基于神经网络多传感器融合的刀具摩损定量监测的研究  

Tool Wear Monitoring on the Basis of Multisensor Fusionby Neural Network

作　　者：梁建成[1] 李圣怡[1] 温熙森[1] 杨叔子[1] 

机构地区：[1]国防科技大学,华中理工大学

出　　处：《机械科学与技术》1995年第6期125-130,150,共7页Mechanical Science and Technology for Aerospace Engineering

基　　金：国家自然科学基金

摘　　要：研究了前馈神经网络（ＦＮＮ）的自构造型学习算法，提出了基于神经网络多传感器融合的一般结构及刀具磨损监测方法，讨论了多传感器的选择、多传感器信号的采集与预处理以及多传感器信号的特征选择与正规化处理，并就铣削过程的刀具磨损监测进行了实验研究，结果表明，所提出的方法可获得９３％的识别率。This paper studies structural learning algorithm of feed forward neural network(FNN)and presents a new methed of tool wear monitoring n the basis of multisensor fusion by neural network. The selections of muItisensor,the acquisition ,preprocess,features selection and the normalization of multisensor signals are discussed. The experiments of tool wear mon-itoring for the milling process are implemented.The results showed that the recognition rateof the method is up to 93%and it is effective on tool wear monitoring.

关 键 词：神经网络 多传感器融合 刀具磨损 监测 

分 类 号：TG710.7[金属学及工艺—刀具与模具] TP212[自动化与计算机技术—检测技术与自动化装置]