Comparison of Swarm Adaptive Neural Network Control of a Coupled Tank Liquid Level System

Abstract

This paper presents the use of neural network control approaches for a two inputs -two outputs (TITO) coupled tank liquid levels with disturbances effects and set-point changes in dynamic system. Hybrid PI-neural network (hybrid PI-NN) and PID neural network (PID-NN) controllers are the techniques used in this investigation to actively control the liquid levels of coupled tank system. Unlike traditional neural network weight adaptation using gradient descent method, Particles Swarm Optimization (PSO) is utilized for adaptive tuning of neural network weights adjustment and fine tuning the controller's parameters. A complete analysis of simulation results for each technique is presented in time domain. Performances of both controllers are examined in terms of disturbance rejection and control performance measures for common input changes. Finally, a comparative assessment on the impact of each controller on the system performance is presented and discussed.