Evidence Theory And Its Applied Research In Information Fusion, In The Early Fault Diagnosis

Aero-engine is the most important component of aircraft, so its security and reliability has close effect on the aircraft flight performance,economy and security of the crew. If the fault of aero-engine can be detected and resolved in the early stage, aero-engine security and reliability can be increased efficiently. Consequently, it is essential to do the research on the early fault of the aero-engine. However, the early fault symptom of the aero-engine is very fragile and it is uncertain in the early fault stage. Consequently, in order to improve accuracy of early fault diagnosis, it is a necessary and efficient way to fusion the early fault information that is attained through sensors which monitor the real-time situation of the aero-engine. Nowadays, with the development of information technology, computer technology and sensor technology, D-S evidence theory has be extensively researched and successfully applied to information fusion. This makes information fusion method based on D-S evidence theory become a very effective way in early fault diagnosis of aero-engine.This research is from the National Natural Science Foundation project: Research on aero-engine early fault diagnosis based on weak feature signal extraction and uncertainties multi-information fusion (No.60672165). Firstly, D-S evidence theory has been introduced carefully. Although it is a good imprecise and uncertain appliance, there are some shortages in its body. Especially the evidence conflict that may cause the counter-intuitive results is one of the most concerns for information fusion by Dempster-Shafer's (D-S) evidence theory. To deal with the issue and remove the evidence conflicts greatly for the improvement of belief convergence, we have investigated the evidence conflict and belief convergence in the following three aspects: 1) The Basic Probability Assignments (BPAs) of evidences are same, 2) For each subset, the BPAs of different evidences are same; the BPAs for different subsets are different, 3) The evidence BPAs are different for a same subset in a discernment frame. In addition, a few methods in dealing with evidence conflict were analyzed and compared. Following that, a new paradox combination algorithm based on a difference factor of evidence absoluteness and a relative difference factor of evidence absoluteness was proposed with the consideration of local attributions to local conflicts. The proposed algorithm was verified by numerical examples. The analysis results showed the efficiency of the proposed method to improve the performance of belief convergence. Comparison studies indicated the advantages of the proposed method as well.Secondly, basic concepts, levels, grades and methods of information fusion are briefly introduced. Necessity of information fusion in diagnosis is illustrated through practical background and mathematics direction with information theory. Moreover, basic frame and methods of information fusion in diagnosis are introduced and summarized.Finally, D-S evidence theory is used in information fusion of early fault diagnosis for aero-engine. Model of fault diagnosis is built using the above method. With the analysis of the characteristic of the early fault and information fusion of aero-engine, we present the method to form the basic belief function and use the evidence theory as information fusion of aero-engine fault symptom level. So the more precise early fault symptom can be attained. Moreover, the attained fault symptom is applied to fuzzy cluster analysis to attain aero-engine fault sources according to the most subjection principle. Furthermore, this method uses fuzzy measure to indicate the characteristics of imprecision, uncertainty and ambiguous to aero-engine early fault symptom. This is more suitable to the practical situation of the early fault of the aero-engine.