Search

The general purpose of this thesis is to focus on a particular industrial process (from the beer industry) which serves as a guidance example for optimal control using different algorithms/methods. At the same time, the aim is to demonstrate the capabilities/features of MATLAB and SIMULINK as tools used in programming algorithms and simulation for optimal control of non linear systems. The thesis shows how to approach an optimisation problem with different techniques and to compare them on the same basis. The main reasons for carrying out research on a beer fermentation process can be summarised as follows: this kind of industry represents an up-to-date example of industrial processes in general, the need to compare and evaluate optimisation methods (well established and "modern") on similar circumstances using the sanle process model and finally, give a good foundation for the control engineer to followup this work with different optimisation techniques and/or any other industrial process. The fundamental features of the methods used involve the viability of known previously tested algorithms for optimal control of beer processes with high nonlinearity and constraints; thus testing the flexibility of some of the known MA TLAB Toolboxes for the optimal control of a particular simulated mathematical model. An important aspect of the experimentation that has been carried out, is the creation of a simulated model of a selected beer process by means of including the mathematical equations, parameters and initial conditions into an s-function block. This SIMULINK model also incorporates the particular objective function that can be calculated directly after the simulation of the process for a particular input temperature profile. Together with the use of some available MA TLAB functions for the formulation of particular optimal control techniques, this facilitates the creation of program routines that can be interfaced with the simulated process. The final results using different optimisation methods such as: the gradient method in function space, DIS OPE algorithm, Genetic Algorithms and Sequential Quadratic programming; show substantial improvement in the perfomance index obtained. The optimised temperature profiles found can be implemented for industrial application to provide a maximised ethanol production under particular restrictions, i.e. final byproducts concentration, contamination risk and brisk changes in temperature.