22 Frequency Response for Linear Systems
Transcript
22 Frequency Response for Linear Systems
210 ME 132, Spring 2005, UC Berkeley, A. Packard 22 22.1 Frequency Response for Linear Systems Theory for Stable System: Complex Input Signal Consider the linear dynamical system ẋ(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) (99) We assume that there are n states, m inputs, and q outputs (so A ∈ Rn×n , B ∈ Rn×m , C ∈ Rp×n , D ∈ Rp×m ). If the system is stable, (ie., all of A’s eigenvalues have negative real parts) it is “intuitively” clear that if u is a sinusoid, then y will approach a steady-state behavior that is sinusoidal, at the same frequency, but with different amplitude and phase. In this section, we make this idea precise. Take ω ≥ 0 as the input frequency, and (although not physically relevant) let ū ∈ C m be a fixed complex vector. Take the input function u(t) to be u(t) = ūejωt for t ≥ 0. Then, the response is x(t) = eAt x0 + 0t eA(t−τ ) Bu(τ )dτ R = eAt x0 + eAt 0t e−Aτ B ūejωτ dτ R = eAt x0 + eAt 0t e(jωI−A)τ dτ B ū R Now, since A is stable, all of the eigenvalues have negative real parts. This means that all of the eigenvalues of (jωI − A) have positive real parts. Hence (jωI − A) is invertible, and we can write the integral as x(t) = eAt x0 + eAt Rt 0 e(jωI−A)τ dτ B ū ¯t = eAt + eAt (jωI − A)−1 e(jωI−A)τ ¯¯ B ū h 0 i = eAt x0 + eAt (jωI − A)−1 e(jωI−A)t − I B ū h i = eAt x0 + eAt e(jωI−A)t − I (jωI − A)−1 B ū h i = eAt x0 + eAt ejωt eAt − I (jωI − A)−1 B ū h i = eAt x0 + ejωt − eAt (jωI − A)−1 B ū h i = eAt x0 − (jωI − A)−1 B ū + (jωI − A)−1 B ūejωt Hence, the output y(t) would satisfy h i y(t) = CeAt x0 − (jωI − A)−1 B ū + C (jωI − A)−1 B ūejωt + Dūejωt In the limit as t → ∞, the first term decays to 0 exponentially, leaving the steady-state response h i yss (t) = D + C (jωI − A)−1 B ūejωt 211 ME 132, Spring 2005, UC Berkeley, A. Packard Hence, we have verified our initial claim – if the input is a complex sinusoid, then the steadystate output is a complex sinusoid at the same exact frequency, but amplified by a complex gain of D + C (jωI − A)−1 B. The function G(jω) G(jω) := D + C (jωI − A)−1 B (100) is called the frequency response of the linear system in (99). Hence, for stable systems, we have proven u(t) := ūejωt ⇒ yss (t) = G(jω)ūejωt More precisely, this is the frequency response from u to y, so we might also write G yu (jω) to indicate what is the input (u) and what is the output (y). Gyu can be calculated rather easily using a computer, simply by evaluating the matrix expression in (100) at a large number of frequency points ω ∈ R. 22.2 MIMO Systems: Response due to real sinusoidal inputs In the case where the system has multiple inputs and outputs, it is a bit more complicated to write out the response due to sinusoidal inputs at a fixed frequency, since the different inputs may all have different magnitudes and phases. Suppose that there are m inputs, and q outputs (so B ∈ Rn×m , C ∈ Rp×n , D ∈ Rp×m ). Take a ∈ Rm , b ∈ Rm , and ω ≥ 0. Consider the input u(t) = a cos ωt + b sin ωt Note that this is the real part of a complex input, namely h u(t) = Re (a − jb)ejωt i Hence, the steady state output must be the real part of a function, specifically, h y(t) = Re G(jω)(a − jb)ejωt i So, in summary: To determine the steady-state response due to an input u(t) = a cos ωt + b sin ωt, let c ∈ Rq and d ∈ Rq be real vectors so that c − jd = G(jω)(a − jb) Then, the steady-state output is yss (t) = c cos ωt + d sin ωt ME 132, Spring 2005, UC Berkeley, A. Packard 22.3 212 Experimental Determination Since the frequency response has the interpretation as a representation of the frequencydependent amplitude gain and phase shift between a sinusoidal input and steady-state output, it is easy to obtain experimentally for a given physical system. This can be done by performing several different forced-response experiments, with the forcing being a sinusoid, each experiment using a different frequency.