Design of Radial Basis Function with PI-Based Supervisory Neural Controller for Liquid Level System

This paper introduces a radial basis function or RBF neural network (NN) for designing a supervisory controller (SRBF) with the combination of discrete-time proportional-integral (PI) controller for a laboratory scale single tank and a coupled tank liquid level system. Here, an RBF-NN is used as a feed forward controller with the combination of discrete-time PI controller in a closed loop structure to develop this SRBF-PI controller. This neural controller updates its weight by applying the gradient descent method to track the set level of water. The simulation results from MATLAB show that this SRBF-PI controller furnishes better performance to track the level than the discrete-time PI controller for the single and coupled tank system. To overcome the slow response of MATLAB for handling a large amount of data, the Google Research Colaboratory platform is used here to design the discrete-time conventional controller for both plants.