Development of a genetically tuned fuzzy dynamic model for nonlinear dynamical systems: Application to reaction section of Tennessee Eastman process

This work presents a new GA-fuzzy method to model dynamic behavior of a process, based on recurrent fuzzy modeling through Mamdani approach, whose inference system is optimized by genetic algorithms. By using the Mamdani approach, the proposed method surmounts the need to solve various types of mathematical equations governing the dynamic behavior of the process. The proposed method consists of two steps: i) constructing a startup version of the model and ii) optimizing the shape of membership functions of the fuzzy sets corresponding to the variables existing in the fuzzy model along with the production rules constituting the inference such that the obtained fuzzy model can predict the dynamic behavior of the process fairly accurately. The proposed method is used to predict the dynamic behavior of the reaction section of the Tennessee Eastman (TE) benchmark. The overall accuracy of the obtained results is shown in comparison with their corresponding counterparts in TE benchmark. The Mean Absolute Percentage Error (MAPE) of the key process variables, which are temperature, pressure, and level of the reactor, and the reactor cooling water outlet temperature are calculated 1.17%, 0.38%, 1.5%, and 1.57%, respectively, showing high prediction capability of the proposed method.