Help with temperature control


Thread Starter


I am new to control system design. I am looking into implementing predictive control for the simple temperature control. I would appreciate if any body could shed some light on this. whether it is practically feasible or is it complex for this simple ON/OFF controls.

THanks for your suggestions and comments.
Well, if this is your first then the first thing you should learn is that nothing is predictable. Now that that is settled.

"Prediction" is based on using a 1st and sometimes higher order derivatives. I've actually seen some propose 8th order systems.

The problem with derivatives is that the higher you go the more susceptible to noise you become. Your tradeoff becomes Response to Disturbance and Noise Rejection. Check out a Bode plot and see how the derivative affects it. 2nd and 3rd order can be highly sensitive to noise.

In general, temperature is a great measurement for using "prediction" (if the measurement system has been properly designed) as opposed to Flow which generally stinks because of the noise.

Now that you understand the principles, you need to pick your "predictive" model. It depends upon the application. Actually, the D in PID is a predictive element of the first order. If you ain't satisfied with that, look at the Smith Predictor, a generalized higher order. In general, you really need to characterize your system well. Know what your disturbances are and where they come from. Know the characteristics of your signal noise. And most of all, know the process. High inertial systems (typical of oven systems) can be difficult to deal with and dangerous when applying predictors.

email me if you don't understand something.

[email protected]