What I want: A Autotuning for a servo control

The Process: The control signal is the current of the motor. The motor has an encoder which is used for the field oriented control and additionaly for position control.

The control loop: there is one PID controler for the position control. The Encoder delivers the real position and the error is calculated and input to the PID. The PID outputs the requested current for the motor.

The used Autotuning method: I make a step response experiment. 10% of the max. current is given to the motor and I sample the encodercounts as distance. This process has a PI1-behaviour: The distance increases quadratic until the max speed is reached and the the slope is constant.

With the values slope of the curve, intersection of the tangent with the time axis and the control input to the process its possible to get PID Parameters with the method of CHIEN-HRONES und RESWICK. This procedure delivers also Parameters for integrating processes, like this one.

The Problem: The Parameters for the PID are extremely weak, so no good control is reached. If I tune the PID manually, then I get a proportional part, which is 100-times stronger. Similarly its with the I and D part.

Question: Whats the reason? I am sure I calculated the parameters correct. Who knows the reason and what kind of procedure delivers better results?

Best regards

Wolfgang

 

I've never had too much luck with autotuning.  You need just the right input data, and it is just easier to manually tune than try to calculate the proper data for the tuning algorithm.