基于神经网络的刀具磨损状态研究  

Research on the State of Tools Weared Based on Artificial Neural Netuorks

作　　者：赵庆新[1] ZHAO Qingxin(Meizhouwan Vocational Technology College,Putian Fujian 351100,China)

机构地区：[1]湄洲湾职业技术学院,福建莆田351100

出　　处：《长春工程学院学报（自然科学版）》2023年第1期86-91,共6页Journal of Changchun Institute of Technology：Natural Sciences Edition

基　　金：福建省教育厅项目(JAT171120)。

摘　　要：为有效预测刀具磨损状态,避免因刀具磨损导致机床无法正常运行的问题,将数控机床作为研究对象,提出了一种基于振动信号和音频信号的刀具磨损状态预测方法。首先,介绍了BP神经网络的基本概况,为了达到更好的刀具磨损状态的预测效果,对BP神经网络进行改进;其次,以改进的BP神经网络为基础,采用时频域统计分析方法对振动信号和音频信号进行分析,以提取刀具的磨损特征值;再次,对特征数据进行融合,并根据融合后的特征数据构建刀具磨损状态模型;最后,通过对比实验,对比优化前后的模型预测结果,结果证明优化后的模型预测精度更好,且收敛速度更快。Aiming at the problem that tool wear state will affect the work of machine tools,this study takes NC machine tools as the research object,and proposes a tool wear state prediction method based on vibration signal and audio signal.Firstly,the basic situation of BP neural network is introduced.In order to achieve better prediction effect of tool wear state,BP neural network is improved.Secondly,based on the improved BP neural network,the vibration signal and audio signal are analyzed by time-frequency domain statistical analysis method to extract the characteristic value of tool wear.Then,the feature data are fused,and the tool wear state model is constructed according to the fused feature data.Finally,through comparative experiments,the prediction results of the model before and after optimization are compared.The results show that the prediction accuracy of the optimized model is better and the convergence speed is faster.

关 键 词：时频域统计 振动信号 音频信号 改进BP预测模型 遗传算法 

分 类 号：TG502[金属学及工艺—金属切削加工及机床]