MPIFNet: A Multi-Path Information Fusion Fault Diagnosis Network Based on Time Series Two-Dimensional Transformation

Vibration signal analysis, which aims to monitor and assess the operational status of mechanical systems in real time, has proven to be an effective method for fault diagnosis. As a crucial component of hoisting equipment, the identification of fault signals in rolling bearings is of paramount importance. However, in practical industrial applications, fault diagnosis often fails to achieve satisfactory results. The challenges stem from the fact that vibration signals generated during the operation of rolling bearings are often accompanied by complex noise, which significantly impairs the accurate identification of fault characteristics. To address this issue, a multi-path information fusion fault diagnosis network (MPIFNet) has been proposed for rolling bearings. Specifically, a time series two-dimensional transformation module is introduced to extract key periodic features from the time series signals, thereby extending the original one-dimensional signal into two-dimensional space. Additionally, the multi-path time series extractor is designed to represent multi-scale features. Experimental results demonstrate the superiority of the proposed method, achieving state-of-the-art performance on a public dataset. The potential extends beyond academic applications, offering significant benefits for industrial settings, including cost savings, improved operational efficiency, and enhanced safety by minimizing the risk of unexpected failures in critical machinery.