Vehicle Bearing Fault Diagnosis Model Based on Speed Information Embedding and 1-D Partial Convolution

Faults in vehicle bearings have the potential to pose a significant risk to life safety, underscoring the critical importance of promptly and effectively detecting and resolving such issues. However, the variable speeds of vehicles present a challenge to the accurate diagnosis of bearing faults due to the instability of the fault signals. This article puts forward a novel diagnostic method based on speed information embedding and 1-D partial convolutional network (SIE_1DPCN). The proposed model reduces the computational costs associated with the process and is capable of automatically extracting speed features and embedding them into fault features, thereby improving the accuracy of fault classification. First, 1DPCN is employed to automatically extract both speed and fault features, with the objective of reducing computational costs while accurately extracting features. Subsequently, the extracted speed information is embedded into the fault-type classifier, thereby enhancing the accuracy of fault classification under varying speeds. Ultimately, a variable-speed dataset was constructed, and the efficacy of the proposed method was validated by introducing vehicle noises with varying signal-to-noise ratios (SNRs). The results demonstrate that the proposed method attains over 90% accuracy in diagnosing bearing faults in vehicles with variable speeds under noise conditions with SNR ranging from 0 to -10 dB.