A fault diagnosis method for variable speed planetary gearbox based on ADGADF and Swin Transformer
Huihui Wang,Zhe Wu,Qi Li,Yanping Cui,Suxiao Cui
DOI: https://doi.org/10.1784/insi.2024.66.4.232
2024-04-01
Insight - Non-Destructive Testing and Condition Monitoring
Abstract:The vibration signal of planetary gearboxes under variable speed conditions shows non-stationary characteristics, indicating that fault diagnosis has become more complex and challenging. In order to more accurately diagnose faults in planetary gearboxes under variable speed conditions, a new method is proposed based on the angular domain Gramian angular difference field (ADGADF) and Swin Transformer. This method initially employs the chirplet path pursuit (CPP) algorithm to fit the speed curve of the original time-domain signal and then combines the speed curve with computed order tracking (COT) to achieve equal angle resampling of the time-domain signal, obtaining a stationary signal in the angular domain. On the basis of the above, the angular domain signal is creatively encoded into the two-dimensional images using the Gramian angular field (GAF), which accurately represents the fault characteristics of the original signal. Finally, the Swin Transformer network, with efficient global feature extraction capability, is used to learn advanced features from the images, achieving accurate fault recognition and classification. The proposed method is verified by experiment on the planetary gearbox and its performance is compared with several common coding methods and intelligent diagnosis algorithms. The experimental results show that the proposed method reaches an accuracy of up to 99.8%. In addition, its performance in accuracy, precision, recall, F 1-score and the confusion matrix is superior to traditional diagnostic methods. It also offers the advantage of strong robustness.
instruments & instrumentation,materials science, characterization & testing