Optimization-based improved kernel extreme learning machine for rolling bearing fault diagnosis

Abstract

Rolling bearing is one of the key components in rotating machinery. The working condition of rolling bearing is complex and non-stationary with shock and noise. Thus, fault diagnosis of rolling bearing is of great significance in rotating machinery. In this paper, a novel method called optimization-based improved kernel extreme learning machine is proposed for fault diagnosis of rolling bearing. Firstly, different signal processing methods and data analysis methods are used as the second layer for feature extraction and vibration data dimension reduction, and the extracted data are sorted by experience pool. Secondly, kernel extreme learning machine (K-ELM) is used as the hidden layer and the output layer to enhance feature learning and classification of the extracted data. Finally, particle swarm optimization is employed to optimize the key parameters of the improved kernel extreme learning machine. The proposed method and five other methods are applied to analyze the raw vibration data of rolling bearing, and the results confirm that the proposed method is more effective than ELM, K-ELM, optimized K-ELM, support vector machine and back-propagation neural network.