基于双通道特征融合的CNN-GSWOA-XGBoost齿轮箱故障诊断方法

打开文本图片集

关键词：齿轮箱；卷积神经网络；优化鲸鱼算法；极端梯度提升树；故障诊断 中图分类号：U260.332；TH133 文献标志码：B doi：10.20213/j.cnki.tdcl.2025.02.08.01

CNN-GSWOA-XGBoost Gearbox Fault Diagnosis Method Based on Dual-channel Feature Fusion

SUN Qiang1，YAO Meng1，LI Chen1，WANG Mingming1，LI Yanzhou² （1. University， O37o09，China; 2. chnology，Xingtai O54Ooo，China）

Abstract：To solve the problems such as difficult extraction gearbox fault features low accuracy fault identification，the article proposes a CNN-GSWOA-XGBoost gearbox fault diagnosis method based on dual-channel feature fusion. Firstly，the original vibration signals are converted into GADF GASF timefrequency maps using GAF image encoding technology. Then，the GADF GASF time-frequency maps are input into a two-dimensional convolutional neural network（CNN）for adaptive fault feature extraction feature fusion， the results are taken as inputs to the XGBoost classifier. Secondly，aiming at the problems the whale optimization algorithm in convergence speed global search ability，an improved whale algorithm is used to optimize the five hyperparameters XGBoost classifier （interation ordinal number，depth trees，minimum weight sum child nodes，learning rate，sample ratio） establish a gearbox fault diagnosis model. Finally，the experimental analysis is conducted on the gearbox data from Southeast University， the results show thatthemodelcan fullyextract thefault features theoriginalvibration signals thegearbox thatcompared with other fault diagnosis models，the model has higher fault recognition accuracy superior performance.

Key words：gearbox；convolutional neural network；whale optimization algorithm；extreme gradient boosting tree;fault diagnosis

齿轮箱作为一种常见的旋转机械，由于其经常工作在复杂恶劣的环境下，不可避免会出现故障，从而严重影响设备的安全性能。（剩余8107字）