Study On Model-Based Fault Diagnosis Method And Its Application In Aerospace

Fault diagnosis technique is very important to ensure the safety and reliability of spacecraft, and model-based fault diagnosis method is an extremely important branch of fault diagnosis technologies. Taking the power subsystem of spacecraft as the research background, this dissertation makes a systematic study on model-based fault diagnosis technique and its application in aerospace.Based on the comprehensively analyzing information of model-based fault diagnosis theory and development home and abroad, this thesis proposes a modified HS-tree method to solve the problems of HS-tree too large and calculation cost during the process of solving the minimum hitting set. The modified HS-tree method generates tree structure using number indexes and the numbers of rules tested are very small. It is simple, very fast, and has smaller calculation cost.This thesis makes a systematic study on the discrete method describing system models and adopts the form of automaton to construct system models. Based on the state transitions inside the system model and the current system states, the state tracking problem and the optimized constraint satisfaction problem are discussed. Meanwhile, this thesis proposes a conflict A* algorithm to search the system states and uses the conflict A* algorithm in the diagnosis of the spacecraft power system. The performance of the proposed method is compared and analysed in different ways.The uncertain problem in fault diagnosis is deeply studied in this paper. This thesis combines Bayesian network with model-based diagnosis framework for the problems that model-based diagnosis method has the difficulty to dealing with the coupled relations between the system components. The adopted Bayesian network method constructs the Bayesian network observation model of the system using the observation from the system, and also studies the reasoning method between the components inside the observation model. The proposed method is validated by the construction of observation model of the two-axis satellite attitude control system and the calculations of the posterior probabilities of the diagnoses. The fault detection method based on the negative selection algorithm is analyzed in-depth. This thesis proposes the fault diagnosis method based on the negative selection algorithm for the difficulties in dealing with unknown faults and modeling fault modes of components existing in model-based fault diagnosis method. The proposed method can produce better detecting set combining the negative selection algorithm detecting anomaly with clone selection theory and mutation theory.Finally under the background of practical engineering application, the model-based fault diagnosis system of satellite power subsystem is developed. The thesis analyzes the designing principles of the system development, the integration strategy of diagnosis, and the qualitative model of satellite power subsystem is constructed.