FLOW & LEVEL CONTROL USING ANN BASED ADAPTIVE CONTROLLER

Abstract

Process controllers are used with adjustable settings in process industries. In the past few decades; digital adaptive controllers have been extensively used. The applications of neural network have drastically changed the performance of these controllers. In this dissertation, the work has been carried out on four ANN based adaptive controllers namely Direct Neural Adaptive Controller (DNAC), DNAC with Linear feedback compensator (LFBC),Direct Neural Model Reference Adaptive Controller (DNMRAC) and DNMRAC with fixed Gain Controller (FGC). Major work has been carried out on DNAC type of controller. For testing the performance of this controller, both simulated model as well as real time process have been used. For experimental study, a hardware setup of CST (continuously stirred tank) process is used, in which the control of temperature is being carried out by varying the flow rate of coolant. A multi layered neural network, (one input layer, one hidden layer and one output layer) has been used. Error Back Propagation training has been used for learning of the controller. The learning of the controller has been done online. An algorithm for the optimal selection for fixed parameters of the neural network has been used which gave satisfactory performance in different type of controllers. The performance curves are given under different process conditions. The performance comparison has also been carried out for all the controllers and the conclusions are drawn on the basis of the outcome of work in respect of each controller