频谱密度函数相似性比较的齿轮箱故障诊断

摘要

In order to reduce the difficulty of gearbox vibration signal frequency spectrum analysis and fault identification,a gearbox fault diagnosis method based on the kernel density estimation and density function similarity comparison of frequency spectrum is proposed.Firstly,multi-set vibration signals for the gearbox under every fault condition are sampled.Then their frequency spectrum density functions are calculated by the kernel density estimation.Next,a portion of the density functions are selected,of which the arithmetic average is performed to obtain the standard density function for this kind of fault condition.Finally,according to the values of cosine similarity and correlation coefficient between the frequency spectrum density function of the test vibration signal and the standard density functions of various fault conditions,the gearbox fault corresponding to the test vibration signal will be identified.The experimental results indicate that compared to the method based on frequency spectral similarity comparison,the proposed method has higher accuracy for the gearbox fault identification.In addition,correlation coefficient shows greater differences among the different fault conditions of gearbox than cosine similarity,thus providing better applicability.%为降低齿轮箱振动信号频谱分析与故障识别的难度,提出了基于频谱核密度估计与密度函数相似性比较的齿轮箱故障诊断方法.首先针对齿轮箱的每一种故障状态采集多组振动信号,利用核密度估计方法对每组振动信号的频谱求取密度函数;然后选取一部分密度函数进行算术平均化,彳导到对应故障状态下的标准密度函数;最后根据测试振动信号频谱密度函数与各种故障状态标准密度函数之间的余弦相似度值与相关系数值,对齿轮箱的故障状态进行识别.试验结果表明:与振动信号的频谱相似性比较方法相比,所提方法对于齿轮箱故障状态的判别具有更高的准确率,同时对应于齿轮箱的不同故障状态,相关系数比余弦相似度显示出更大的差异性,具有更好的适用性.