基于两类特征和最小二乘支持向量机的齿轮故障诊断方法

发布时间：2019-02-12 07:27

【摘要】：针对齿轮振动信号非线性非平稳特性,为避免传统"时-频"分析方法在表征设备状态时的不足和样本数量少易造成故障辨识模型"欠学习"的问题,提出一种基于峭度、本征模式分量(Intrinsic mode function,IMF)能量两类特征和最小二乘支持向量机(Least squares support vector machine,LS-SVM)的齿轮故障诊断方法。首先,对所测齿轮振动信号在集合经验模式分解(Ensemble empirical mode decomposition,EEMD)的基础上提取有效IMF分量计算其能量特征和峭度值,据此构建时频域两类特征向量;其次,将融合后的齿轮正常、齿根裂纹、断齿3种状态下的时频域两类特征向量作为输入,基于LS-SVM建立齿轮故障诊断模型,进行齿轮故障识别。实验结果表明,该方法能够准确地识别齿轮的工作状态,与基于BP、SVM的故障诊断模型相比,其具有更高的识别率,为齿轮状态识别和故障诊断提供了一种新途径。
[Abstract]:Aiming at the nonlinear and non-stationary characteristics of gear vibration signal, a kurtosis based on kurtosis is proposed in order to avoid the deficiency of the traditional time-frequency analysis method in characterizing the state of the equipment and the problem that the small number of samples can easily lead to the problem of "underlearning" in the fault identification model. Two kinds of characteristic of intrinsic mode component (Intrinsic mode function,IMF (energy) and least square support vector machine (Least squares support vector machine,LS-SVM) are used for gear fault diagnosis. Firstly, the effective IMF component is extracted from the measured gear vibration signal on the basis of the set empirical mode decomposition (Ensemble empirical mode decomposition,EEMD) to calculate its energy characteristics and kurtosis, and then two kinds of eigenvectors in time-frequency domain are constructed. Secondly, two kinds of characteristic vectors in the time-frequency domain of the fused gear under normal condition, root crack and broken tooth are taken as input, and the gear fault diagnosis model is established based on LS-SVM to identify the gear fault. The experimental results show that the method can accurately identify the working state of gears, and it has a higher recognition rate than the fault diagnosis model based on BP,SVM, which provides a new way for gear state identification and fault diagnosis.
【作者单位】：内蒙古科技大学机械工程学院;浙江大学信息学部控制科学学院;包头钢铁(集团)有限公司设备动力部;
【基金】：国家自然科学基金(51565046) 内蒙古自然科学基金(2015MS0512) 内蒙古高等学校科学研究项目(NJZY146) 内蒙古科技大学创新基金(2015QDL12)
【分类号】：TH132.41

【相似文献】