基于声振融合的二次EWT-CNN刀具磨损监测
Secondary EWT-CNN Tool Wear Monitoring Based on Vibroacoustic Fusion
郝旺身 1娄永威 1董辛旻 1李继康 1娄本池1
作者信息
- 1. 郑州大学机械与动力工程学院,郑州 450001
- 折叠
摘要
为了实现加工过程中对刀具磨损状态的监测,提出一种基于协同过滤融合的方法.首先,对工作刀具振动信号和声音信号进行特征相关性分析后进行数据层融合;然后,将得到的声振融合信号进行二次经验小波变换(EWT)后去噪重构;最后,将重构信号进行信号增强并送入CNN实现特征提取及刀具故障识别.通过对不同故障类型的麻花钻头进行故障识别实验,在声音、振动以及声振融合信号和不同信号去噪重构方法的对比下,该方法对不同故障类型的钻头作出了98.96%的高识别率.验证了所提方法在刀具故障识别方面的优越性.
Abstract
In order to realize the fault monitoring of the working state of the tool during the machining process,a method based on collaborative filtering and fusion is proposed to analyze the feature correlation between the vibration signal and the sound signal of the working tool and then fuse the data layer,and the obtained acoustic-vibroacoustic fusion signal is denoising and reconstructed after the second empirical wavelet transform(EWT),and finally the reconstructed signal is enhanced and sent to CNN to realize fea-ture extraction and tool fault identification.Through the fault identification experiment of twist drill bits of different fault types,the method has a high recognition rate of 98.96%for drill bits of different fault types under the comparison of sound,vibration,vibroacoustic fusion signals and different signal denoising recon-struction methods.The superiority of the proposed method in tool fault detection is verified.
关键词
声振融合信号/刀具磨损/故障识别/经验小波变换/卷积神经网络Key words
vibroacoustic fusion signal/tool wear/fault identification/empirical wavelet transform/convo-lutional neural networks引用本文复制引用
基金项目
国家重点研发计划项目(2016YFF0203104-5)
河南省科技攻关资助项目(202102210075)
出版年
2024
NETL
NSTL
维普
万方数据