Cyclostationary Feature Mode Decomposition and Its Application in Fault Diagnosis of Planetary Gearboxes via Built-In Information

Fault detection of planetary gearboxes is of vital importance for the operation safety of large equipment. Built-in information, including the encoder signal, which can detect the slight torsional vibration caused by the machinery fault, has an outstanding superiority for the fault feature extraction. Yet, the differentiate operation of the encoder signal can enhance the fault information and also amplify the different interference components, which would further undermine the efficiency of the traditional methods. To overcome this issue, a new decomposition method, cyclostationary feature mode decomposition (CFMD), is proposed in this article. First, a filter bank with different initialization is constructed to provide the direction of the decomposition. Then, the maximum second-order cyclostationary acts as the optimal objective of decomposition to accurately lock the fault information. Finally, an entropy criterion is applied to select the informative mode and eliminate the mode mixing. Compared with the traditional decomposition methods, the proposed CFMD has obvious advantages in the process of the built-in encoder signal. Without any preprocessing enhanced methods, CFMD can extract weak fault information, and the proposed method focused on the periodic fault information without any prior knowledge. The superiority has been verified by the simulated and experimental data collected from planetary gearbox faults with the built-in encoder signal.