Study On Flight Conditions Monitoring Based On Time-frequency Analysis

Work in harsh environments with high temperature and high-speed, the aircraft engine easily catchvarious faults. This not only affects the safety of flight, but also increases the aircraft’s operating costs.In order to ensure reliable operation of the engines, it is necessary to identify potential or impendingengine faults in advance. The flight data of the gas circuit parameters is closely related with theengine. This paper aims to study this data by time-frequency analysis, to find fault characteristicparameters laid the foundation for the assessment and prediction engine health.The flight data is nonlinear and non-stationary, so the Hilbert-Huang transform (HHT) analysismethod proposed by Huang et al. in1998was chosen as the time-frequency analysis method. TheHHT consists of two aspects: empirical mode decomposition (EMD) and Hilbert spectral analysis.The signal will be decompose into a finite number of intrinsic mode functions and a trend by EMD.But there are many problems of EMD itself; mode mixing is one of them. So here use empirical modedecomposition (EEMD) to replace the EMD, EEMD is a method based on assistant analysis of thewhite noise proposed by Huang et al. in recent years, and complementary forms of EEMD (CEEMD)can effectively eliminate the added white noise, and improve the decomposition efficiency.Before the detailed analysis, the flight data was firstly denoised by singular value decompositionand smoothed by five-spot triple smoothing method. Then contraposed the different characteristics ofthe different flight phases, this paper gave different analysis for each phase. Depended on theintegrated information of the time, frequency and amplitude gave by the Hilbert spectrum; everymutation was clearly displayed. And gave a method to evaluate the degree of mutations to thesequence by singular value decomposite the Hankel, IMFs or HHS matrix.Specific implementations, this paper used Visual C++to complete the pretreatment of the flightdata and data analysis software needed for HHT. The pretreatment including: flight dividing, errorcorrection and de-noising operation. HHT analysis completed CEEMD, calculating Hilbert spectrumof the results from CEEMD, and the reconstruction of the signal sequence and marginal spectrumanalysis. In addition, the Matlab was used to explain the mask signal post processing.