I have to design a PI-controller with zero order hold for the plant $G_s(s) = \frac{0.24}{(s+4)(s+12)}$ to meet the specifications of $\omega_n=5\,s^{-1}$ and $\zeta = 0.6$.

I choose a $T=0.12\,s$ which leads to a approx. 13 samples per oscillation (and $T \leq 0.125 T_d$)and applied the Z.O.H. to $H_o(s)G_s(s)=G(s)\approx \frac{1}{1+0.5 T s}\frac{0.24}{(s+4)(s+12)}$.

With the transfer function of the PI-controller: $G_D(s) = K_c\frac{a + s}{s}$ I get the characteristic equation with $H_c(s) = \frac{G(s)}{1+G_D(s)G(s)}$ to $\frac{4s}{s^3 + 32.667 s^2 (314.672 + 4 K_c)s + 800.016 + K_c a 4}$.

With the desired system of: $\frac{1}{(s^2 + 2 \omega_n \zeta s + \omega_n^2)(s + \alpha)}= \frac{1}{s^3 + 32.6s^2 + 184.6 s + 665}$ where I choose $\alpha=26.6$ with $\alpha \approx 4 \ldots 5 \omega_n$ (hence, the $s^2$-term is equal).

Now I did the comparison of coefficents:

$314.672 + 4 K_c = 184.6 \rightarrow K_c=-32.518$

$800.916 + K_c a 4 = 665 \rightarrow a = 1.038 $

Theoretically this should be now a proper $K_c$ and $a$. But simulation shows undesired behavior (step response goes to infinity). Should I introduce another pole? Is the general approach right?

(Remark: I edited the problem considerable with help from Design of a PI controller.)

  • $\begingroup$ Why do you have to design a closed loop system with a given natural frequency and damping factor? Since usually you have bandwidth and phase margin specifications. Also why do you use $\frac{G}{1+CG}$ as closed loop and not $\frac{CG}{1+CG}$ (which I think is much more intuitive way of controlling)? $\endgroup$
    – fibonatic
    May 12 '16 at 23:27
  • $\begingroup$ It is for an exercise example so my guess the lecturer wanted to spare (us) the calculation of $\omega_n$ and $\zeta$. For the second question, I don't know which way is more common, and was using the common description of our course. $\endgroup$
    – Lageos
    May 13 '16 at 7:07
  • $\begingroup$ In that case I would assume that the dominant poles should have those $\omega_n$ and $\zeta$. You do not need to bother about zero-pole-cancellation in the open loop, just make sure that the two other poles die out faster. $\endgroup$
    – fibonatic
    May 13 '16 at 7:24
  • $\begingroup$ So the general approach - the values for $K_c$ / $a$ - are right? I don't see my mistake. $\endgroup$
    – Lageos
    May 13 '16 at 7:50
  • $\begingroup$ You should always check whether your results are correct when you get negative gains. $\endgroup$
    – fibonatic
    May 13 '16 at 8:25

I think you've made a mistake when setting up your characteristic equation. In MATLAB I get the following for your system transfer function:

enter image description here

for your system transfer function.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.