Validation of ANN-based model for binary distillation column

Abstract

An artificial neural network model for Binary Distillation Column (BDC) is presented in this work. The recurrent neural networks have been used for the modeling to represent the nonlinear behavior of distillation process. The data for neural network training has been acquired from continuous BDC setup available in laboratory. The available model contains nine trays. The neural network model is composed of two layers. The activation function chosen for the first layer is a hyperbolic tangent sigmoid function, whereas a pure linear function is utilized as activation functions in the second layer. The validation of developed neural network based model has been done by an extensive data set of real-time data acquired from the BDC set up.