Correlation-constrained sparse decomposition for fault diagnosis of large-scale bearings

Abstract

Accurate fault diagnosis of large-scale bearings under strong noise interference remains a significant challenge, particularly in multi-sensor monitoring scenarios where redundant information among channels is insufficiently utilized. To address this issue, a correlation-constrained sparse decomposition method for fault diagnosis of large-scale bearings is proposed in this paper. By incorporating inter-channel correlation characteristics as prior constraints into the sparse representation framework, the proposed method establishes a collaborative sparse decomposition model for multi-sensor vibration signals. Specifically, the common sparse support shared among correlated channels is extracted from a reference signal and subsequently reused to characterize the common fault-related components of other channels, while channel-specific sparse components are independently solved. Through this correlation-constrained strategy, weak fault impulses can be effectively enhanced while significantly reducing the computational burden of sparse decomposition. Simulation and experimental results demonstrate that the proposed method achieves superior noise suppression and fault feature enhancement compared with conventional sparse decomposition and traditional denoising methods.