基于频率切片小波变换和注意力增强ConvNeXt模型的行星齿轮箱故障诊断

doi:10.12382/bgxb.2024.0329

摘要/Abstract

摘要：

针对传统手工提取故障特征过度依赖专家的先验知识,导致信息提取不完全、效率低、成本高、漏诊误诊的问题,提出一种基于频率切片小波变换(Frequency Slice Wavelet Transform,FSWT)和注意力增强ConvNeXt模型的新方法,用于行星齿轮箱故障诊断。该方法在ConvNeXt模型基础上融合卷积注意力模块(Convolutional Block Attention Module,CBAM),使网络更加聚焦于关键区域的特征,减小无关目标的干扰。通过FSWT将一维振动信号转化为具有二维特征的时频谱图像,输入到改进后的网络中进行自动特征提取,并建立特征空间与故障空间之间的映射关系,实现不同故障模式的准确区分。利用动力传动模拟实验台数据对所提方法进行实验验证,结果表明:相较于其他网络模型,改进后的ConvNeXt模型能够准确识别出齿轮特定类型的故障,且噪声干扰下依旧展现出良好的鲁棒性。所得研究成果可为行星齿轮箱智能故障诊断提供参考。

关键词: 行星齿轮箱, 频率切片小波变换, 注意力机制, ConvNeXt模型, 故障诊断

Abstract:

The excessive reliance on experts prior knowledge in traditional manual fault feature extraction leads to incomplete information extraction,low efficiency,high cost,and missed or misdiagnosed cases.A fault diagnosis method based on frequency slice wavelet transform (FSWT) and attention-enhanced ConvNeXt is proposed for planetary gearboxes.This method integrates the convolutional block attention module (CBAM) into the ConvNeXt model,enabling the network to focus more on key regional features and reduce the interference from irrelevant targets.The one-dimensional vibration signals are transformed into the two-dimensional time-frequency spectrum images by applying FSWT,which are then input into the improved network for automatic feature extraction.A mapping relationship between the feature space and the fault space is established to accurately distinguish different fault modes.The proposed method is validated experimentally using the data from a dynamic drive simulation experimental platform.The results show that compared to other network models,the improved ConvNeXt model can accurately identify specific types of gear faults and exhibit good robustness even under noise interference.The research findings provide valuable reference for intelligent fault diagnosis of planetary gearboxes.

Key words: planetary gearbox, frequency slice wavelet transform, attention mechanism, ConvNeXt model, fault diagnosis

崔素晓, 武哲, 崔彦平, 张强, 赵月静. 基于频率切片小波变换和注意力增强ConvNeXt模型的行星齿轮箱故障诊断[J]. 兵工学报, 2025, 46(2): 240329-.

CUI Suxiao, WU Zhe, CUI Yanping, ZHANG Qiang, ZHAO Yuejing. Fault Diagnosis of Planetary Gearbox Based on Frequency Slice Wavelet Transform and Attention-enhanced ConvNeXt Model[J]. Acta Armamentarii, 2025, 46(2): 240329-.

图/表 15

图1 FSWT转化过程示意图

Fig.1 Schematic diagram of FSWT conversion process

图2 CBAM结构图

Fig.2 CBAM structure diagram

图3 CBAM-ConvNeXt模型结构

Fig.3 Structure of CBAM-ConvNeXt model

图4 智能故障诊断流程

Fig.4 Flow chart of intelligent fault diagnosis

图5 动力传动模拟实验台

Fig.5 Power transmission simulation test bench

表1 行星齿轮箱规格

Table 1 Specification of planetary gearbox

部件	齿轮齿数
部件	第1级	第2级
齿圈	100	100
行星轮	40×3	36×4
太阳轮	20	28

图6 5种工况下振动信号的时域波形图及频谱图

Fig.6 Time-domain waveforms and spectra of vibration signals under 5 working conditions

图7 5种工况下振动信号的FSWT

Fig.7 FSWT diagram of vibration signals under 5 working conditions

图8 不同二维时频图转化方法诊断结果对比

Fig.8 Comparison of diagnostic results of different two-dimensional time-frequency image conversion methods

图9 CBAM-ConvNeXt模型的准确率和损失率

Fig.9 Accuracy and loss rate curves of CBAM-ConvNeXt model

图10 特征降维可视化

Fig.10 Feature dimensionality reduction visualization

图11 各对比网络模型的准确率和损失率

Fig.11 Accuracy and loss rate curves of each comparison network model

表2 各网络模型测试集评价指标结果对比

Table 2 Comparison of evaluation indicators for different network model test sets %

网络模型	准确率	精确率	召回率	F₁
CBAM-ConvNeXt模型	99.98	99.98	99.98	99.98
ConvNeXt-T模型	99.89	99.88	99.90	99.89
Resnet34模型	99.10	99.10	99.10	99.10
Densnet121模型	95.21	95.34	95.20	95.27
Swin Transformer网络模型	99.31	99.32	99.30	99.31

图12 各网络模型的混淆矩阵对比

Fig.12 Comparison of confusion matrices for different network models

图13 5dB信噪比下各网络模型的准确率

Fig.13 Accuracy of each network model at 5dB signal-to-noise ratio

参考文献 20

[1]	赵文辉, 孙晓恒, 张伟东, 等. 四旋翼飞行器齿轮箱-支臂组件动态特性分析[J]. 兵工学报, 2022, 43(5):1175-1184. doi: 10.12382/bgxb.2021.0001
	ZHAO W H, SUN X H, ZHANG W D, et al. Analysis of dynamic characteristics of gearbox support arm assembly for quadrotor aircraft[J]. Acta Armamentarii, 2022, 43(5):1175-1184. (in Chinese)
[2]	TANG S N, YUAN S Q, ZHU Y. Deep learning-based intelligent fault diagnosis methods toward rotating machinery[J]. IEEE Access, 2019, 8:9335-9346.
[3]	SHI J C, PENG D K, PENG Z X, et al. Planetary gearbox fault diagnosis using bidirectional-convolutional LSTM networks[J]. Mechanical Systems and Signal Processing, 2022, 162:107996.
[4]	王卫国, 孙磊. 基于EEMD-CWD的齿轮箱振动信号故障特征提取[J]. 兵工学报, 2014, 35(8):1288-1294. doi: 10.3969/j.issn.1000-1093.2014.08.023
	WANG W G, SUN L. Fault feature extraction of gearbox vibration signals based on EEMD-CWD[J]. Acta Armamentarii, 2014, 35(8):1288-1294. (in Chinese)
[5]	BRITO L C, SUSTO G A, BRITO J N, et al. An explainable artificial intelligence approach for unsupervised fault detection and diagnosis in rotating machinery[J]. Mechanical Systems and Signal Processing, 2022, 163:108105.
[6]	YANG J L, YIN S Y, CHANG Y Q, et al. An efficient method for monitoring degradation and predicting the remaining useful life of mechanical rotating components[J]. IEEE Transactions on Instrumentation and Measurement, 2020, 70:3504214.
[7]	SAHA D K, AHMED S, SHAUROV M S. Different machine maintenance techniques of rotary machine and their future scopes:a review[C]// Proceedings of 2019 4th International Conference on Electrical Information and Communication Technology. Khulna,Bangladesh: IEEE, 2019:1-6.
[8]	MOHAMMED S A, GHAZALY N M, ABDO J. Fault diagnosis of crack on gearbox using vibration-based approaches[J]. Symmetry, 2022, 14(2):417.
[9]	BARBIERI N, BARBIERI G S A V, MARTINS B M, et al. Analysis of automotive gearbox faults using vibration signal[J]. Mechanical Systems and Signal Processing, 2019, 129:148-163. doi: 10.1016/j.ymssp.2019.04.028
[10]	LI Y Z, DING K, HE G L, et al. Non-stationary vibration feature extraction method based on sparse decomposition and order tracking for gearbox fault diagnosis[J]. Measurement, 2018, 124:453-469.
[11]	熊鹏, 汤宝平, 邓蕾, 等. 基于动态加权密集连接卷积网络的变转速行星齿轮箱故障诊断[J]. 机械工程学报, 2019, 55(07):52-57.
	XIONG P, TANG B P, DENG L. Fault diagnosis of variable speed planetary gearbox based on dynamic weighted dense connected convolutional network[J]. Journal of Mechanical Engineering, 2019, 55(7):52-57. (in Chinese) doi: 10.3901/JME.2019.07.052
[12]	YAO G, WANG Y C, BENBOUZID M, et al. A hybrid gearbox fault diagnosis method based on GWO-VMD and DE-KELM[J]. Applied Sciences, 2021, 11(11):4996.
[13]	王辉, 徐佳文, 严如强. 基于多尺度注意力深度强化学习网络的行星齿轮箱智能诊断方法[J]. 机械工程学报, 2022, 58(11):133-142. doi: 10.3901/JME.2022.11.133
	WANG H, XÜ J W, YAN R Q. Intelligent diagnosis method for planetary gearbox based on multi-scale attention deep reinforcement learning network[J]. Journal of Mechanical Engineering, 2022, 58(11):133-142. doi: 10.3901/JME.2022.11.133
[14]	RAVIKUMAR K N, YADAV A, KUMAR H, et al. Gearbox fault diagnosis based on Multi-Scale deep residual learning and stacked LSTM model[J]. Measurement, 2021, 186:110099.
[15]	黄佳欢, 余建波. 基于形态帽积卷积自编码器模型的齿轮箱故障诊断[J]. 机械工程学报, 2023, 59(23):132-145. doi: 10.3901/JME.2023.23.132
	HUANG J H, YÜ J B. Gearbox fault diagnosis based on morphological cap convolutional autoencoder model[J]. Journal of Mechanical Engineering, 2023, 59(23):132-145. (in Chinese)
[16]	师恩容. 采用改进Transformer的挤压机齿轮箱故障诊断技术研究[J]. 制造业自动化, 2024, 46(3):142-146.
	SHI E R. Research on fault diagnosis technology for extruder gearbox using improved Transformer[J]. Manufacturing Automation, 2024, 46(3):142-146.
[17]	古莹奎, 石昌武, 陈家芳. 基于格拉姆角场与深度卷积生成对抗网络的行星齿轮箱故障诊断[J]. 噪声与振动控制, 2024, 44(1):111-118.
	GU Y K, SHI C W, CHEN J F. Fault diagnosis of planetary gearbox based on Gram angle field and deep convolutional generative adversarial network[J]. Noise and Vibration Control, 2024, 44(1):111-118. (in Chinese)
[18]	CHEN Z Y, MAURICIO A, LI W H, et al. A deep learning method for bearing fault diagnosis based on cyclic spectral coherence and convolutional neural networks[J]. Mechanical Systems and Signal Processing, 2020, 140:106683.
[19]	ZHANG L, ZHANG H, CAI G W. The multiclass fault diagnosis of wind turbine bearing based on multisource signal fusion and deep learning generative model[J]. IEEE Transactions on Instrumentation and Measurement, 2022, 71:3514212.
[20]	YAN Z H, MIYAMOTO A, JIANG Z W, et al. An overall theoretical description of frequency slice wavelet transform[J]. Mechanical Systems and Signal Processing, 2010,491-507.