摘要
针对非线性、非平稳的行星齿轮箱振动信号故障特征"难提取"和基于核参数随机生成的高斯核极限学习机状态辨识模型分类精度低的问题,提出一种改进多尺度排列熵(Enhence Multi-scale Permutation Entropy,EMPE)与核极化高斯核极限学习机(Kernel Extreme Learning Machine,KELM)结合的行星齿轮箱状态辨识方法。首先,将经由形态平均滤波的行星齿轮箱行星齿轮的振动信号,借助于EMPE来获取多尺度下的排列熵值(Permutation Entropy,PE)构建高维特征向量集;其次,利用核极化(Kernel Polarization,KP)优化高斯核极限学习机的核参数σ;最后,将EMPE特征向量集作为输入,通过KP优化KELM算法的训练建立行星齿轮状态辨识模型。实验结果表明,与基于SVM和KELM的状态辨识模型相比,基于EMPE和KP-KELM的行星齿轮故障诊断方法具有更高的分类精度。
According to the nonlinear and non-stationary of the planetary gearbox vibration signal fault characteristics is difficult to extract and the problem of low classification accuracy Gaussian kernel extreme learning machine based on random generation kernel parameters,a method for identifying the state of planetary gearboxes with enhence multi-scale permutation entropy(EMPE)and nuclear-polarized Gaussian kernel extreme learning machine(KELM)is proposed. Firstly,noise reduction of vibration signals of planetary gearbox planetary gears by morphological average filtering and using EMPE to obtain permutation entropy values at multiple scales to constructing eigenvector set. Secondly,kernel parameter σ of Gaussian kernel extreme learning machine is optimized by kernel polarization. Finally,using the EMPE eigenvector set as input,the planetary gear state identification model is constructed by the training of KP-KELM algorithm. The experiment results show that,compared with the fault classification model based on SVM and KELM,the EMPE and KP-KELM planetary gear fault diagnosis method has higher classification accuracy and stronger generalization ability.
引文
[1]冯志鹏,褚福磊,左明健.行星齿轮箱振动故障诊断方法[M].北京:科学出版社,2015:1-3.
[2]邵毅敏,陈再刚,周晓君,等.冲击振动能量通过“齿轮-轴-轴承-轴承座”多界面传递损耗研究[J].振动与冲击,2009,28(6):60-65.
[3]丁闯,张兵志,冯辅周,等.局部均值分解和排列熵在行星齿轮箱故障诊断中的应用[J].振动与冲击,2017,36(17):55-60.
[4]WWI Y,XU M,LI Y,et al.Gearbox fault diagnosis based on local mean decomposition,permutation entropy and extreme learning machine[J].Journal of Vibroengineering,2016,18(3):1459-1473.
[5]LI Y,LI G,YANG Y,et al.A fault diagnosis scheme for planetary gearboxes using adaptive multi-scale morphology filter and modified hierarchical permutation entropy[J].Mechanical Systems&Signal Processing,2018,105:319-337.
[6]WANG Y,CHENG Y.An Approach to Fault Diagnosis for Gearbox Based on Image Processing[J].Shock and Vibration,2016,2016(4):1-10.
[7]XING Z Q,QU J F,CHAI Y.Gear fault diagnosis under variable conditions with intrinsic time-scale decomposition-singular value decomposition and support vector machine[J].Journal of Mechanical Science and Technology,2017,31(2):545-553.
[8]AZIZ W,ARIF M.Multiscale Permutation Entropy of Physiological Time Series[C]//International Multitopic Conference,IEEE Inmic.New York:IEEE,2005:1-6.
[9]郑近德,程军圣,杨宇.多尺度排列熵及其在滚动轴承故障诊断中的应用[J].中国机械工程,2013,24(19):2641-2646.
[10]AZAMI H,ESCUDRERO J.Improved multiscale permutation entropy for biomedical signal analysis:Interpretation and application to electroencephalogram recordings[J].Biomedical Signal Processing and Control,2016,23:28-41.
[11]MATILLA M.A non-parametric test for independence based on symbolic dynamics[J].Journal of Economic Dynamics&Control,2007,31(12):3889-3903.
[12]LI Y B,XU M.Q,WEI Y,et al.A new rolling bearing fault diagnosis method based on multiscale permutation entropy and improved support vector machine based binary tree[J].Measurement,2016,77:80-94.
[13]ZHENG J,CHENG J,YANG Y.Multiscale permutation entropy based rolling bearing fault diagnosis[J].Shock and Vibration,2014(1):1-8.
[14]HUANG G B.An insight into extreme learning machines:Random neurons,random features and kernels[J].Cognitive Computation,2014,6(3):376-390.
[15]WANG T H,TIAN S F,HUANG H K,et al.Learning by local kernel polarization[J].Neurocomputing,2009,72(13/14/16):3077-3084.