Algorithm self-tuning

Foxboro developed a self-tuning PID controller that is based on a so-called expert system approach for adjustment of the controller parameters. The on-line tuning of K, Xi, and Xo is based on the closed-loop transient response to a step change in set point. By evaluating the salient characteristics of the response (e.g., the decay ratio, overshoot, and closed-loop period), the controller parameters can be updated without actually finding a new process model. The details of the algorithm, however, are proprietary... 

Self-Tuning Algorithm This algorithm incorporates the method described just above with the added feature of... 

Consequently, the self-tuning algorithm would respond to the introduction of a setpoint change or another disturbance faster than the original model, illustrating the value of the self-tuning algorithm. 

Figure 25. Simulated response of third reactor of a continuous vinyl acetate polymerization to a step disturbance at high emulsifier feed concentration (0.06 mol/L H,0) and manipulation of initiator flow rate to the third reactor at 50°C ((-------j optimum PID) (--------) Z transform (XXX) self-tuning algorithm)...

Even starting off with poor initial controller parameter values, the self-tuning algorithm usually converges quite rapidly to a stable controller. After a certain amount of data has been collected one can test whether or not the assumed controller form is optimal (in the sense of the minimum variance) and then change it if necessary. 

Likewise, electronic and digital controls will demand more attention, depending on their vintage. Self-tuning algorithms were available in most systems in the late 1980s, but there have been improvements in the past few years, making them more reliable and less prone to performance problems at the extremes of their control band. Better controls mean less maintenance and repair costs. 

Figures 3.31 to 3.33 give the recommended tuning for the preferred algorithm (noninter-active, proportional-on-PV, integral-on-Zin, derivative-on-PV and no derivative filtering). It is assumed that the scan interval is small compared to the process dynamics. The mning is designed to minimise ITAE subject to a maximum MV overshoot of 15 % on a self-...

---