MODELLING AND SIMULATION OF A PRESSURE AND TEMPERATURE PROCESS

摘要

The present paper tries to deal with the problem of correct modelling a multivariable process (temperature/pressure). One of the most important facts to ensure the optimal design of a controller, is to have a good model of the physical system which is going to be controlled. Usually, industrial processes are multivariable systems with important coupling effects between the controlled variables and consequently, correct models are quite difficult to obtain. In order to model a pressure and temperature system (a laboratory process control system), two different types of strategies have been used. The first one was based on an experimental study of the system dynamic performance, using time-response and frequency domain methods. When trying to obtain the time-response of a system, step-response methods relating the different controlled variables with the applied signals were used. On the other hand, for the frequency domain response, the Ziegler-Nichols frequency response method for determining interesting points of the Nyquist curve was analysed. From the data resulting of both analysis, the transfer-function matrix was derived. Then, simulations of the model performance using MATLAB/SIMULINK were carried out, to check the correct modelling of the system. The second method used for the calculation of the correct model of the system, was the one based on the Recursive Least Squares method, used to identify the correct parameters of each transfer function of the matrix which represents the system. This method could be specially interesting when dealing with time-varying systems, because in this particular case the parameters of the system can change under different working conditions. Some of the variations are related to nonlinearities, process ageing and other changing properties in the system. As adaptive controllers work with identifying techniques trying to update the control parameters as the process changes, a correctly used RLS method would in fact, be very useful. It has to be pointed out that even if this method requires more computations than the experimental previous ones, the resulting models are normally more accurate than the ones experimentally obtained. After having completed the calculations using the two different techniques, a comparative study was done in order to ensure the effectiveness of the model to be used for control purposes. The presented paper has described and used two well-known modelling techniques to obtain a correct model for the temperature/level process control. The most important task which we have accomplish is the use of simple transfer functions for each component of the G(s) or G(z) matrix, in order to avoid using too much computation time in the identification and when making the model work in a more complete controller design method. As the first controller to be applied to the system is a PID one (multivariable), it was enough to model the system in a very simple way, without taking into account specific features. After this task has been achieved, it is our aim to implement a more complex adaptive PID controller. That is the reason for it to be specially interesting to start with a model obtained from an identification method.