摘要
为了实现滚动轴承故障的智能诊断,提出了一种基于经验模态分解(Empirical mode decomposition,EMD)和深度信念网络(Deep belief network,DBN)的轴承故障诊断模型。首先,采用经验模态分解对振动信号进行处理,选取有效的本征模态函数(Intrinsic mode function,IMF)分量及其Hilbert包络谱、边际谱,计算其统计参数,构造原始特征集;然后,提出了一种基于极限学习机(Extreme learning machine,ELM)的特征选择方法 (Features selection base on ELM,FSELM),以去除原始特征集中的冗余和干扰特征,选取出故障状态敏感特征;最后,利用深度学习在高维、非线性信号处理方面的优势,完成基于DBN的故障特征自适应分析与故障状态智能识别。通过对12种轴承状态进行分类实验,表明FSELM方法能够选取出故障的敏感统计特征,DBN方法的自适应特性能够有效提高故障状态识别准确率。
In order to realize the intelligent diagnosis of rolling bearing failure,a fault diagnosis model of vibration signal based on empirical mode decomposition（ EMD） and deep belief network（ DBN） is proposed.Firstly,the vibration signal is processed by empirical mode decomposition,and the statistical parameters of the effective intrinsic modal function（ IMF） component and its Hilbert envelope are obtained as the original feature set. Then,the extreme learning suite（ ELM） classifier feature selection method is proposed to remove the redundancy and interference characteristics of the original feature set and to select the fault state sensitive features. Finally,by using the depth learning advantages in high-dimensional and non-linear processing,the DBN-based fault feature adaptive analysis and fault state intelligent identification is completed. The results show that the ELM method can select the sensitive statistical characteristics of the fault,and the adaptive characteristic of the DBN method can effectively improve the accuracy of fault state recognition.
作者
俞啸
范春旸
董飞
丁恩杰
吴守鹏
王昕
YU Xiao;Fan Chunyang;Dong Fei;Ding Enjie;Wu Shoupeng;Wang Xin(IOT Perception Mine Research Center,China University of Mining and Technology,Xuzhou 221008,China;School of Information and Control Engineering,China University of Mining and Technology,Xuzhou 221008,China;School of Medicine Information,Xuzhou Medical University,Xuzhou 221009,China;School of Information Media and Art,Jiangsu Vocational Institute of Architectural Technology,Xuzhou 221008,China)
出处
《机械传动》
CSCD
北大核心
2018年第6期157-163,共7页
Journal of Mechanical Transmission
基金
国家重点研发计划:"矿山安全生产物联网关键技术与装备研发"(2017YFC0804400
2017YFC0804401)
国家重点基础研究发展计划(973):"深部危险煤层无人采掘装备关键基础研究"(2014CB046300)
关键词
经验模态分解
极限学习机
深度信念网络
滚动轴承
故障诊断
Empirical modal decomposition
Extreme learning machine
Deep belief network
Rollingbearing
Fault diagnosis
