Research On Fault Detection And Analysis Algorithm For QAR Data

Seeing all the way to the development of the aviation industry, aviation safety incidents has never been eliminated, it's obviously that the security situation is not optimistic. However, as the saying goes, forewarned is forearmed, through collecting and analyzing aircraft flight data, summarizing safety characteristics of the aircraft flight data and studying the data got from plane, it can be fully prepared for aviation development by providing objective data basis for flight risks determined and appropriate precautions taken. It is definitely possible to further improve the level of aviation safety and reliability of scientific management, reducing the likelihood of accidents.QAR(Quick Access Recorder) data is a kind of timing flow data representing aircrafts' flying states. It is a kind of data which is high-dimensional, complex, great volume and holding mutiple characteristics, the complexity will undoubtedly handicap the efficiency of processing research heavily. In order to reduce the amount and complexity of data to efficiently retrieval current QAR data to determine whether it includes hidden faults or not. If included, faults type should be told. In this paper, a summary data structure for QAR data is created to compress data scale, sub-tree search method is used to retrieval processed data, with the purpose of improving retrieval efficiency. In addition, uncertain relationship lies between different dimensions of QAR data, subtle changes of the relationship is also an important characterization of aircraft status changement, it should not be ignored. Within this paper, method of extracting data by outlining the data is used to reduce its scale, for which a summary data structure for great number of QAR data is created; method of pattern growth is used to mine summary data to create the unique QAR FP-Tree(Frequent Pattern Tree) for QAR data, finally, method of sequences query is used based on FP-Tree structure to achieve the purpose of fault locatization.In this paper following tasks are completed:1.Using summary data structure to compress QAR data scale.Through symbolizing QAR data into discrete data, segmenting the data symbols and calculating the frequency of each symbol appears in each segment, at last there comes finite symbols and their frequency of occurrence, which can be used as a summary data structure of each data segment.2.Using pattern growth method to mine QAR summary data structure, modifying FP-Growth algorithm to get them fit for summary data structure of QAR data to create a unique pattern growth tree(FP-Tree).3.For a given fault model, focusing on the special structure of FP-Tree, using subsequence queries basing on the tree structure to achieve the purpose of locating fault-point segment both quickly and efficiently.