Design and implementation of model based controller for systems with input saturation and model mismatch

This research deals primarily with the design and implementation aspects of model based digital control algorithms. The performance of model based digital controllers degrades whenever the process input saturates or the process model is significantly different from the true process. Simulations and experiments have been performed to identify the saturation problems in model based controllers and four algorithms were proposed to compensate for this problem. The saturation problem was generalized to cover all situations where the output computed by the model based controller cannot be applied to the actual process. This includes actuator saturation, operator override, temporary actuator failure and asynchronous control. A generalized implementation scheme of model based controllers was proposed and a dramatic performance improvement was observed by using such an implementation scheme. The second part of the thesis focuses on the robustness aspect of model based controllers. The work reported here takes a time domain, rather than the traditional frequency domain approach. A new stability theorem and corollaries for uncertain systems have been derived and applied to simulated processes. A new analysis method, Error Band Method (EBM), has been developed to analyse the performance of a dynamic uncertain control system. Monte Carlo simulations have been performed to verify the validity of this approach. The method was extended to include controller design and controller comparisons. This methodology provides a different perspective to the analysis and design of uncertain model based digital control system.