基于空间坐标与振动特征融合的机床切削状态分类方法  

Machine tool cutting state classification method based on fusion of spatial coordinate and vibration features

作　　者：王晶[1,2] 程晓斌 高艳[1] 王勋 杨军 WANG Jing;CHENG Xiaobin;GAO Yan;WANG Xun;YANG Jun(CAS Key Lab of Noise and Vibration,Institute of Acoustics,Chinese Academy of Sciences(CAS),Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院噪声与振动重点实验室(声学研究所),北京100190 [2]中国科学院大学,北京100049

出　　处：《振动与冲击》2022年第23期249-256,306,共9页Journal of Vibration and Shock

基　　金：中国科学院先导专项项目(XDC02020400)。

摘　　要：t分布的随机邻域嵌入(t-distributed stochastic neighbor embedding, t-SNE)常被用作机床切削状态分类中的特征选择方法,以学习切削参数之间的潜在关系。为了提高切削状态分类的精度,融合振动信号特征与切削激励点的空间坐标,提出了空间坐标嵌入的t分布的随机邻域嵌入方法(spatial coordinate embedded t-SNE, Ct-SNE)。该方法采用振动信号构建高维特征空间,将空间坐标作为物理信息嵌入至特征空间,以优选出类内相似度高、类间差异性大的特征。试验采集了三轴立式铣床加工的数据,对比了传统t-SNE方法与Ct-SNE方法的可视化结果和切削状态分类的准确性。结果表明,与传统方法相比,切削激励点的空间坐标的引入可以提高振动特征的可区分度,显著提升切削状态分类的准确率。Thet-distributed stochastic neighbor embedding(t-SNE) is often used as a feature selection method in machine tool cutting state classification to learn potential relations among cutting parameters.Here, to improve the accuracy of cutting state classification, a spatial coordinate embedded t-SNE(Ct-SNE) method was proposed to fuse vibration signal features and spatial coordinates of cutting excitation points.In this method, vibration signals were used to construct a high-dimensional feature space, and spatial coordinates were embedded into the feature space as physical information to select features with high in-class similarity and large inter-class difference.Processing data of a three-axis vertical milling machine in testswere collected, and visualization results and accuracies of cutting state classification of traditional t-SNE methodand Ct-SNE method were compared. The results showed that compared with the traditional method, introducing spatial coordinates of cutting excitation points can improve the distinguishability of vibration features, and significantly improve the accuracy of cutting state classification.

关 键 词：状态监测 t分布的随机邻域嵌入 特征选择 振动监测 空间坐标 

分 类 号：TH1[机械工程]