Optimization of bioreactor profile control

Abstract

The use of Nonlinear Autoregressive - moving average (NARMA) neuro controller for temperature control and two degree of freedom (2DOF) PID controllers also known as the ISA-PID controllers for pH and dissolved oxygen control of a biochemical reactor has been presented in this paper. The process model for yeast fermentation has been described in terms of temperature, pH and dissolved oxygen. The NARMA neuro controller has been trained by Levenberg-Marquardt training algorithm. The ISA-PID controllers have been tuned by MATLAB's auto tune feature along with manual tuning. Random profiles have been used for simulation. The performance of the controllers for following desired random profiles has been shown in contrast to the anti-windup PID (AWU-PID) controllers. The NARMA neuro controller and the ISA-PID controllers have shown overall improvement in controlling the process.