lanari: CS - Frequency domain analysis Monday, November 3, 2014 9 Transfer function yZS (t) = Input/Output behavior x0 = 0 (ZSR) � t 0 w(t − τ )u(τ )dτ YZS (s) = W (s) U (s) Transfer function W (s) Input/Output behavior = C(sI − A)−1 B + D = YZS (s) U (s) = L [w(t)] independent from state choice? independent from state space representation? Lanari: CS - Frequency domain analysis Monday, November 3, 2014 10 Transfer function z = Tx (A, B, C, D) det(T ) �= 0 x˙ = Ax + Bu z˙ y = Cx + Du y � = T A T −1 A �=TB B � B, � C, � D) � (A, = = � = C T −1 C � + Bu � Az � + Du � Cz � =D D � � −1 B �+D � W (s) = C(sI − A)−1 B + D = C(sI − A) For a given system, the transfer function is unique Lanari: CS - Frequency domain analysis Monday, November 3, 2014 11 Shape of the transfer function W (s) = C(sI − A)−1 B + D inverse through the adjoint (transpose of the cofactor matrix) 1 C (adjoint of (sI − A)) B + D det(sI − A) cofactor(i, j) = (−1)i+j minor(i, j) polynomial of order n - 1 polynomial of order n - 1 1 T C (cofactor of (sI − A)) B + D det(sI − A) polynomial of order n strictly proper rational function: proper rational function: Lanari: CS - Frequency domain analysis Monday, November 3, 2014 rational function (strictly proper) degree of numerator < degree of denominator degree of numerator = degree of denominator 12 Shape of the transfer function W (s) = strictly proper rational function + D proper rational function N (s) W (s) = D(s) N (s) W (s) = D(s) D=0 strictly proper rational function D �= 0 proper rational function zeros poles roots (for coprime N(s) & D(s)) i.e. no common roots from previous analysis the poles are a subset of the eigenvalues of A {poles} Lanari: CS - Frequency domain analysis Monday, November 3, 2014 ⊆ {eigenvalues} more on this later 13 Laplace transform (other properties) Integral property L �� t 0 � 1 1 f (τ )dτ = L [f (t)] = F (s) s s Time shifting property L [f (t − T )δ−1 (t − T )] = e−sT L [f (t)δ−1 (t)] = e−sT F (s) N(s) and D(s) are said to be coprime if they have no common factor Lanari: CS - Frequency domain analysis Monday, November 3, 2014 14 Laplace transform tables δ(t) 1 δ−1 (t) 1 s e at tk k! tk at e k! 1 s−a 1 sk+1 1 (s − a)k+1 sin ωt 1/2j ω 1/2j = − 2 2 s +ω s − jω s + jω cos ωt 1/2 s 1/2 = + 2 2 s +ω s − jω s + jω sin(ωt + ϕ) s sin ϕ + ω cos ϕ s2 + ω 2 eat sin ωt ω (s − a)2 + ω 2 at e cos ωt Lanari: CS - Frequency domain analysis Monday, November 3, 2014 (s − a) (s − a)2 + ω 2 15 Realizations (A, B, C, D) ok W (s) = C(sI − A)−1 B + D how ? W (s) = C(sI − A)−1 B + D • • • • realization infinite solutions (A, B, C, D) state dimension ? how can we easily find one state space representation (A, B, C, D) ? may be complicated for MIMO systems (here SISO) we see only one, obtainable directly from the coefficients of the transfer function (others are obtainable by simple similarity transformations) Lanari: CS - Frequency domain analysis Monday, November 3, 2014 16 Realizations N (s) W (s) = D(s) coprime N(s) & D(s) bn−1 sn−1 + bn−2 sn−2 + · · · + b1 s + b0 W (s) = n +D n−1 n−2 s + an−1 s + an−2 s + · · · + a1 s + a0 - find D - state with dimension n - one possible realization for A, B, C Ac = Cc = � 0 0 ··· 0 −a0 b0 1 0 ··· 0 −a1 b1 b2 0 1 ··· 0 −a2 ··· ··· ··· ··· ··· ··· bn−1 Lanari: CS - Frequency domain analysis Monday, November 3, 2014 0 0 ··· 1 −an−1 � 0 0 .. . Bc = 0 1 Controller canonical form (useful for eigenvalue assignment) 17