Research On QAR Data Anomaly Detection Algorithm Based On Wavelet Clustering

With the development of society progress,more and more people choose to travel by air passenger plane,the flight safety of the aircraft received more and more attention,directly or indirectly affect the development of civil aviation.How to effectively find the potential safety hazards in the flight process is particularly important.At present,QAR is used to store a series of parameters in aircraft flight.Therefore,it is great significance to analyze the potential fault information of QAR data for the follow-up fault location of the aircraft and to solve the fault.In this paper,In order to detect outliers information,studied the application of wavelet clustering algorithm in outlier detection of QAR data and find out the cause of the failure by matching the obtained anomaly point data with the standard fault sample under the expert experience,so as to effectively help the aircraft maintenance personnel locate the cause of the fault and solve the fault.Completed the following described for this study.1?Completed the outlier detection algorithm and fault analysis of QAR data.The method firstly obtains the data of the cruise phase during the flight process,and then uses the wavelet clustering algorithm to cluster the data to find the hidden information in the QAR data.Then according to the standard fault model,we use the outlier data and the standard fault model to do the similarity matching to locate the fault type.The effectiveness of the method is verified by expert experience and maintenance manual.2?By analysis the wavelet clustering algorithm in-depth,found that the wavelet clustering algorithm mark the connected unit density threshold for the same cluster,and the grid connected unit in there may be several cluster boundary,also not belong to any cluster,resulting in different clusters are labeled with the same cluster because MUAP,Also grid does not meet the density threshold may also exist data objects that belong to the cluster and the cluster boundary that we called it fault division.so as to put forward improved wavelet clustering algorithm divisive non-uniform grid,and further refinement of the boundary grid and further processing of the grid which does not satisfy the density value,finally form a cluster and the feasibility of this method is validated.Compared with the traditional wavelet clustering algorithm,the improved wavelet clustering algorithm can improve the efficiency of the algorithm without increasing its order of time complexity.3?Researching similarity measure and detailing analysis of the principle of the method,combining with the analysis of the characteristics of the QAR data found thatcorrelation between the QAR data attributes,and selected the similarity match based on the Mahyagra distance to the anomaly point data and the standard fault model to find the fault type.