Description

The fault-tolerant capability is an important performance specification for technical systems. Examples showing its importance are some catastrophes in civil aviation. According to official investigations, some air accidents due to failures are technically avoidable if the pilots can take right measures. But, relying on the skill and experience of the pilots, it cannot be guaranteed that reliable flight decisions are always made. Instead, if fault-tolerant strategies can be included in the decision-making procedure, it will be very useful for safer flight. Fault-tolerant control is generally classified into passive and active fault-tolerant control. Passive fault-tolerant control relies on the robustness of the controller, which can only provide limited fault-tolerant ability, while active fault-tolerant control turns to a fault detection and isolation module to obtain fault information and then to actively take actions to tolerate the effect of faults. Generally, active fault-tolerant control has more powerful fault-tolerant ability than passive fault-tolerant control.

In this dissertation, one focuses on active fault-tolerant control, which for this case considers model predictive control and set-based fault detection and isolation. Model predictive control is a successful advanced control strategy in process industry and has been widely used for processes such as chemistry and water treatment, because of its ability to deal with multivariable constrained systems. However, the performance of model predictive control has deep dependence on model accuracy. Realistically, it is impossible to avoid the eect of modeling errors, disturbances, noises and faults, which always result in model mismatch. Comparatively, model mismatch induced by faults is possible to be eectively handled by suitable fault-tolerant strategies. The objective of this dissertation is to endow model predictive control with fault tolerant ability to improve its eectiveness. In order to reach this objective, set-based fault detection and isolation methods are used in the proposed fault-tolerant schemes. The important advantage of set-based fault detection and isolation is that it can make robust fault detection and isolation decisions, which is the key for taking right fault-tolerant measures.

This dissertation includes four parts. The first part introduces this research, presents the state of the art and gives an introduction of used research tools. The second part proposes set-based fault detection and isolation for actuator and sensor faults, which is involved in interval observers and invariant sets. In the second part, the relationship between interval observers and invariant sets is firstly investigated. Then, actuator and sensor faults are separately coped with depending on their own features. The third part focuses on actuator and sensor fault-tolerant model predictive control, where the control strategy is robust model predictive control. The last part draws some conclusions, summarizes this research and gives clues for the future work.

The work is under the scope of the following projects:

• WATMAN: Analysis and Design of distributed optimal control strategies applied to large-scale WATer systems MANagement (web)