激光抛光过程声发射信号的降噪与分类预测方法研究  

Study on Signal Denoising and Classification Prediction for On-line Monitoring of Acoustic Emission in Laser Polishing Process

作　　者：王德辉 刘俨后 郭繁明 麻娟 刘建 韩金国 Wang Dehui;Liu Yanhou;Guo Fanming;Ma Juan;Liu Jian;Han Jinguo(School of Mechanical Engineering,Shandong University of Technology,Zibo,Shandong 255000,China;不详)

机构地区：[1]山东理工大学机械工程学院,山东省淄博市255000 [2]山东省精密制造与特种加工重点实验室

出　　处：《工具技术》2025年第1期125-131,共7页Tool Engineering

基　　金：国家自然科学基金(51905322);山东省科技型中小企业创新能力提升工程(2023TSGC0971)。

摘　　要：激光抛光(LP)是一种激光束作用于材料表面使其快速熔凝的新型表面处理技术,抛光过程中激光作用时间短(10^(-4)~10^(-2)s)、材料加工层小(10^(-4)~10^(-2)m),激光抛光过程状态的监测十分困难。本文采用声发射传感器对激光抛光过程进行了状态信号的采集和分析,用三种降噪方法对声发射信号进行降噪处理,并用信噪比与均方根差值对降噪效果进行评价,经对比得出,小波硬阈值算法的降噪效果最好。建立了基于BP神经网络算法的激光抛光工艺参数分类预测模型,将时域信息中提取的峰值与持续时间作为时域特征值与小波包变换得到的节点能量作为频域特征值来共同组成特征向量,用于BP神经网络模型训练,并评价模型的性能。预测结果表明,BP神经网络模型对激光抛光工艺参数分类预测具有可行性,研究成果可用于优化激光抛光工艺参数,并为激光抛光在线监测提供方法基础。Laser polishing(LP)is a new type of surface treatment technology,which utilizes laser beam to rapidly melt the surface of materials.Due to the short laser action time(10^(-4)to 10^(-2)s)and small material processing layer(10^(-4)to 10^(-2)m),monitoring the status of laser polishing process is very difficult.To solve this problem,this paper adopts acoustic emission sensor to collect and analyzes the status signals of laser polishing process.Three noise reduction methods are used to process acoustic emission signals,and the noise reduction effect is evaluated by signal-to-noise ratio and root mean square difference.The results show that the wavelet hard threshold algorithm has the best noise reduction effect.A model based on BP neural network algorithm is established to predict the classification of laser polishing process parameters.The peak and duration extracted from the time domain information are used as the time domain characteristic values,and the node energy obtained from wavelet packet transform is used as the frequency domain characteristic value to form the feature vector for BP neural network model training and performance evaluation.The prediction results show that BP neural network model is feasible in predicting the classification of laser polishing process parameters.This research achievement can be applied to optimize the parameters of laser polishing process,and provide a method basis for online monitoring.

关 键 词：激光抛光 声发射 降噪处理 神经网络 

分 类 号：TG665[金属学及工艺—金属切削加工及机床] TG115