Does anybody know what might be the best algorithm to determine the derivative component used in a PID controller?

I would really appreaciate any help!

Thanks.

 

I don't know about an algorithm, but for manual tuning, you generally don't need a derivative component unless the system exhibits a large amount of inertia. In that case, use trial and error to adjust the other components first, leaving the derviative component at zero. Then gradually bring up the derivative component to suppress overshoot.

Robert Scott
Real-Time Specialties
Embedded Systems Consulting

 

You can use a number of "canned" tuning approaches (just search the web).

Cohen-coon
Ziglar-Nichols
IMC (Internal Model control)

These are canned formula based approaches and are very effective for linear processes.

Most chemical processes don't require a derivative term (PI normally is sufficient).

An example of where a "D" term is useful:

Many thermal processes, for example, exhibit a significant process lag (delay between the application of control and seeing its effect in the PV movement) and a "+D" term is sometimes used to push the process faster toward the SP.

The "D" term can be problematic because it reacts to noise and other spurious signals (like motor coupling). To avoid the negative effects, use a filter on the PID or use an equation that that filters "D" through the process.

You will sometimes require a “D” term for an open loop unstable process, but these are the exception rather than the rule.

 

As mentioning above, most chemical process require only PI for getting the stability. How about the controlling of a DC motor?

What is normally sufficient for getting the stability in controling a motor?

agus