Tesis doctoral de Pedro Luis Guerra Brito
Model-based fault detection is based on the use of mathematical models of the monitored system. Reliability and performance of fault detection algorithms depend on the quality of the model used. These algorithms can often be improved by refining the models they are based on. However, high fidelity models are costly and modelling errors and disturbances in complex engineering systems are inevitable. Hence there is a need to develop robust fault detection algorithms where model uncertainty is explicitly taken into account. The robustness of a fault detection system indicates its ability to distinguish between faults on the one hand and model uncertainty and disturbances on the other hand. the robust fault detection research has roughly focused on two distinct approaches. In one of the approaches, characterized as active, the central idea is to decouple the effect of the uncertainty. The other approach, known as passive, is based on enhancing the robustness of the fault detection system at the decision-making stage. The aim with the passive approach is usually to determine, given a set of models, if there is any member in the set that can explain the measurements. when a measurement is found to be inconsistent with this set, a fault is assumed to have occurred. As an exact representation of the set of parameters or states consistent with the measurements is hard to calculate, outer bounds are often used instead, using algorithms coming from set-membership estimation. This is the approach adopted in this thesis. set-membership estimation methods have been the subject of a number of publications. they can be classified according to how the approximation of the feasible solution set is represented or parameterized. In other study the set was over bounded by an ellipsoid. other authors have focused on orthotopic approximations. Set-membership algorithms have been already applied to fault detection. in the current thesis, the robust fault detection is studied using zonotopes, which include as a special case parallelotopes. This thesis will show that the zonotope representation of the uncertainty combined with the above mentioned identification methods, is particularly suitable for fault detection. as an extension of the use of set representation with zonotopes, this thesis presents a computational procedure to evaluate fault-tolerance of linear constrained model predictive control (lcmpc) in case of a given actuator fault configuration (afc). faults in actuators cause changes in the constraints on the control signals which in turn change the set of feasible solutions. This may derive in an empty set of admissible solutions for the control objective. Therefore, the admissibility of the control law facing the actuator faults can be determined knowing the feasible solutions set. This thesis presents a method to compute this set and then evaluate the admissibility of the control law. In particular, the admissible solutions set for the predictive control problem including the effect of faults is determined using an algorithm that uses set computations based on zonotopes.
Datos académicos de la tesis doctoral «Robust fault detection and tolerance evaluation using zonotopes«
- Título de la tesis: Robust fault detection and tolerance evaluation using zonotopes
- Autor: Pedro Luis Guerra Brito
- Universidad: Politécnica de catalunya
- Fecha de lectura de la tesis: 27/07/2009
Dirección y tribunal
- Director de la tesis
- Vicenc Puig Cayuela
- Tribunal
- Presidente del tribunal: fatiha Nejjari
- José Manuel Bravo caro (vocal)
- (vocal)
- (vocal)