DELAY PROCESSING IN NETWORKED CONTROL SYSTEMS USING SMITH PREDICTOR AND ANN BASED ERROR PREDICTOR

Abstract

Networked control system is essentially a closed ioop system with a plant and controller, with the plant located in a remote area and controlling done through data communication networks. When the network used is shared like Ethernet, time delay is unavoidable. Also delay is stochastic because time delay depends on multiple factors. This depicts a nonlinear plant with stochastic time delay. Three models are developed and compared for processing the delay in NCSs; a reference model which is simple, an existing model and the indigenous model from novel method of error prediction. Fuzzy logic controller is the best option as controller amongst the conventional controllers like P1, PlD being flexible and nonlinear. Promising robust control systems are possible with Fuzzy Logic Controller in combination with new adaptive systems. This report presents comparison of three such combinations. Neural network is used to make Fuzzy Logic Controller adaptive; however the method used is new. With the help of artificial neural network, a concept of error predictor is introduced. Classical method for time delay processing is Smith predictor which is also implemented here and compared with the novel ANN error predictor. Time delay can go up to 600ms, hence the operating range for simulations is between Oms and 600ms. Plant selected to run the simulations is an AC 400W servomotor. Simulation results favors ANN error predictor as it can handle up to 600ms time delay.