基于判别字典学习的刀具磨损状态分类  

Tool Wear Condition Classification Based on Discriminant Dictionary Learning

作　　者：罗长源 袁德志 李申申 朱锟鹏 LUO Changyuan;YUAN Dezhi;LI Shenshen;ZHU Kunpeng(School of Mechanical Automation,Wuhan University of Science and Technology,Wuhan Hubei 430081,China;Institute of Intelligent Machines of Hefei Institutes of Physical Science,Chinese Academy of Sciences,Changzhou Jiangsu 213164,China)

机构地区：[1]武汉科技大学机械工程学院,湖北武汉430081 [2]中国科学院合肥物质科学研究院智能机械研究所,江苏常州213164

出　　处：《机床与液压》2025年第4期47-53,共7页Machine Tool & Hydraulics

基　　金：国家自然科学基金面上项目(52175528)。

摘　　要：传统的数据驱动刀具磨损状态分类依赖于特征的提取与选择,极大影响分类性能。为了实现自动提取刀具磨损状态对应的特征,提出一种基于判别字典学习的刀具磨损状态分类模型(DDLC),此模型联合用于稀疏表示的判别字典和用于模式识别的线性分类器,模型的结构简单、复杂度低、准确率高。在训练阶段,为了增强字典学习的可判别性,在字典学习过程中引入判别稀疏编码误差、重构误差和分类误差,建立了统一的字典学习优化目标。同时将多方向力进行数据级融合作为模型的输入信号。与其他经典的刀具磨损状态监测模型进行比较,所提模型的准确率和F1分数分别为98.46%和97.62%,证明了DDLC方法在刀具磨损状态分类方面的有效性和优越性,其检测精度满足实际加工需求,为刀具磨损状态监测提供了一种新方法。The traditional data-driven classification of tool wear states relies on the extraction and selection of features,which greatly affects the classification performance.In order to automatically extract the corresponding features of tool wear state,a novel tool wear state classification model based on discriminative dictionary learning classification(DDLC)was proposed,combining the discriminant dictionary for sparse representation and the linear classifier for pattern recognition,with simple structure,low complexity and high accuracy.A unified dictionary learning optimization goal was established.In the training stage,in order to enhance the discriminability of dictionary learning,the discriminant sparse coding error,reconstruction error and classification error were introduced in the dictionary learning process.At the same time,the data-level fusion of multidirectional forces was used as the input signal of the model.Compared with other classical tool wear state monitoring models,the accuracy and F1 score of the proposed model are 98.46%and 97.62%,respectively,which proves the effectiveness and superiority of the DDLC method in the classification of tool wear condition.Its detection accuracy meets the actual processing requirements and provides a new method for the monitoring of tool wear state.

关 键 词：稀疏表示 判别字典学习 数据融合 刀具磨损状态监测 

分 类 号：TP23[自动化与计算机技术—检测技术与自动化装置]