Gearbox Fault Diagnosis Based on Fast Iterative Filtering Decomposition and Modified Permutation Entropy

Gearbox is an important mechanical component. After the gearbox fails, the vibration signal is nonlinear and nonstationary. In addition, due to the prominent impact of both disruptive noise and components, it can be observed that the fault-related information embedded within the vibration signal is very weak, In order to diagnose gearbox faults accurately and quickly, a fault feature extraction model founded on fast iterative filter decomposition (FIFD) and modified permutation entropy (MPE) is established. Firstly, the vibration signal is decomposed into several intrinsic mode function (IMF) by FIFD with considerable computational efficiency. Subsequently, the IMF components with strong correlation were selected as main components by a newly proposed evaluation index called correlation energy kurtosis (CEK). Then, the fault information is extracted by modified permutation entropy (MPE), which introduced mean square value into entropy calculation to construct fault characteristics. Finally, random forest (RF) classification model is used to intelligently diagnose the fault type. It is verified by operating a set of experimental data from a fault gearbox that the proposed method enjoys a high accuracy of fault diagnosis and identify the fault type of the gearbox quickly.