Fault Diagnosis Method Based on Multi-Attention Convolutional Neural Network

Rolling bearings play an essential role in mechanical systems, yet their fault vibrations are often obscured by noise and other forms of interference. A refined convolutional neural network approach has been introduced to bolster fault diagnosis reliability and decrease fault frequency in equipment operations to tackle this challenge. The proposed algorithm incorporates several attention mechanism modules into the convolutional neural network structure to create a lightweight and accurate model. The enhanced algorithm has been shown to attain a remarkable accuracy rate of 96.14% in experimental evaluations.