Abstract
Industrial practice navigates a robust controller tuning due to fairly uncertain process information. A comprehensive process identification is often avoided due to operating conditions, continuous manufacturing, etc. When process properties or operating conditions change, controller re-tuning is often ignored resulting in long-term lower productivity outcomes. We introduce a novel methodology to extract relevant process information from short and minimally rich datasets. Specifically, a sine test of duration related to settling time of the loop is superimposed on nominal process inputs. The method can be applied under both open and closed-loop operating conditions. The process impulse response coefficients are estimated and consequently provides a band limited, but accurate, frequency response. Based on this information, a robust PID controller can be (automatically) tuned. Numerical examples are provided to validate the theory. Additional experimental validation on an integrating and nonminimum phase process and on a poorly damped system is also provided. A comparison with one of the most widely used autotuning methods illustrates the relevance of the proposed approach.