Litcius/Paper detail

Intelligent Fault Diagnosis Method for Gearboxes Based on Deep Transfer Learning

Zhenghao Wu, Huajun Bai, Hao Yan, Xianbiao Zhan, Chiming Guo, Xisheng Jia

2022Processes18 citationsDOIOpen Access PDF

Abstract

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

Topics & Concepts

Computer scienceFault (geology)Artificial intelligenceTransfer functionPattern recognition (psychology)WaveletInterference (communication)Transfer of learningSIGNAL (programming language)Noise (video)Feature (linguistics)AlgorithmEngineeringImage (mathematics)GeologyPhilosophySeismologyComputer networkProgramming languageElectrical engineeringChannel (broadcasting)LinguisticsMachine Fault Diagnosis TechniquesGear and Bearing Dynamics AnalysisEngineering Diagnostics and Reliability