Fuzzy Real-Time Optimization of the Tennessee Eastman Challenge Process

Abstract

<em>A Real-Time Optimization (RTO) strategy incorporating the fuzzy sets theory is developed, where the problem constraints obtained from process considerations are treated in fuzzy environment. Furthermore, the objective function is penalized by a fuzzified form of the key process constraints. To enable using conventional optimization techniques, the resulting fuzzy optimization problem is then reformulated into a crisp programming problem. The crisp programming problem is solved using both Sequential Quadratic Programming (SQP) and Heuristic Random Optimization (HRO) techniques for comparison purposes. The proposed fuzzy RTO strategy is demonstrated via the Tennessee Eastman benchmark process, and is also compared with a crisp RTO strategy from the literature. Remarkable economical improvement is found over the crisp RTO. In spite of the fuzzified constraints, the proposed strategy yields smooth operation of the process, while maintaining the product quality within the acceptable range.</em>