Document Type

Dissertation

Rights

Available under a Creative Commons Attribution Non-Commercial Share Alike 4.0 International Licence

Disciplines

1.2 COMPUTER AND INFORMATION SCIENCE, 2. ENGINEERING AND TECHNOLOGY

Publication Details

Submitted in partial fulfilment of the requirements of the Masters in Engineering in Pharmaceutical Process Control and Automation.

Abstract

The aim of this project was to investigate the performance of a number of key control strategies in the temperature control of batch reactors. A bench scale model was built and a batch production system was then implemented on this model. As there was no a priori knowledge of the system a number of common system identification methods were investigated. The system was controlled using a Mitsubishi FX(2)N Programmable Logic Controller which was interfaced with a PC running ICONICS, a Supervisory Control And Data Acquisition software package. The system identification methods produced two different models for the system and these models were examined against the actual system using Matlab/SIMULINK, a software package used for technical computing. Then a number of tuning rules were investigated and implemented on both models with the results compared and contrasted. The standard Industry criteria were used to compare the performance of the servo response for each controller. The PI controller using Zeigler-Nichols tuning rules was set as the bench mark. The Cascaded control strategy offered no increase in performance in the servo response in either the actual process or the SIMULINK models. However the regulatory response of the Cascaded strategy would offer an improvement on the performance of the PI controller. The performance of the Smith Predictor was limited due to the minimal time delay relative to the time constant. The Integrating method proved to offer an improvement on the two point method in terms of system performance and in the time required to identify the initial controller. Also the Smith Predictor offered a slight improvement in both the laboratory model and in the Matlab/SIMULINK simulations.

Download


Share

COinS