Virtual Flight Data Recorder Based on Least Squares Support Vector Machines
Some recent research efforts to improve aviation safety are concerned with analysis of data that are routinely recorded on Flight Data Recorder (FDR) of aircraft. Earlier FDRs on most aircraft do not record the angular deflections of the control surfaces which are crucial in the investigation of aircraft crash. The absence of the critical dynamic parameters has hampered the accident investigation efforts. In the paper a Least Squares Support Vector Machines (LS-SVM) approach is proposed for the reconstruction of the time histories of the control surfaces deflections. The scheme consists of two LS-SVM, The first is named LS-SVM Simulator (LS-SVMS) which is used to simulate the aircraft dynamics, while the second is named LS-SVM Reconstructor (LS-SVMR) which is used to reconstruct the primary control surfaces deflections. The LS-SVMS is trained off-line, using available flight data for the particular aircraft, for the purpose of simulating any desired dynamic output recorded in current FDR, and then interfaced with an LS-SVMR. The LS-SVMR outputs are the estimates of the control surfaces deflections which minimize weighted quadratic difference performances between the time histories data from the LS-SVMS and the actual FDR data. The scheme has been introduced, tested, and validated with flight data from a J8 with an FDR with extended recording capabilities. The simulation results show that the proposed method provides a powerful tool for reconstruction of the control surface deflections data and has promises application in the investigation of aircraft crashes.
Least Squares (LS) Support Vector Machines (SVM) Virtual Flight Data Recorder (VFDR) Investigation of Aircraft Crash
Jingshun Duanmu Xusheng Gan Baicheng Zhang Chao Shi
College of Automation, Northwestern Polytechnical University, Shaanxi, Xi’an 710072 Air Force University of Engineering, Shaanxi, Xi’an 710038 Military Representative Office by PLA in 624 Factory, Heilongjiang, Harbin 150030
国际会议
武汉
英文
2007-09-21（万方平台首次上网日期，不代表论文的发表时间）