Identification of Hollow Fiber Membrane Manufacturing System Incorporating Artificial Neural Network

An appropriate model is required for the purpose of estimation ae well as control which is obtained from system identification methods. A number of approaches are available for the linear systems however, the complex plant with unknown structure in the real word are frequently nonlinear systems. Overall porosity of hollow fiber membranes (HFMs) is the main key factors to evaluate the membranes performance and their specific applications. Hence, this study aims to introduce attractive and convenient method prediction the overall porosity of the membranes. System identification method was used to obtain appropriate model for predicting the mentioned parameters. Different polyvinylidene fluoride (PVDF) membranes were fabricated under various polymer compositions and spinning conditions. The fabricated HFMs were examined in terms of water pycnometry method to find the membranes porosity experimentally. The neural network training algorithm is based on the least square error and developed based on the Levenberg–Marquardt method. The predicted overall porosity of the membranes was compared with the actual values achieved from the experimental test. There was no significant difference between the results of both methods confirming the applicability of ANN for the study of HFMs overall porosity. This work presents a novel approach in order to evaluate the overall porosity of HFM in different range.