Application of Neural Network based Control Strategies to Binary Distillation Column

amit singh, Barjeev Tyagi, Vishal kumar

Abstract


This paper presents three different neural network based control schemes to the control of the Distillate composition of binary distillation column. The main goal is to control a single output variable, the Distillate composition, by changing two manipulated input variables, reflux flow rate and steam flow rate. A first-principle equation based model of binary distillation column is developed in SIMULINK® and validated by the experimental results. This model is used here as a reference model on which the developed neural control schemes have been applied. Three approaches Neural Network based Direct Inverse control (NN-DIC), Neural network based model reference adaptive control (NN-MRAC) and Neural network based internal model control (NN-IMC), are simulated and their performances are assessed. Comparison was also made with conventional PID cascade control. The results demonstrate that NN-IMC strategy provides a better performance than PID, NN-DIC and NN-MRAC for the cases analyzed.


Keywords


Direct Inverse Control, Feedforward networks, Internal model control, model reference adaptive control, Distillation Column

Full Text: PDF