基于随机森林法的铣刀磨损状态监测  

Wear Condition Monitoring of Milling Cutter Based on Random Forest Method

作　　者：张丹[1] 隋文涛[2] 李志永[2] 陈锦 ZHANG Dan;SUI Wen-tao;LI Zhi-yong;CHEN Jin(School of Electrical and Electronic Engineering,Shandong University of Technology,Shandong Zibo 255000,China;School of Mechanical Engineering,Shandong University of Technology,Shandong Zibo 255000,China)

机构地区：[1]山东理工大学电气与电子工程学院 [2]山东理工大学机械工程学院

出　　处：《机械设计与制造》2024年第9期12-15,共4页Machinery Design & Manufacture

基　　金：山东省重点研发计划(重大科技创新工程)项目(2018CXGC0602);山东省自然科学基金项目(ZR2021ME160)。

摘　　要：针对机床加工过程中铣削刀具状态监测困难,以及单传感器监测法存在信息不全的问题,提出了一种利用随机森林完成多传感器信息融合的刀具磨损状态监测研究方法。在不同的切削参数下,以声发射传感器和振动传感器为信号采集元件,多方位采集信号,并对信号进行时域、频域、小波包分析,提取其对刀具状态敏感的共计23个特征量,构建随机森林模型对铣刀状态进行监测。实验结果表明,该研究方法识别刀具准确率达90.0%,具有可行性。Aiming at the problem of tool condition monitoring is difficult in milling machine tool processing and incomplete infor⁃mation in single sensor monitoring method,this paper proposes a tool wear condition monitoring method which uses Random For⁃ests to complete multi-sensor information fusion.Under different cutting parameters,acoustic emission sensor and vibration sen⁃sor are used as signal acquisition elements to collect multi-directional signals,and the signals are analyzed in time domain,fre⁃quency domain and wavelet packet.A total of 23 characteristic quantities that are sensitive to the cutting tool state are extracted,and Random Forests is constructed to monitor the milling cutter state.The experimental results show that the accuracy of the method is 90.0%,which is feasible.

关 键 词：声发射 振动信号 刀具磨损状态 特征提取 随机森林 

分 类 号：TH16[机械工程—机械制造及自动化] TP301[自动化与计算机技术—计算机系统结构]