基于时域-频域双路网络的深度学习电火花线切割放电状态识别  

Discharge State Recognition of Wire EDM Based on Time Domain and Frequency Domain Two-way Network

作　　者：王锋 李佳旺 张粤东 张永俊[1] 黄志刚[1] WANG Feng;LI Jiawang;ZHANG Yuedong;ZHANG Yongjun;HUANG Zhigang(Guangzhou Key Laboratory of Non-traditional Manufacturing Technology and Equipment,School of Mechanical and Electrical Engineering,Guangdong University of Technology,Guangzhou 510006,China)

机构地区：[1]广东工业大学机电工程学院,广州市非传统制造技术及装备重点实验室,广东广州510006

出　　处：《电加工与模具》2022年第5期23-29,40,共8页Electromachining & Mould

基　　金：国家自然科学基金项目(51275098)。

摘　　要：针对电火花线切割间隙放电波形的时序特性,采用长短期记忆网络结合小波变换提出了一种时域-频域双路记忆网络模型,用于识别电火花线切割间隙放电状态。该模型对原时域数据和经小波变换后的时频域数据分别进行了特征提取,而后输出类别置信度,再对双路输出融合进行分类训练。结果表明:经过十折交叉验证,双路输出融合模型的分类准确率达到99.69%,可较好地满足电火花线切割放电状态识别需求。Aiming at the timing characteristics of the discharge waveform of WEDM,a long short-term memory network combined with wavelet transform is used to propose a time domain-frequency domain dual memory network model,which is used for WEDM gap discharge state recognition.The model extracts features from the original time domain data and the time-frequency domain data transformed by wavelet,and then outputs the class confidence,and finally fuses the dual outputs for classification training.The results show that the classification accuracy rate reaches 99.69%after ten-fold cross-validation average,which can better meet the needs of wire cutting discharge status recognition.

关 键 词：电火花线切割 放电状态识别 小波变换 长短期记忆网络 

分 类 号：TG661[金属学及工艺—金属切削加工及机床]