Page 252 - 《应用声学》2025年第2期

512 2025 年 3 月

[3] Brinksmeier E, Heinzel C, Meyer L. Development and ap-
4 结论 plication of a wheel based process monitoring system in
grinding[J]. CIRP Annals, 2005, 54(1): 301–304.
本文通过分析插齿刀磨削过程中不同砂轮磨 [4] Yoshida T, Karasawa H, Fukui R, et al. Analysis of chip
size distribution using image processing technology to esti-
损状态下的 AE 信号，基于 AE 和 SVM 建立了可实
mate wear state of cylindrical grinding wheel[J]. Tribology
时监测砂轮磨损监测平台。首先分析了插刀磨磨削 International, 2021, 153: 106600.
原理，这种非连续磨削行为对实时监测造成不便，并 [5] Nikiforov I, Maltsev P, Kulakova M. Grinding wheel mon-
且该磨削方式和传感器的安装方式会使磨削信号 itoring system[C]. Environment. Technology. Resources.
Proceedings of the International Scientific and Practical
产生一定的噪声。为了提取有效磨削信号，采用滤
Conference, 2019.
波的方法对其展成分量及其他因素产生的噪声进 [6] Ahmer M, Marklund P, Gustafsson M, et al. Integra-
行过滤，最终得到了接近实际工况下的磨削 AE 信 tion of process monitoring and machine condition diagnos-
tics to improve quality prediction in grinding[J]. Procedia
号。对实验过程磨削信号进行时域 RMS 能量分析，
CIRP, 2021, 101: 170–173.
根据理论砂轮钝化能量曲线划分了插刀磨砂轮钝 [7] Wang S, Zhao Q L, Wu T. An investigation of monitor-
化状态节点，将其分类标签，为多分类模型做好数据 ing the damage mechanism in ultra-precision grinding of
准备。采用小波包变换提取磨削信号频域特征，并 monocrystalline silicon based on AE signals processing[J].
Journal of Manufacturing Processes, 2022, 81: 945–961.
优化各特征参数组合对模型性能的影响，最终提取 [8] 尹国强, 王东, 关云匀, 等. 基于声发射监测的砂轮磨损实
了9 维故障特征参数作为多分类 SVM 模型的输入， 验 [J]. 东北大学学报 (自然科学版), 2022, 43(8): 1127–1133.
砂轮磨损状态作为输出，实现了对非连续磨削产生 Yin Guoqiang, Wang Dong, Guan Yunyun, et al. Grind-
ing wheel wear experiment based on acoustic emis-
时变非稳定信号的实时监测。
sion monitoring[J]. Journal of Northeastern Univer-
最终，模型准确率可达 0.91，ROC 曲线性能指 sity(Natural Science), 2022, 43(8): 1127–1133.
标AUC高达0.96，同时该多分类模型性能优于朴素 [9] 钟利民, 李丽娟, 杨京, 等. HDP-HSMM 的磨削声发射砂轮
贝叶斯和 K最近邻其他两种多分类模型。在后续工 钝化状态识别 [J]. 应用声学, 2019, 38(2): 151–158.
Zhong Limin, Li Lujuan, Yang Jing, et al. The blunt state
作中，如何改进加工工艺和优化模型来提高识别准 identification of acoustic emission for grinding wheel based
确率、进一步提高磨削精度是研究重点。 on HDP-HSMM[J]. Journal of Applied Acoustics, 2019,
38(2): 151–158.
[10] Bagga P J, Chavda B, Modi V, et al. Indirect tool wear
measurement and prediction using multi-sensor data fu-
参 考 文 献
sion and neural network during machining[J]. Materials
Today: Proceedings, 2022, 56(Pt1): 51–55.
[1] Yang Z S, Yan W, Jin L, et al. A novel feature represen- [11] Liu C Y, Meerten Y, Declercq K, et al. Vibration-based
tation method based on original waveforms for acoustic gear continuous generating grinding fault classification
emission signals[J]. Mechanical Systems and Signal Pro- and interpretation with deep convolutional neural net-
cessing, 2020, 135: 106365. work[J]. Journal of Manufacturing Processes, 2022, 79:
[2] Azarhoushang B, Kitzig-Frank H. Principles of grinding 688–704.
processes[M]//Azarhoushang B, Ioan Marinescu D, Rowe [12] 杨亚森. 插齿刀专用数控磨床数控代码自动编程软件开
W B, et al. Tribology and Fundamentals of Abrasive Ma- 发 [D]. 西安: 西安工业大学, 2022.
chining Processes. Third Edition. New York: William [13] 张泽. 插齿刀专用数控磨床模块化设计及动力学分析 [D]. 西
Andrew Publishing, 2022: 351–468. 安: 西安工业大学, 2022.