一种具有强抗噪性的深度学习故障诊断模型

吴春志; 吴守军; 冯辅周; 朱俊臻

doi:10.7652/xjtuxb202104007

您当前的位置：

首页 >

文章列表页 >

一种具有强抗噪性的深度学习故障诊断模型

更新时间：2025-07-09

- 一种具有强抗噪性的深度学习故障诊断模型
- A Deep Learning Fault Diagnosis Model with Strong Anti-Noise Ability
- 西安交通大学学报 2021年55卷第4期页码：61-68
- 作者机构：
  
  陆军装甲兵学院车辆工程系,北京,100072
- 作者简介：
- 基金信息：
  
  : 2020-07-23。作者简介: 吴春志(1991—),男,博士生;冯辅周(通信作者),男,教授,博士生导师。基金项目: 国家自然科学基金资助项目(51875576)
- DOI：10.7652/xjtuxb202104007
  中图分类号： TP206
- 网络首发：2021-04-10，
  
  纸质出版：2021
- 稿件说明：
移动端阅览
吴春志, 吴守军, 冯辅周, 等. 一种具有强抗噪性的深度学习故障诊断模型[J]. 西安交通大学学报, 2021,55(4):61-68.

A Deep Learning Fault Diagnosis Model with Strong Anti-Noise Ability[J]. 2021, 55(4): 61-68.
吴春志, 吴守军, 冯辅周, 等. 一种具有强抗噪性的深度学习故障诊断模型[J]. 西安交通大学学报, 2021,55(4):61-68. DOI： 10.7652/xjtuxb202104007.

A Deep Learning Fault Diagnosis Model with Strong Anti-Noise Ability[J]. 2021, 55(4): 61-68. DOI： 10.7652/xjtuxb202104007.

摘要

针对旋转机械运行环境噪声大

现有的基于深度学习的故障诊断模型抵抗强噪声干扰能力不足的问题

提出了一种具有强抗噪性的深度学习故障诊断模型(HAT-MSCNN)。首先将齿轮箱训练集正常状态下样本的频谱信号由大到小排列

取排序的第k个的幅值α为基准值

将训练集中所有样本的频谱信号低于α的值置0

降低了易与噪声耦合的低幅值频率成分对诊断模型的干扰

突出了高幅值频率成分对故障诊断模型的贡献

同时保留零值所在的位置信息。然后将经过此处理的训练集输入到由多种移动步长卷积核组合的多尺度卷积神经网络中进行训练

最终得到强抗噪性的故障诊断模型。以精确率和召回率的调和平均值F作为模型的评价指标

并使用加权梯度类激活映射方法可视化模型的分类权重

在具有丰富故障模式的PHM2009齿轮箱数据集上验证模型。将所提模型与以原始振动信号作为输入的Raw-MSCNN模型、以频谱信号作为输入的Frequency-MSCNN模型、以集合经验模态分解(EEMD)降噪后的信号作为输入的EEMD-MSCNN模型和以所提训练集样本构造方法处理后的信号作为输入的HAT-CNN模型进行对比

结果表明

所提模型在测试集信噪比为-4 dB时

诊断的F值为98.6%

比第二名的Frequency-MSCNN高了6.4%

在更强噪声干扰的测试集信噪比为-6 dB时

F值为89.1%

比Frequency-MSCNN高了10.9%

证明了所提的HAT-MSCNN模型具有在强噪声干扰下的抗噪能力。

Abstract

A deep learning fault diagnosis model with strong anti-noise ability(HAT-MSCNN)is proposed to solve the problem that the existing fault diagnosis model based on deep learning has insufficient resistance to strong noise interference due to the high noise in the operating environment of rotating machinery. Firstly

the frequency spectrum signals of the samples under normal state of the gearbox training set are arranged from large to small

and the kth amplitude value α of the sort is taken as the reference value. The frequency spectrum signals of all samples in the training set with values being lower than the value of α are set to zero

which reduces the interference of low-amplitude frequency components that are easy to be coupled with noise on the diagnosis model

and highlights the contribution of high-amplitude frequency components to the fault diagnosis model and retains the position information where the zero value is located. Then the processed training set is inputted into a multi-scale convolutional neural network composed of multiple moving step size convolution kernels for training

and finally a fault diagnosis model with strong noise immunity is obtained. The harmonic average value F of precisions and recalls is used as an evaluation index of the model

the weighted gradient activation mapping method is used to visualize the classification weights of the model

and the model is verified on the PHM2009 gearbox data set with rich failure modes. The proposed method is compared with the Raw-MSCNN

Frequency-MSCNN

and EEMD-MSCNN models that use vibration signals

frequency signals

and EEMD denoised signals as input respectively

and the HAT-CNN model that uses the signal processed by the proposed training set sample construction method as input. The results show that when the SNR of the test set is -4 dB

the F value of the proposed HAT-MSCNN is 98.6%

which is 6.4% higher than that of the second-place Frequency-MSCNN. When the SNR of the test set with stronger noise interference is -6 dB

the F value of the proposed HAT-MSCNN is 89.1%

which is 10.9% higher than that of Frequency-MSCNN. It proves that the proposed model has the anti-noise capability under strong noise interference.

关键词

Keywords

references

LI Chuan, SANCHEZ R V, ZURITA G, et al. Gearbox fault diagnosis based on deep random forest fusion of acoustic and vibratory signals [J]. Mechanical Systems and Signal Processing, 2016, 76/77(8): 283-293.

汤宝平, 熊学嫣, 赵明航, 等. 多共振分量融合CNN的行星齿轮箱故障诊断 [J]. 振动、测试与诊断, 2020, 40(3): 507-512.

TANG Baoping, XIONG Xueyan, ZHAO Minghang, et al. A multi-resonance component fusion based convolutional neural network for diagnosis of planetary gearboxes [J]. Journal of Vibration, Measurement Diagnosis, 2020, 40(3): 507-512.

周兴康, 余建波. 基于深度一维残差卷积自编码网络的齿轮箱故障诊断 [J]. 机械工程学报, 2020, 56(7): 96-108.

ZHOU Xingkang, YU Jianbo. Gearbox fault diagnosis based on one-dimension residual convolutional auto-encoder [J]. Journal of Mechanical Engineering, 2020, 56(7): 96-108.

魏巍宏. 基于振动监测和油液分析的齿轮箱多参数故障诊断方法优化研究 [D]. 江苏徐州: 中国矿业大学, 2019: 7-8.

张弘斌, 袁奇, 赵柄锡, 等. 采用多通道样本和深度卷积神经网络的轴承故障诊断方法 [J]. 西安交通大学学报, 2020, 54(8): 58-66.

ZHANG Hongbin, YUAN Qi, ZHAO Bingxi, et al. A bearing fault diagnosis with multi-channel sample and deep convolutional neural network [J]. Journal of Xi'an Jiaotong University, 2020, 54(8): 58-66.

卢昱奇. 基于卷积神经网络的行星齿轮箱复合故障诊断方法研究 [D]. 成都: 电子科技大学, 2020: 2-4.

丁闯, 冯辅周, 张兵志, 等. 行星变速箱振动信号的线性量子信息熵特征 [J]. 兵工学报, 2018, 39(12): 2306-2312.

DING Chuang, FENG Fuzhou, ZHANG Bingzhi, et al. Feature of linear quantum information entropy for vibration signals of planetary transmission gearbox [J]. Acta Armamentarii, 2018, 39(12): 2306-2312.

JANSSENS O, SLAVKOVIKJ V, VERVISCH B, et al. Convolutional neural network based fault detection for rotating machinery [J]. Journal of Sound Vibration, 2016, 377: 331-345.

DING Xiaoxi, HE Qingbo. Energy-fluctuated multiscale feature learning with deep ConvNet for intelligent spindle bearing fault diagnosis [J]. IEEE Transactions on Instrumentation Measurement, 2017, 66(8): 1926-1935.

万齐杨, 熊邦书, 李新民, 等. 基于DCAE-CNN的自动倾斜器滚动轴承故障诊断 [J]. 振动与冲击, 2020, 39(11): 273-279.

WAN Qiyang, XIONG Bangshu, LI Xinmin, et al. Fault diagnosis for rolling bearing of swashplate based on DCAE-CNN [J]. Journal of Vibration and Shock, 2020, 39(11): 273-279.

JING Luyang, ZHAO Ming, LI Pin, et al. A convolutional neural network based feature learning and fault diagnosis method for the condition monitoring of gearbox [J]. Measurement, 2017, 111: 1-10.

李涛, 段礼祥, 张东宁, 等. 自适应卷积神经网络在旋转机械故障诊断中的应用 [J]. 振动与冲击, 2020, 39(16): 275-282, 288.

LI Tao, DUAN Lixiang, ZHANG Dongning, et al. Application of adaptive convolutional neural network in rotating machinery fault diagnosis [J]. Journal of Vibration and Shock, 2020, 39(16): 275-282, 288.

刘冲. 基于联合抗噪算法的滚动轴承故障诊断研究 [J]. 华东交通大学学报, 2020, 37(4): 82-87.

LIU Chong. Research on joint anti-noise algorithm and its application in rolling bearing fault diagnosis [J]. Journal of East China Jiaotong University, 2020, 37(4): 82-87.

吴静然, 丁恩杰, 崔冉, 等. 采用多尺度注意力机制的旋转机械故障诊断方法 [J]. 西安交通大学学报, 2020, 54(2): 51-58.

WU Jingran, DING Enjie, CUI Ran, et al. A diagnostic approach for rotating machinery using multi-scale feature attention mechanism [J]. Journal of Xi'an Jiaotong University, 2020, 54(2): 51-58.

赵小强, 梁浩鹏. 使用改进残差神经网络的滚动轴承变工况故障诊断方法 [J]. 西安交通大学学报, 2020, 54(9): 23-31.

ZHAO Xiaoqiang, LIANG Haopeng. Fault diagnosis method for rolling bearing under variable working conditions using improved residual [J]. Journal of Xi'an Jiaotong University, 2020, 54(9): 23-31.

彭鹏, 柯梁亮, 汪久根. 噪声干扰下的RV减速器故障诊断 [J]. 机械工程学报, 2020, 56(1): 30-36.

PENG Peng, KE Liangliang, WANG Jiugen. Fault diagnosis of RV reducer with noise interference [J]. Journal of Mechanical Engineering, 2020, 56(1): 30-36.

WU Chunzhi, JIANG Pengcheng, DING Chuang, et al. Intelligent fault diagnosis of rotating machinery based on one-dimensional convolutional neural network [J]. Computers in Industry, 2019, 108: 53-61.

WU Chunzhi, FENG Fuzhou, WU Shoujun, et al. Research on multi-scale batch normalized convolutional neural networks for gearbox fault diagnosis [C]∥10th Prognostics and System Health Management Conference. Piscataway, NJ, USA: IEEE, 2019: 8942850.

SELVARAJU R R, COGSWELL M, DAS A, et al. Grad-CAM: visual explanations from deep networks via gradient-based localization [J]. International Journal of Computer Vision, 2019, 128(8): 618-626.

PHM Society. PHM09_competition_1 [DB/OL].[2017-12-01]. https:∥c3.nasa.gov/dashlink/resour ces/997.

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

蜉蝣优化双通道网络在齿轮箱故障诊断中的应用

利用改进卷积神经网络的滚动轴承变工况故障诊断方法

采用参数化解调的变转速下柱塞泵故障诊断方法

一维卷积神经网络在往复式压缩机气阀故障诊断中的应用

电力变压器故障诊断多智能体大模型构建与应用