Gearbox Fault Diagnosis Based on ALIF-PE-GOLSSVM

Abstract

An adaptive local iterative filtering (ALIF) and permutation entropy (PE) fault diagnosis method based on gene optimized least squares support vector machine (GOLSSVM) is proposed. The method is applied to the diagnosis of gearboxes, and the identification of four fault types of gearboxes is successfully realized. Aiming at the problem that permutation entropy cannot directly identify different fault categories of gearboxes, the advantages of ALIF method in removing residual noise and suppressing mode aliasing compared with EMD method are used, and the ALIF method is used to reduce noise and extract effective components. Then the PE value with component (C-PE value) is calculated to obtain the multi-scale characteristics of vibration signals. Then the genetic algorithm is used to optimize the least squares support vector machine (LSSVM). Finally, the feature vector is input into GOLSSVM to classify the fault features. The results show that the proposed method has advantages in fault recognition accuracy compared with BP neural network and SVM.

Keywords

• Gene optimization;Support vector machine;Adaptive local iterative filtering;Permutation entropy