Research On Aeroengine Fault Diagnosis Based On Multivariate Statistical Analysis

The aero-engine, whose slight fault will cause fatal accident directly, is important component of airplane. Moreover, the aero-engine's structure is complex and works in harsh environment, so the research of aero-engine fault diagnosis is a focal point. In this thesis, Multivariate statistical analysis is used to obtain an overall performance parameter in order to monitor the turbofan engine's gas path. When fault occurs, Least Squares Support Vector Machine (LSSVM) is used to implement fault accommodation, which provides the theory to component's condition based maintenance.The main contents are as follows. First, Aeroengine fault diagnosis based on multivariate statistical analysis is studied. Principal Component Analysis (PCA), Kernel Principal Components Analysis (KPCA) and Independent Component Analysis (ICA) are validated. Second, Multi-scale Independent Component Analysis (MSICA), which combines wavelet transform with ICA to detect aeroengine path fault, is introduced into fault diagnosis. Third, ICA-LSSVM is introduced into aeroengine fault accommodation. ICA and Least Squares Support Vector Machine (LSSVM) is combined to identify fault component. Simulation results show that ICA is effective, MSICA methods can detect slight fault further, and Fault accommodation method base on ICA-LSSVM is feasible.