Neural network-based control strategies for a two-CSTR-in-series system

The goal of this research was to develop an effective control law for a process consisting of two continuous-stirred-tank-reactors in series using neural networks. A total of five types of controllers were developed to show the improvement in control with controllers using neural networks over conventional controllers. Conventional feedforward, neural network feedforward, PID feedback, conventional feedforward plus PID feedback and neural network feedforward plus PID feedback were the five controllers designed in this project.; Based on the closed loop responses with the five controllers, it was concluded that the neural network feedforward controller gave better control in comparison to the conventional feedforward controller, and the neural network feedforward plus PID feedback controller outperformed the PID feedback and conventional feedforward plus PID feedback controllers. Therefore, based on the results from this research, the control strategy designed using neural networks showed improvement in the control for the process of two continuous-stirred-tank-reactors in series.