Research On Airborne Data Filtering,Fusion And Clustering Detection Algorithm

The Next Generation Air Transportation System(Next Gen)is the FAA-led modernization of America's air transportation system to make flying even safer,more efficient,and more predictable.Modern airborne systems include a large amount of data,the sources can be navigation,route planning,collision avoidance system and so on.The redundancy and complementarity was used to excavate the data value,which improved the safety and efficiency of air transportation.It was in line with the goal of Next Gen.Based on the application background of Integrated Environmental Monitoring System(ISS),collision avoidance system and quick access recorder(QAR)data were the main research objects.The input data filtering and fusion,abnormal track detection were achieved after simulation.The main contents includes the following:1.The functions and technologies of the current typical collision avoidance system were introduced,including the Traffic Alert and Collision Avoidance System(TCAS),Automatic Dependent SurveillanceBroadcast(ADS-B)and ACAS X.2.For more safe and precise flight,IMM interactive multi-model was used in filtering and fusion of collision avoidance system input data.The Variational Bayesian algorithm was used to estimate the noise variance,and based on this,the sampling period was adaptive.The filtering fusion accuracy was improved and fault location of fusion system was realized.It was proved that VSVB-IMM(Variable Sampling Variational Bayesian-IMM)based fusion was superior to FSVB-IMM(Fixed Sampling Variational Bayesian-IMM)and CS model algorithm by analyzing RMSE and alarm condition after typical ADS-B failure modes injected.In some conditions,although the ADS-B information deteriorated,the fusion system could ensure the normal operation,which reduced the probability of system failure.Those were the positive gains brought by fusion.3.For more predictable and efficient requirements,research on abnormal track detection was carried out based on QAR data.In the absence of threshold criteria,the abnormal track detection was realized by fast search and find of density peaks.The decision support for flight optimization of airlines was supplied.Experimental results showed that the abnormal track can be detected without prior parameter setting.The results were consistent with Density-Based Spatial Clustering of Applications with Noise.Different from exceedance detection method,some new abnormal patterns may be detected by clustering mining.4.A digital simulation system was established in the laboratory,ADS-B data was connected to the ISS simulation system.On the one hand,the filtering research of input data was carried out and it was also the data incentive for collision avoidance system.On the other hand,the airspace situation monitoring could be realized and the recorded data was the basis for abnormal track detection.