On-line Tuning Strategy for PI Control Algorithms Based on Simple Linear Models

On-line model-based tuning of the parameters of conventional PI control algorithms is considered. This is accomplished to satisfy the requirement of preset performance specifications. The specifications are defined as time-domain response envelopes, which are appealing to the practitioner. The parameter adaptation at each sampling point is accomplished via using a linear relationship between process outputs and PI tuning parameters. Thus the adaptation strategy directly uses the sensitivity of the closed-loop response to the parameters. In this paper, the model used for calculating the sensitivity is a first order model developed by a simple step testing of the plant. Such a model is very common in industrial practice. The efficiency of the method is presented through simulated implementation on two nonlinear reactor examples.