Analysis on fault diagnosis of gearbox based on FDBO+Informer-ECANet
LI Tingting
JIA Dong
Abstract：[Objective]The fault diagnosis method for gearboxes based on intelligent optimization algorithms and deep neural networks has gradually become a research hotspot,but there are still many problems.To address the challenges of fault feature extraction and low diagnostic accuracy for gearboxes under strong noise environments,a novel gearbox fault diagnosis method was proposed based on a fusion-enhanced dung beetle optimization(FDBO)algorithm,the Informer model,and the efficient channel attention network(ECANet)module.[Methods]Firstly,to overcome the limitations of the conventional dung beetle optimization(DBO)algorithm,such as insufficient global search capability and tendency to fall into local optima,a fusion strategy integrating Fuch chaotic mapping combined with inverse learning,an adaptive step size strategy,convex lens imaging,and a stochastic differential mutation strategy was introduced,significantly enhancing the algorithm's global search performance.Secondly,benefiting from its excellent long-term time series processing capability,the Informer model was enabled to efficiently extract global features and local features from sequence data;especially for fault signals involving long-term dependencies,the model was able to demonstrate extremely high classification performance.Thirdly,an ECANet module was incorporated into the encoder of the Informer model to perform channel-wise adaptive calibration of extracted features,enhancing the model's attention to critical features,strengthening feature representation,and reducing noise interference.Finally,the FDBO algorithm was employed to optimize multiple hyperparameters of the Informer-ECANet model,determining the optimal parameter combination to improve the diagnostic accuracy and generalization capability of the model.[Results]Test results demonstrate that the proposed method achieves an accuracy of 100%under noise-free conditions,and maintains a high accuracy of 94.4%even when subjected to Gaussian white noise at a-6 dB signal-to-noise ratio,thereby validating the superior performance and robustness of the model.This study provides an effective intelligent approach for gearbox fault diagnosis under challenging noisy environment.
Keywords：Fusion-enhanced dung beetle optimization algorithmInformer modelECANet moduleStochastic differen-tial variation strategy
Publication Date：2026-03-15
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：11( 161-171 )