Emotional Learning Based Controller for Quadruple Tank System—An Improved Stimuli Design for Multiple Set-Point Tracking

Accurate set-point tracking across a broad operating region with minimum transients is a crucial requirement during the start-up and shutdown phases of a process plant operation. The emotional learning based controller, with its model-independent and nonlinear features, is extremely suitable for plant-based operations. However, the design of most of the existing emotional-learning strategies is based on narrow operating regions, and, hence, they may not yield satisfactory tracking performance when operated on a broader region. In this correspondence, a new design approach is adopted, which takes into consideration the specific requirements of multiple set-point tracking. This approach fuses a regression model with a constrained optimization problem to generate stimuli weights as well as the learning and inhibition parameters. Experimental validation of the proposed method on a quadruple tank system and comparison of its performance with an optimized second-order sliding mode controller illustrates its effectiveness.