College of Automation,Shenyang Aerospace University,Shenyang 110136,China
Show less
文章历史+
Received
Accepted
Published
2022-09-08
Issue Date
2025-08-05
PDF (1015K)
摘要
针对滚动轴承在实际运行环境中的振动信号复杂且具有非平稳性等问题,为提高滚动轴承故障诊断的准确性,提出了一种基于参数优化变分模态分解(Variational Mode Decomposition,VMD)和卷积神经网络(Convolutional Neural Networks,CNN)的滚动轴承故障诊断方法。在特征提取上,提出灰狼优化(Grey Wolf Optimization,GWO)算法对VMD进行优化,确定VMD中模态分量个数K及惩罚参数α的最佳组合,并将原始振动信号分解获得K个特征模态分量(Intrinsic Mode Function,IMF),根据峭度指标进行筛选,从而提取最佳主成分特征;在故障诊断上,构建参数优化后的卷积神经网络故障诊断模型,将筛选出的模态分量IMF转化为特征向量作为卷积神经网络故障诊断模型的输入,达到对故障状态准确识别的目的。通过与传统方法诊断结果的比较,极大提高了滚动轴承的诊断准确率,证明了该方法的可行性。
Abstract
In view of the problems such as complex vibration signals and non-stationarity of rolling bearing in the actual operating environment, in order to improve the accuracy of fault diagnosis of rolling bearing, a new method based on parameter optimization Variational Mode Decomposition (VMD) and convolutional neural Network(CNN)rolling bearing fault diagnosis method was proposed. In feature extraction, a Grey Wolf Optimization (GWO) algorithm was proposed to optimize the VMD, the optimal combination of the number of modal components K and the penalty parameter α in the VMD was determined, and decomposed the original vibration signal to obtain K.The eigenmode component (Intrinsic Mode Function, IMF) was screened according to the kurtosis index to extract the best principal component features; in fault diagnosis, a convolutional neural network fault diagnosis model with optimized parameters was constructed to select the modal.The component IMF was transformed into a feature vector as the input of the convolutional neural network fault diagnosis model, so as to achieve the purpose of accurately identify the fault state. By comparing the diagnostic results with the traditional method, the diagnostic accuracy of the rolling bearing was greatly improved, the feasibility of the method was proved.
(5) 构建故障诊断模型:通过构建特征向量 F 并输入到卷积神经网络故障诊断模型中进行训练,确定卷积核数量n和卷积核尺寸c×r的大小。先在卷积层提取特征,在完成卷积计算后采用修正线性单元(Relu)激活函数,对卷积计算后的结果进行非线性映射来提高模型的训练速度和精度;再输送到池化层中,选用最大池化方法消除无关信息并减少模型参数的数量;最后输送到全连接层特征分类,并利用测试样本对训练模型进行测试和验证,完成对滚动轴承的故障诊断。
JiaoZ Y, PanL R, FanW,et al.Partly interpretable transformer through binary arborescent filter for intelligent bearing fault diagnosis[EB/OL].(2022-08-16)[2022-08-22].
[13]
GuoZ R, YangM, HuangX. Bearing fault diagnos- is based on speed signal and CNN model [J]. Energy Reports, 2022, 8(13):904-913.
[14]
HanC K, LuW, WangP X,et al.A recursive sparse representation strategy for bearing fault diagnosis[EB/OL].(2021-10-25)[2022-08-22].
[15]
JinZ Z, HeD Q, WeiZ X.Intelligent fault diagnosis of train axle box bearing based on parameter optimization VMD and improved DBN[EB/OL].(2022-02-10)[2022-03-15].
[16]
LuoS Y, HuangX F, WangY Z,et al.Transfer learning based on improved stacked autoencoder for bearing fault diagnosis[EB/OL].(2022-09-03)[2022-09-04].
[17]
YuC M, WangF, LiuY H,et al.Research on knowledge graph alignment model based on deep learning[J].Expert Systems with Applications,2021,21(8):3456-3462.
[18]
KumarA, KumarR.Role of signal processing,modeling and decision making in the diagnosis of rolling element bearing defect:a review[J].Journal of Nondestructive Evaluation,2019,38(1):5.
[19]
LiX, ZhangW, DingQ, et al. Multi⁃layer domain adaptation method for rolling bearing fault diagnosis [J]. Signal Processing, 2019, 157:180-197.
[20]
HeM, HeD.Deep learning based approach for bearing fault diagnosis[J].IEEE Transactions on Industry Applications,2017,53(3):3057-3065.