Fault diagnosis of bearings under variable working conditions based on strip pooling VGG19

To address the challenge of extracting fault features and achieving accurate classification of rolling bearings under time-varying speed conditions, this paper proposes a fault diagnosis model based on a StripPooling-improved VGG19 network. Vibration signals are first converted into images through mapping. The VGG19 network is then enhanced using the StripPooling strategy, and the resulting RGB feature maps are input into the SP-VGG19 model for fault classification. The model is experimentally validated using the time-varying bearing dataset from the University of Ottawa. Results show that the proposed model accurately classifies faults under mixed operating conditions and demonstrates good generalization.