Fault feature extraction method for rolling bearing in gas turbine engines based on comprehensive dynamic screening

Abstract To address the issue that gas turbine engine (GTE) rolling bearings are prone to intense background noise during operation and the transmission path is highly complex, thereby making it challenging to extract and characterize vibration features, a fault feature extraction method based on optimized wavelet packet decomposition and comprehensive dynamic screening index is put forward. Firstly, Shannon entropy is employed as the objective function to input 100 wavelet basis functions into the database, and the Shannon entropy of the signal after decomposition and reconstruction of each wavelet basis function is calculated respectively. The wavelet basis function with the lowest Shannon entropy is chosen as the optimal wavelet basis function of the signal, and the collected bearing vibration signal is decomposed through wavelet packet. The kurtosis value, phase relation value, and energy ratio of each node component are calculated. Subsequently, the variance is utilized as the index to find the most suitable weight for different vibration signals and fuse it into a comprehensive dynamic screening index. Then, the contribution degree of each component is computed, and the first i node components whose contribution degree reaches the threshold are reconstructed to obtain the new signal after noise removal. Eventually, the weak fault characteristics of the bearing are extracted by envelope spectrum. The theory and tests confirm that it can serve as one of the effective means for complex signal processing and diagnosis of the GTE rolling bearings.