Intelligent Fault Diagnosis Method for Gearboxes Based on Deep Transfer Learning

The complex operating environment of gearboxes and the easy interference early fault feature information make identification difficult. This paper proposes a diagnosis method based on combination whale optimization algorithm (WOA), variational mode decomposition (VMD), deep transfer learning. First, VMD is optimized by using WOA, minimum sample entropy used as fitness function to solve for K value penalty parameter α corresponding optimal VMD, correlation coefficient reconstruct signal. Second, reconstructed signal after reducing noise generate two-dimensional image continuous wavelet transform learning target domain data. Finally, AlexNet model object, which pretrained fine-tuned with parameters it suitable crack in gearboxes. experimental results show that proposed this can effectively reduce gearbox vibration signals under working environment, effective achieves high accuracy diagnosis.