Fault diagnosis of rolling bearings based on enhanced optimal morphological gradient product filtering

Due to the interference of various irrelevant information in the environment, the early bearing fault features are difficult to detect. To enhance the fault-associated feature extraction performance in the process of bearing fault diagnosis, a signal processing method named enhanced optimal gradient product filtering (EOGPF) is proposed. First, the filtering capabilities of eight morphological gradient operators are investigated and compared to excavate the optimal morphological operators. Then, a new optimal gradient product operator (OGPO) is developed to improve the extraction performance of bearing fault-induced transient impulse information in the vibration signal. Finally, a novel EOGPF method combining noise removal and feature extraction is proposed to diagnose bearing faults. The OGPO-based morphological filtering is applied to remove noise and extract fault-induced impulse features from the vibration signals. Moreover, a two-stage denoising strategy based on me-dian filtering and autocorrelation is used to enhance the noise removal performance of OGPO-based morpho-logical filtering when processing the signal with strong noise interference. The analysis results of simulation signal, bearing accelerated life test data and measured railway bearing data verify the EOGPF can effectively enhance the extraction performance of fault-associated features. The comparison results of the EOGPF with several existing methods show its superiority in bearing fault diagnosis.

Automated summary

This study proposes an enhanced optimal gradient product filtering (EOGPF) method to improve the extraction performance of fault-associated features in bearing fault diagnosis. By optimizing morphological operators and developing an optimal gradient product operator, the EOGPF method effectively extracts fault-induced impulse features and combines noise removal and feature extraction techniques for bearing fault diagnosis.