基于单值中智集相似性及熵度量的轴承故障诊断

Bearing Fault Diagnosis Based on Similarity and Entropy Measure of Single Value Neutrosophic Sets

文章以轴承的故障识别和分类作为研究对象,建立特征提取和分类方法。首先通过小波包分解得到各频段的能量谱,根据多个样本的能量谱得到已知故障类型单值中智集(SVNS)矩阵和实验数据的SVNS。其次,通过SVNS理论,建立了三种分类计算方式,分别是:Euclidean距离、余弦相似度、单值中智集对称交叉熵(SVNSSCE)。最后,实验数据证明了文章采用方法的可行性和有效性,故障识别率达到96%以上,且余弦相似度、SVNSSCE分类方式相对于Euclidean距离在文章数据集中有更好效果。文章分类方法具有以数据驱动为主、人工参数少、计算量小的特点。

Bearing fault recognition and classification was the research object in the paper and it established the method of feature extraction and classification.Firstly,the energy spectrum of each frequency band was obtained by wavelet packet decomposition.And,the single value neutrosophic sets(SVNS)matrix of known fault type and the SVNS of experimental data were obtained according to the energy spectrum of samples.Secondly,based on the SVNS theory,three kinds of classification calculation methods were established including Euclidean distance,cosine-similarity and the single value neutrosophic sets symmetric cross entropy(SVNSSCE).Finally,the experimental proved the feasibility and effectiveness of the method.The fault recognition rate was over 96%,and the classification methods of cosine similarity and SVNSSCE were proved better.The classification methods had the characteristics of data-driven,less manual parameters and less calculation in the thesis.