CONTROLLER DESIGN FOR CONTINUALLY STIRRED TANK HEATER

Abstract

Industrial process control always needs a control algorithm that can adapt itself in changing environment of industries in terms of its parameters. That means, it can control and coordinate the process over wide range of operating conditions. The famous conventional control theory is the basis for simple automatic control system using P, PI, P11) controllers. It has a great deal of advantages like its easily comprehensibility, relatively simple implementable scheme and anywhere applicability and with these features in the wings, it is flying so high and popular in the sky of industrial control. To compensate the parameter variations and as well as to adapt the environmental changes, controllers are tuned after control system has been installed using trial and error algorithm after getting an initial value using any of simple stability checking tools. Now, for significantly changes of parameters controllers must return to its back state to obtain satisfactory control. And this every time tuning of these controllers is a difficult part of this conventional control algorithm. One more different problem is we must model the plant for these controllers and plants are not always so simple to be modelled mathematically, its linearization is sometime computationally impractical. But herein, we have. used this popular class of controller to control our CSTH in spite of its drawbacks while handling high dynamic and complex plants. Thus, the conventional controllers are very sensitive to variation in parameters and have drawbacks in approximating the plant properties. In order to get a relatively perfect controller, work has to be done to find a new intelligent controller and that is based on Neural Network Training. In an intelligent group of controller, once expertise has done its work well, controller has in-built properties of adaptation and learning and decision making capability to meet the required performance specification over a wide range of change in parameters and operating conditions. Here, used ANN controller can be summed up in following properties: 1.They acquire the plant properties through non-linear mapping with an algorithm of having minimum of errors in doing so. 2. They don't require modelling of plant before designing control strategies, they do it by expertise. 3. As they depend on observance of data collected out of test done on plant, control strategies don't get affected if we miss any knowledge of dynamics of plant and do not pose any hindrance in development and implementation of control strategies. 4. Adaptation and learning are obtained off-line and on-line iv weight adaptation to keep the network updated to current environment, and 5. Artificial Neural Network processes multiple numbers of inputs and outputs and thus is readily available for and kind of system. - Until now, what we come to know of Neural Network is much of on paper information, we have to study it through its various aspects and need to modify them to implement it in real time process control applications. In this work, both the controllers, conventional PI and ANN controllers have been implemented to same problem of continually stirred tank heater and comparisons are made to select a better controller to meet a particular set of desired performance specification. Both can be useful and have separate significances for the different task intended to perform. The subject process CSTH has got wide application in this industrial era, particularly in chemical and process industries. For example, CSTH may be a reacting vessel where two or more reactants are mixes to carry out a particular reaction which demands a fix temperature for its occurrence. Hence, attainment of this desired level of temperature to serve the reaction process is our main control objective. Reaction can be of any type, sometimes it will sink the heat out of its surrounding (endothermic process) and other time it sources the environment with its unleashed energy during its reaction, so this process requires an isolating medium with surrounding and same time which can be used as a controlling parameter for controlled variable temperature of tank. And here, a jacket is used to surround the tank in which the hot liquid is made to run through. The amount of liquid flowing into tank will decide its temperature, hence, according.to type of reaction and temperature requirement we regulate the inflow of hot fluid in jacket. Both ANN and Conventional PI have been used mainly to control the flow of hot liquid into jacket which in turn, affects the tank temperature. Now, strategies lie in understanding of how jacket temperature affect the tank's one and how variation in amount of inflow liquid and its initial temperature determine the jacket temperature and there both controllers have different algorithm and we observed each of their ways to control the system. v On the basis of comparison of the performances of both controllers, it can be stated as the developed controller ANN has effective results with a bit of more complexity and show high promise of its industrial application in future with even good modifications than that of old and popular conventional controller which has got some real limitations like coping with high dynamics of plant and increasing sluggishness with complexity of process.