Data-Driven Soft Sensing of Polyoxymethylene Melt Index using Takagi-Sugeno Fuzzy Neural Network and Quantum Genetic Algorithm

As a frequently used polymer material, polyoxymethylene (POM) has played an indispensable role in a variety of industrial fields. For production efficiency and safety, online monitoring of POM melt index is of importance in industrial practice. However, current data-driven soft sensors for POM melt index cannot address the pure delays within process variables. In this work, we propose time-delayed Takagi-Sugeno (T-S) fuzzy neural network as a novel data-driven soft sensor model, and improved quantum genetic algorithm (IQGA) for parameter estimation. By considering time-delayed process variables, the predictive performance and interpretability of soft sensor models can be improved. However, it is computationally cumbersome to find an optimal choice of time delays, and thus IQGA is proposed to resolve this problem. Case studies on industrial data from a real-world POM plant illustrate the usefulness of the proposed soft sensing method for POM melt index prediction.