结合GA-BP与集成学习的钻削过程刀具状态实时监测

马晶; 白峥言; 刘献礼; 刘强; 贾儒鸿; 周强

doi:10.13433/j.cnki.1003-8728.20220122

结合GA-BP与集成学习的钻削过程刀具状态实时监测

doi: 10.13433/j.cnki.1003-8728.20220122

马晶^{1, 3,},
白峥言¹,
刘献礼¹,
刘强^{1, 2},
贾儒鸿¹,
周强¹

1.
哈尔滨理工大学先进制造智能化技术教育部重点实验室, 哈尔滨 150080
2.
哈尔滨理工大学电气工程博士后流动站, 哈尔滨 150080
3.
哈尔滨理工大学仪器科学与技术博士后流动站, 哈尔滨 150080

基金项目:

国家重点研发计划项目 2019YFB1704800

国家自然科学基金项目 52005141

国家自然科学基金项目 51805122

2020年度黑龙江省普通本科高等学校青年创新人才培养计划项目 UNPYSCT-2020193

详细信息

作者简介:
马晶(1989-), 讲师, 硕士生导师, 博士, 研究方向为智能刀具, majing2211@163.com

中图分类号: TG713⁺.1;TH16
计量
- 文章访问数: 50
- HTML全文浏览量: 44
- PDF下载量: 15
- 被引次数: 0
出版历程
- 收稿日期: 2021-08-21
- 刊出日期: 2023-10-25

Real-time Monitoring of Tool States in Drilling Process Combined with GA-BP and Ensemble Learning

MA Jing^{1, 3
,},
BAI Zhengyan¹,
LIU Xianli¹,
LIU Qiang^{1, 2},
JIA Ruhong¹,
ZHOU Qiang¹

1.
The Ministry of Education, Key Laboratory of Advanced Manufacturing and Intelligent Technology, Harbin University of Science and Technology, Harbin 150080, China
2.
Postdoctoral Research Station of Electrical Engineering, Harbin University of Science and Technology, Harbin 150080, China
3.
Postdoctoral Mobile Station of Instrument Science and Technology, Harbin University of Science and Technology, Harbin 150080, China

摘要

摘要: 为了能够有效识别钻削过程中刀具的磨损状态，为工厂实际加工过程提供刀具磨损的及时预警，开发了一种基于LabVIEW的钻削刀具磨损状态监测平台。平台可以实现实时采集振动信号并进行时域、频域和时频域的特征提取和数据保存。通过将遗传优化算法、BP神经网络与集成学习结合，构建了GA-BP-Adaboost模型，借助LabVIEW与MATLAB混合编程实现了模型搭建。最后，经过钻削实验分析实时信号及其多种特征对钻头刀具磨损状态的的表征情况，选择三层小波包分解的1、6、8频带作为模型的输入数据训练模型，经测试，模型的分类精度在90%以上。同时，平台的实时响应时间不超过3 s，可以满足实际加工过程的要求。
- 钻削过程 /
- 状态监测 /
- LabVIEW /
- 集成学习 /
- GA-BP-Adaboost
Abstract: In order to effectively identify the tool wear states in the drilling process and provide timely warning of tool wear for the actual machining process in the factory, a condition monitoring platform for tool wear state of drilling based on LabVIEW is developed. The platform can realize real-time vibration signal acquisition, feature extraction and data preservation in time domain, frequency domain and time-frequency domain. By combining genetic optimization algorithm (GA), Back propagation (BP) neural network and ensemble learning, the GA-BP-Adaboost model is constructed, and the model is built by LabVIEW and MATLAB mixed programming. Finally, through actual drilling experiments, the characterization of real-time signals and their various characteristics on the wear states of drill tools is analyzed, the 1, 6 and 8 frequency bands of three-layer wavelet packet decomposition are selected as the input data of the model to train the GA-BP-Adaboost model, the classification accuracy of the model is above 90%. At the same time, the real-time response time of the platform is not more than 3 seconds, which can meet the requirements of the actual machining process.
- drilling process /
- condition monitoring /
- LabVIEW /
- ensemble learning /
- GA-BP-Adaboost

HTML全文

图 1 钻头磨损程度

Figure 1. Wear degree of drill bit

下载: 全尺寸图片幻灯片

图 2 平台整体框架

Figure 2. Framework of the platform

下载: 全尺寸图片幻灯片

图 3 小波包分解结构

Figure 3. Wavelet packet decomposition structure

下载: 全尺寸图片幻灯片

图 4 平台前面板可视化效果

Figure 4. Visual effect of the front panel of the platform

下载: 全尺寸图片幻灯片

图 5 BP神经网络结构

Figure 5. Structure of BP neural network

下载: 全尺寸图片幻灯片

图 6 Adaboost算法流程

Figure 6. Adaboost algorithm procedures

下载: 全尺寸图片幻灯片

图 7 监测算法流程图

Figure 7. Flow chart of the monitoring algorithm

下载: 全尺寸图片幻灯片

图 8 钻削实验装置

Figure 8. Drilling experimental device

下载: 全尺寸图片幻灯片

图 9 刀具磨损情况

Figure 9. Tool wear condition

下载: 全尺寸图片幻灯片

图 10 钻头磨损曲线

Figure 10. Wear curve of drill bit

下载: 全尺寸图片幻灯片

图 11 原始振动信号

Figure 11. Original vibration signal holes

下载: 全尺寸图片幻灯片

图 12 振动信号时域特征

Figure 12. Time domain characteristics of vibration signals

下载: 全尺寸图片幻灯片

图 13 小波包分解能量图

Figure 13. Energy graph of wavelet packet decomposition

下载: 全尺寸图片幻灯片

图 14 模型分类结果

Figure 14. Classification results of the model

下载: 全尺寸图片幻灯片

图 15 BP神经网络分类结果

Figure 15. BP neural network classification results

下载: 全尺寸图片幻灯片

图 16 GA-BP分类结果

Figure 16. GA-BP classification results

下载: 全尺寸图片幻灯片

图 17 实时监测界面

Figure 17. Real time monitoring interface

下载: 全尺寸图片幻灯片

表 1 钻头的主要磨损形式

Table 1. Wear form of a drill bit

磨损形式	具体类型	产生原因
	前刀面磨损	切削温度过高, 运动中的钻头将与其粘接的工件材料带走, 从而产生“月牙洼”
正常磨损	后刀面磨损	后刀面与工件持续挤压与摩擦, 产生较大的压力, 引起后刀面磨损
	横刃磨损	横刃主要承担定心的作用, 受到的轴向力最大, 且回转半径较小, 切削时挤压严重, 容易造成磨损
	崩刃	切削刃上应力较集中且受机械振动及切屑的影响, 容易产生崩刃
非正常磨损	剥落	切削不稳定或钻头承受交变接触应力, 易产生表层剥落
	折断	钻头材料有缺陷, 在恶劣条件下加工, 可能造成钻头折断

下载: 导出CSV

表 2 预实验切削参数

Table 2. Pre experimental cutting parameters

编号	主轴转速/(r·min^-1)	进给量/(mm·min^-1)	钻孔深度/mm	机床负载/%
1	270	15	12	30~40
2	270	10	12	20~30
3	270	15	18	30~50
4	300	10	12	25~35
5	300	15	12	30~60

下载: 导出CSV

表 3 标签划分标准

Table 3. Label classification standards

刀具磨损值VB	刀具磨损状态	标签
VB≤0.2 mm	初期磨损	[1, 0, 0]
0.2 mm≤VB≤0.3 mm	正常磨损	[0, 1, 0]
VB≥0.3 mm	急剧磨损	[0, 0, 1]

下载: 导出CSV

表 4 弱分类器累计误差和分类权重结果

Table 4. Cumulative error and classification weight results of the weak classifier

序号	累计误差e_n	分类权重a_n
1	0.350 0	0.309 5
2	0.304 4	0.413 3
3	0.307 9	0.405 1
4	0.321 8	0.372 8
5	0.332 4	0.348 6
6	0.312 0	0.395 4
7	0.308 7	0.403 2
8	0.309 1	0.402 2
9	0.302 1	0.418 7
10	0.344 4	0.322 0

下载: 导出CSV

表 5 模型测试结果

Table 5. Model test results

加工孔数	磨损状态	期望输出	实际输出
2	初期磨损	[1, 0, 0]	[0.872 3, 0.008 9, 0.045 3]
7	初期磨损	[1, 0, 0]	[0.694 5, 0.397 7, 0.001 5]
15	正常磨损	[0, 1, 0]	[0.006 2, 0.816 0, 0.117 8]
24	正常磨损	[0, 1, 0]	[-0.115 7, 1.215 9, 0.390 1]
35	急剧磨损	[0, 0, 1]	[0.005 4, 0.334 2, 0.641 9]
38	急剧磨损	[0, 0, 1]	[0.140 5, -0.064 2, 1.168 0]

下载: 导出CSV

参考文献(15)

[1]	徐彦伟, 陈立海, 袁子皓, 等. 基于信息融合的刀具磨损状态智能识别[J]. 振动与冲击, 2017, 36(21): 257-264. https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCJ201721037.htm XU Y W, CHEN L H, YUAN Z H, et al. Intelligent recognition of tool wear conditions based on the information fusion[J]. Journal of Vibration and Shock, 2017, 36(21): 257-264. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCJ201721037.htm
[2]	郝碧君. 钻削刀具状态监测技术的研究[D]. 南京: 南京航空航天大学, 2019. HAO B J. Research on tool condition monitoring technology in drilling[D]. Nanjing: Nanjing University of Aeronautics and Astronautics, 2019. (in Chinese)
[3]	赵京鹤, 刘宏岩, 胡晶. 基于BP神经网络的超声振动钻削钻头磨损状态监测试验研究[J]. 机械设计与研究, 2020, 36(2): 83-86. https://www.cnki.com.cn/Article/CJFDTOTAL-JSYY202002023.htm ZHAO J H, LIU H Y, HU J. Experimental study on wear monitoring of ultrasound vibration drilling bit based on BP neural network[J]. Machine Design & Research, 2020, 36(2): 83-86. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSYY202002023.htm
[4]	穆殿方, 刘献礼, 岳彩旭, 等. 深度学习与多信号融合在铣刀磨损状态识别中的研究[J]. 机械科学与技术, 2021, 40(10): 1581-1589. doi: 10.13433/j.cnki.1003-8728.20200209 MU D F, LIU X L, YUE C X, et al. Study on wear state recognition of milling cutter via deep learning and multi-signal fusion[J]. Mechanical Science and Technology for Aerospace Engineering, 2021, 40(10): 1581-1589. (in Chinese) doi: 10.13433/j.cnki.1003-8728.20200209
[5]	张学忱, 刘红岩, 史尧臣, 等. 基于声发射信号的振动钻削刀具磨损状态监测试验研究[J]. 机床与液压, 2020, 48(13): 189-192. https://www.cnki.com.cn/Article/CJFDTOTAL-JCYY202013045.htm ZHANG X C, LIU H Y, SHI Y C, et al. Experimental study on vibration drilling tool wear monitoring based on AE signal[J]. Machine Tool & Hydraulics, 2020, 48(13): 189-192. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JCYY202013045.htm
[6]	WANG G F, YANG Y W, XIE Q L, et al. Force based tool wear monitoring system for milling process based on relevance vector machine[J]. Advances in Engineering Software, 2014, 71: 46-51. doi: 10.1016/j.advengsoft.2014.02.002
[7]	YU J S, LIANG S, TANG D Y, et al. A weighted hidden Markov model approach for continuous-state tool wear monitoring and tool life prediction[J]. The International Journal of Advanced Manufacturing Technology, 2017, 91(1-4): 201-211. doi: 10.1007/s00170-016-9711-0
[8]	郝碧君, 陈妮, 李亮, 等. 基于小波包分析和LS-SVM的钻削刀具状态识别研究[J]. 工具技术, 2019, 53(12): 3-9. https://www.cnki.com.cn/Article/CJFDTOTAL-GJJS201912001.htm HAO B J, CHEN N, LI L, et al. Tool wear condition recognition in drilling based on wavelet packet analysis and LS-SVM[J]. Tool Engineering, 2019, 53(12): 3-9. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-GJJS201912001.htm
[9]	樊红卫, 杨一晴, 马宏伟, 等. 一种转子振动故障诊断及预警的虚拟仪器系统开发与验证[J]. 机械设计与制造, 2019(6): 77-79. https://www.cnki.com.cn/Article/CJFDTOTAL-JSYZ201906020.htm FAN H W, YANG Y Q, MA H W, et al. Development and validation of virtual instrument system for vibration fault diagnosis and warning of rotor[J]. Machinery Design & Manufacture, 2019(6): 77-79. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSYZ201906020.htm
[10]	杨一晴, 马宏伟, 樊红卫, 等. 煤矿旋转机械在线故障诊断及预警系统设计[J]. 工矿自动化, 2019, 45(10): 104-108. https://www.cnki.com.cn/Article/CJFDTOTAL-MKZD201910020.htm YANG Y Q, MA H W, FAN H W, et al. Design of online fault diagnosis and early warning system for coal mine rotating machinery[J]. Industry and Mine Automation, 2019, 45(10): 104-108. (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-MKZD201910020.htm
[11]	刘子安, 刘建春, 苏进发, 等. 刀具磨损感知数据驱动下的DBN预测模型研究[J]. 机械科学与技术, 2021, 40(7): 1043-1050. doi: 10.13433/j.cnki.1003-8728.20200178 LIU Z A, LIU J C, SU J F, et al. Study on DBN prediction model driven by tool wear sensing data[J]. Mechanical Science and Technology for Aerospace Engineering, 2021, 40(7): 1043-1050. (in Chinese) doi: 10.13433/j.cnki.1003-8728.20200178
[12]	刘红岩. 多传感器融合振动钻削磨损状态监测研究[D]. 长春: 长春理工大学, 2020. LIU H Y. Research on wear condition monitoring of multi-sensor fusion vibration drilling[D]. Changchun: Changchun University of Science and Technology, 2020. (in Chinese)
[13]	陈洪涛. 基于多参量信息融合的刀具磨损状态识别及预测技术研究[D]. 成都: 西南交通大学, 2013. CHEN H T. Study on tool wear monitoring and prediction technology based on multi-parameter information fusion[D]. Chengdu: Southwest Jiaotong University, 2013. (in Chinese)
[14]	宗姝. 振动钻削钻头状态监测技术研究[D]. 长春: 长春理工大学, 2019. ZONG S. Research on drill bit condition monitoring technology of vibration drilling[D]. Changchun: Changchun University of Science and Technology, 2019. (in Chinese)
[15]	ZHAO D B, WANG K S, KRIMMEL O. Intelligent tool state detection system based on neural network[J]. Transactions of Nanjing University of Aeronautics & Astronau, 2000(2): 169-175.