Structure Identification of Adaptive Network Model for Intensified Semibatch Process

Methods for the application of an adaptive network model are investigated for the estimation of product properties in a semibatch process. The semibatch process exhibits nonlinear behavior, although the process is intensified by inlet flow rate scheduling (IFRS). In the present article, an estimation model based on the adaptive-network-based fuzzy inference system (ANFIS) is considered to flexibly deal with the multiplicity of the semibatch process. We focus on the structure identification of the ANFIS model, and thus propose a model structure where multiple membership functions are set with respect to a measured variable. Then, the adoption of the subtractive clustering method (SCM) is investigated for the determination of initial forms and the number of membership functions. This method results in improving the estimation performance, whereas poor robustness to change in the constant characteristic time in IFRS is seen. Thus, we have come up with the idea of the cascade mode for modifying the model structure. By using the monomer conversion as a cascaded variable, the model structure in the cascade mode is demonstrated to enhance robustness to disturbance and multiplicity in the intensified semibatch process.