Abstract

The feedback control system design technique of Model Predictive Control (MPC) has been vastly used in the chemical and process engineering industry, due to its ability to handle dynamics with multiple inputs and multiple outputs, which are essentially the majority of today's engineering systems. In addition, the field of Fault Detection and Diagnosis (FDD) in control systems also has been extensively researched over the past decades as it is critical for the controller to realize when and if a fault has occurred within a system. However, due to the high computational requirements, it is often challenging to implement FDD based MPC algorithms in resource limited real world systems.

This thesis addresses the development of MPC algorithms with combined state and fault estimation. Firstly, a novel Quadratic Programming (QP) formulation is developed for a recently proposed efficient MPC method along with simultaneous state and fault estimation. Another contribution is the enhancement of a standard integral action MPC algorithm (which has an implicit fault tolerance capability), to provide state and actuator fault estimation in real time. This work focuses on faults which are modeled as a Loss Of Effectiveness (LOE). The algorithm to estimate the system faults and states simultaneously is a simple observer based method which can be tuned beforehand, thus eliminating the need for on-line real time complex calculations. Lastly, a third contribution of this thesis is the application of the above methods to design MPC based flight control systems for fixed wing aircraft. Simulations are presented to demonstrate the effectiveness of the proposed methods.