Multiple Input-Single Output (MISO) Feedforward Artificial Neural Network (FANN) Models for Pilot Plant Binary Distillation Column

Abstract

Distillations column control becomes the main subject of control research due to the intensive energy usage in the industry and the nonlinearity behavior in control variables. The growing importance of "green technology" and sustainability has triggered researchers to focus on this matter. Therefore, a method of modeling and controlling of the column is certainly indispensible in this matter. Neural networks are a powerful tool especially in modeling nonlinear and intricate process. Hence, in this paper Feed forward Artificial Neural network (FANN) have been chosen to model the multiple input-single output (MISO) for the distillation column predicting top and bottom composition. The performance and the accuracy of the models have been presented in term of correlation coefficient (R value) and the smallest sum squared error (SSE). It has been found that FANN can model MISO in representing the process. The results obtained also show that the MISO model is suitable to be used to represent the distillation process accurately.