Bearing Fault Diagnosis Method for Single-Channel Vibration Signal

Abstract

Conventional methods for diagnosing bearing faults typically depend on multiple sensors to monitor the operational state of equipment. However, in practical applications, space constraints often limit sensor placement, making it difficult to ensure diagnostic accuracy. This study proposes a fault diagnosis method based on the fusion of one-dimensional and two-dimensional features extracted from a singlechannel vibration signal. Firstly, one-dimensional time-domain and frequencydomain features are extracted from the vibration signal. The minimum Redundancy Maximum Relevance (MRMR) method is employed to select key features that have strong correlation with fault types while minimizing redundancy. Secondly, the single-channel vibration signal is converted into a two-dimensional image using the Gramian Angular Difference Field (GADF) method, and a Convolutional Neural Network (CNN) is employed to automatically extract high-level image features. The optimized one-dimensional feature vector is then fused with the two-dimensional image feature vector to create a high-dimensional feature representation, and the Support Vector Machine (SVM) is utilized to identify fault types based on the fused feature vector. The experimental results show that the feature fusion strategy using only single-channel vibration signals significantly improves the accuracy of fault diagnosis and provides an effective and interpretable solution for industrial fault diagnosis.