令狐克进 杜茂华 王沛鑫

摘  要： 刀具状态监测技术是实现自动化、智能化生产的关键技术。该技术发展至今，仍然不能够真正应用于实际加工中，也未能很好的解决变工况加工过程中;精确识别刀具磨损状态的问题。因此，本文通过对车削刀具磨损状态监测技术的研究。建立了刀具磨损状态识别的极限学习机（Extreme learning machine，ELM）模型。首先，选择了振动信号、AE信号作为监测信号;对采集的信号分别进行时域分析、振动信号进行小波分析、AE信号进行多分辨率分析并提取与刀具磨损相关性强的特征作为原始特征。其次，采用Relief-F算法对原始特征进行特征过滤得到最终特征样本。最后，将训练样本输入建立的ELM模型进行训练，并输入测试样本查看仿真结果。模型的正确识别率为96.296%，表明建立的ELM模型对车削刀具状态识别具有很好的分类效果。

关键词： 振动信号;AE信号;Relief-F算法;极限学习机

中图分类号： TP183;O235    文献标识码： A    DOI：10.3969/j.issn.1003-6970.2020.06.042

本文著录格式：令狐克进，杜茂华，王沛鑫，等. 基于极限学习机的车削刀具状态监测技术研究[J]. 软件，2020，41（06）：208213

【Abstract】： Tool condition monitoring technology is the key technology to realize automated and intelligent production. To date， the technology has not been able to be used in actual machining， and it has not been able to solve the problem of machining under variable working conditions; the problem of accurately identifying the tool wear status. Therefore， in this paper， we study the technology of monitoring the wear status of turning tools. An extreme learning machine （ELM） model for tool wear state recognition was established. First， vibration signals and AE signals were selected as the monitoring signals; time-domain analysis was performed on the collected signals， wavelet analysis was performed on the vibration signals， multi-resolution analysis was performed on the AE signals， and features with strong correlation with tool wear were extracted as original features. Secondly， the Relief-F algorithm is used to filter the original features to obtain the final feature samples. Finally， the training sample is input into the established ELM model for training， and the test sample is input to view the simulation results. The correct recognition rate of the model is 96.296%， which indicates that the established elm model has a good classification effect for turning tool state recognition.

【Key words】： Vibration signal; AE signal; Relief-F algorithm; Extreme learning machine

0  引言

刀具狀态监测技术作为先进制造技术的重要组成部分，尚未形成完整、成熟的理论体系，不能很好的解决各种变工况加工条件下刀具磨损状态识别模型的识别精度低的问题[1]。

在传统的机加工中，大多数零件是通过切削形成的。研究表明，使用刀具监视技术可以将自动化加工处理系统的生产率提高10%到60%，且停机时间减少75%。机床的利用率达到50%以上[2]。刀具状态监测技术已受到世界各国的广泛关注，其成功无疑带来了巨大的经济和社会价值。……