An Adaptive Fault Feature Extraction Method for Aeroengine Intershaft Bearing Based on Casing Vibration

Fault diagnosis of the intershaft bearing of aeroengine is quite challenging because vibrations can only be measured on the distant external casing, resulting in weak fault features and complex interference. To address this issue, this paper proposes an adaptive multipoint optimal minimum entropy deconvolution adjusted (MOMEDA) method to extract fault impulses of intershaft bearings based on casing signals. First, envelope harmonic product spectrum (EHPS) is used to estimate the bearing fault frequency and determine the parameter search range. Additionally, an evolved firefly algorithm (EFA) is introduced, incorporating mutation and step-size factors to enhance the convergence speed and global search capability of the algorithm. Furthermore, a new objective function EP is proposed that considers both the time-domain complexity and characteristics of periodic pulses, which is designed to enhance the capability to localize fault periodic impulses. Simulation results show that the proposed EPEFA-MOMEDA can accurately extract the fault period from simulated signals containing high-energy harmonics and random impulse noise without prior knowledge of the fault period. Finally, verification was conducted using real aero-engine casing vibration data. Comparative analysis demonstrates that this method is practical for engineering applications in extracting intershaft bearing faults from casing vibrations.