Fast fault diagnosis method of rolling bearings based on compression features in multi-sensor redundant observation environment

Abstract

High-precision synchronous condition detection and fault diagnosis for bearings are important to reduce the failure rate of rotating machinery products. Therefore, this paper proposes a fast and high-precision diagnosis method based on multi-sensing fusion and compression features. First, the traditional random weighting method is optimized. The fluctuations of each signal are calculated, and used as the basis for balancing the weighting relationship between the current and historical sampling values, in order to achieve high precision fusion of bearings signals. Second, based on the traditional compression sensing method, the reconstruction part that would further increase the diagnosis error and time is omitted. The partial Hadamard matrix is constructed to retain the feature trend in the compressed signal, and the bearings fault diagnosis based on the compressed features is realized. Finally, the combination of these two methods can reduce the number of signal samples during the collection and transmission process, and realize a direct, fast and accurate diagnosis of the bearings state. Simulation and experimental results verify the superiority and effectiveness of the proposed method.