Let's go one step at a time
Thanks Peter for your time and effort.
What you have done is exactly the same to what I have done. You use Z-transform approach to get A and B constants in your PV equation using ZOH method or method of step invariance how it is also called. (It's called step invariance vecause you matching step response of continuos system to response of discrete system in sampling points). I used state space approach and we got exactly the same equation (update time is 1 sec rather then 1 minute). That's good, very good. Also in my work you can find proof for Ac and Bc using state space.
You have done just like me, first you have found equation by assuming there is no time delay.
First thing that is little unclear is CO when introducing delay. In your article from controlguru, you wrote CO(n-thetap) (in the begining, but later use n*thethap), and in this pdf you wrote thetap/T. Since it is discrete equation
I agree there should be CO(n-trunc(thetap/T)). I do understand that this is only approximation with trunc function that has no big influence because theta is much larger than T. Theory in book about discrete system with dead time is full of advanced math, but we can leave it this way.
So, until part with headline „Steady State Compensation. THIS IS NOT IN THE TEXT BOOKS!!!!!!“ everything is more or less same like I did in my pdfs. You described it and it makes perfect sense why is needed to have that C part, so I'm accepting this. Unfortunately many of real time things aren't described in textbooks, I'm living proof, I have very good grades, but it seems I'm going too deep into theory trying to find explanations with constraints I found in textbooks and simply have no sense what some formula means in real life.
I'm becoming frustrated because, althoug I make good controller (PID, or state feedback) in Matlab, when specifications are stated, for example, design controller for system described with G(s) so that output (step response) has steady state error zero, overshoot <10%, settling time <3s, when I face real time problem, I'm not able do solve it on my own.
However I need formal method to apply to wide range of processes and I found that
www.controlguru.com website, where is shown how to obtain model (FOPTD) from real process and how to calculate PID gains.
Now I need to fully understand everything that is going on.
PVss=140-39*Kp // STEADY STATE TEMPERATURE
You wrote this:
From the
heat exchanger data plot, we can see that the heat exchanger temperature (PV) is 140 °C degrees when the controller output (CO) is 39%. If these values are plugged into the formula y=mx+b, we can solve for the steady state temperature, PVss:
140 = 39Kp + PVss
I don't understand how. If Kp is process gain, then you assumed linear relationship:
PV = Kp*CO+PVss and since 39% of CO will give 140 deg you got this relationship:
140 = 39Kp + PVss. Since Kp is process gain and it's Kp = -0.53, this means that PVss will be 160.7 deg???? What is conclusion of this? Temperature is system's output, so steady state value means that system will have this value after some specific amount of time. So one can conclude that when valve is opened 39%, after some time, output temperature will have have value of 160 deg. But that doesn't happen. Steady state temperature is 140 not 160 Temperature 160 will be when CO goes to 0%. Maybe steady state is not the right therm, it's better to call it offset or something.
Also, you assumed linear relationship y = mx+b? I understand this is an approximation and because of nature of system it's likely that PVss will be little different than 160 deg when CO is 0%. Relationship is more exponential than linear but I assume thi gives good results in practice, right?
Please, explain this!