Noise-robust adaptive feature mode decomposition method for accurate feature extraction in rotating machinery fault diagnosis

Abstract

Rotating machinery typically consists of multiple rotating components, and its fault signals contain not only periodic impulse components caused by local defects but also periodic noise components generated by the normal operation of other rotating parts. Especially in the case of compound faults, the vibration signals exhibit the characteristics of simultaneous coupling of multiple periodic components and multiple pulse components, greatly affecting the accuracy of compound fault diagnosis. In order to accurately separate and extract individual fault components from the rotating machinery's compound fault signals under strong periodic noise interference, this paper proposes a noise-robust adaptive feature mode decomposition method for compound fault diagnosis in rotating machinery. In addressing the challenge of existing decomposition methods, which heavily rely on accurate fault period estimation and initialization of decomposition number, an efficient strategy has been developed within the proposed method. This strategy remains effective even under intense periodic disturbances by accurately pinpointing the resonance bands induced by faults. It simultaneously acquires the essential prior knowledge necessary for mode decomposition, resolving the issue of prevailing fault period estimation methods being prone to failure in the presence of strong periodic noise. Furthermore, a feature mode decomposition method with the second-order indicators of cyclostationarity as the objective function is introduced. This, coupled with the devised parameter optimization strategy, facilitates precise decomposition of compound fault components in the presence of strong periodic noise. Finally, the robustness of the proposed method against periodic noise and its outstanding ability to extract compound fault features undergo validation through simulations and experiments, highlighting its potential for advancement in the field of rotating machinery fault diagnosis.