基于多信号融合与深度学习的铣刀磨损量预测方法  

Prediction Method of Milling Tool Wear Based on Multi-signal Fusion and Deep Learning

作　　者：罗波[1] Luo Bo(Beijing Polytechnic College,Beijing 100042,China)

机构地区：[1]北京工业职业技术学院

出　　处：《工具技术》2022年第8期131-135,共5页Tool Engineering

基　　金：北京工业职业技术学院校科研项目(BGY2020KY-13Z)。

摘　　要：为了给铣刀的维修工作提供有效参考数据,解决现有磨损量预测方法精度低的问题,利用多信号融合与深度学习算法,实现铣刀刀具磨损量预测方法的优化设计。根据铣刀结构和摩擦工作机制,模拟其铣削加工过程。采集刀具在工作状态下的切削力、振动、电流以及声发射信号,采用自适应加权的方式实现多信号的融合处理。以信号融合结果为基础,利用深度学习算法提取信号特征,得到降维特征处理结果。确定铣刀的信号特征与磨损量之间的关系,通过刀具信号的估计,得出铣刀磨损量的预测结果。通过与现有预测方法的对比试验发现,采用该设计预测方法得出的磨损量与实际测定值之间的相对误差更小,设计预测方法的精度更高。In order to provide effective reference data for the maintenance of milling cutter and solve the problem of low accuracy of existing wear prediction methods,the optimal design of milling cutter wear prediction method is realized by using multi-signal fusion and deep learning algorithm.According to the structure and friction working mechanism of milling cutter,the milling process is simulated.The cutting force,vibration,current and acoustic emission signals of the tool in the working state are collected,and the adaptive weighting method is used to realize the fusion processing of multiple signals.Based on the signal fusion results,the signal features are extracted by deep learning algorithm,and the dimension reduction feature processing results are obtained.The relationship between milling cutter tool signal characteristics and wear is determined.Through the estimation of tool signal,the prediction result of milling cutter tool wear is obtained.Through the comparison test with the existing prediction methods,it is found that the relative error between the wear amount obtained by this design prediction method and the actual measured value is smaller,and the accuracy of the design prediction method is higher.

关 键 词：多信号融合 深度学习 铣刀刀具磨损 磨损量预测 

分 类 号：TG714[金属学及工艺—刀具与模具] TH161[机械工程—机械制造及自动化] TP391[自动化与计算机技术—计算机应用技术]