A study on the feature separation and extraction of compound faults of bearings based on casing vibration signals

The autocorrelation function is combined with wavelet transform and cyclostationary theory (WT-AF-CT) in place of threshold denoising, and meanwhile the mean power ratio (MPR) is calculated by the proposed method. Furthermore, extracted characteristic as well as calculated MPR is used to identify compound faults of rolling bearings in aero-engine based on casing vibration acceleration signal-including the ones of common rolling bearing (inner race rotates and outer race is constant) and intershaft bearing (co-rotates with outer and inner race). A comparative analysis was carried out between conventional researches (cyclostationary theory (CT) or wavelet transform combined with threshold value denoising (WT-TD)) and proposed WT-AF-CT method. Additionally, the effect of sensors installation direction for feature separation and extraction of compound faults is considered. The results indicate that the proposed WT-AF-CT method can separate and extract characteristics of compound faults exactly and identify fault types of bearings precisely, no matter sensors are installed horizontally or vertically, while CT or WT cannot.