基于自步学习的刀具加工过程监测数据异常检测方法

doi:10.16183/j.cnki.jsjtu.2022.234

上海交通大学学报 ›› 2023, Vol. 57 ›› Issue (10): 1346-1354.doi: 10.16183/j.cnki.jsjtu.2022.234

所属专题：《上海交通大学学报》2023年“机械与动力工程”专题

基于自步学习的刀具加工过程监测数据异常检测方法

张建^a, 胡小锋^a(), 张亚辉^b

a.上海交通大学机械与动力工程学院, 上海 200240
b.上海交通大学海洋装备研究院, 上海 200240

收稿日期:2022-06-21 修回日期:2022-07-22 接受日期:2022-09-08 出版日期:2023-10-28 发布日期:2023-10-31
通讯作者: 胡小锋 E-mail:wshxf@sjtu.edu.cn
作者简介:张建(1997-),硕士生,现主要从事刀具磨损预测研究.
基金资助:
国防基础科研计划项目(JCKY2021110B048);国家重点研发计划资助项目(2018YFB1700502)

Abnormal Detection Method of Tool Machining Monitoring Data Based on Self-Paced Learning

ZHANG Jian^a, HU Xiaofeng^a(), ZHANG Yahui^b

a. School of Mechanical Engineering, Shanghai 200240, China
b. Institute of Marine Equipment, Shanghai Jiao Tong University, Shanghai 200240, China

Received:2022-06-21 Revised:2022-07-22 Accepted:2022-09-08 Online:2023-10-28 Published:2023-10-31
Contact: HU Xiaofeng E-mail:wshxf@sjtu.edu.cn

摘要/Abstract

摘要：

针对零件加工过程的监控数据异常导致刀具剩余寿命预测准确性下降的问题,提出一种基于自步学习的数据异常检测方法.首先建立多层感知机模型关联刀具加工过程监测数据和所对应的刀具剩余寿命;其次在模型权重更新过程中,先固定模型权重参数,预测损失拟合高斯分布得到异常样本的损失阈值,然后构建基于自步学习方法的损失函数,迭代更新模型参数.在模型训练结束后,根据损失阈值划分出异常样本.最后利用汽轮机转子轮槽的实际加工监测数据进行验证,并与局部异常因子算法、基于密度的聚类算法、 K 均值算法、孤立森林算法、一分类支持向量机等方法进行对比分析,验证本方法的有效性.

关键词: 刀具加工监测, 数据质量, 异常检测, 自步学习

Abstract:

Aimed at the problem that the accuracy of tool remaining life prediction was reduced due to the abnormal monitoring data in machining process, a data anomaly detection method based on self-paced learning was proposed. First, a multi-layer perceptron model was established to correlate the tool processing monitoring data with the tool remaining life. Next, in the process of updating model weight, the model weight parameters were fixed first, and the loss threshold of abnormal samples was obtained by predicting loss fitting Gaussian distribution. Then, the loss function based on the self-paced learning method was constructed to update model parameters iteratively. At the end of the model training, abnormal samples were divided according to the loss threshold. Finally, the actual processing monitoring data of turbine rotor groove were used to verify the proposed method, and compare with the local anomaly factor algorithm, the density-based clustering algorithm, the K-means algorithm, the isolated forest algorithm, and the one-class support vector machines.

Key words: tool machining monitoring, data quality, anomaly detection, self-paced learning

中图分类号:

TH166

张建, 胡小锋, 张亚辉. 基于自步学习的刀具加工过程监测数据异常检测方法[J]. 上海交通大学学报, 2023, 57(10): 1346-1354.

ZHANG Jian, HU Xiaofeng, ZHANG Yahui. Abnormal Detection Method of Tool Machining Monitoring Data Based on Self-Paced Learning[J]. Journal of Shanghai Jiao Tong University, 2023, 57(10): 1346-1354.

图/表 9

图1

图2

表1

表2

图3

表3

图4

表4

图5

参考文献 20

[1]	孙宇. 电力设备监测数据处理和数据库设计[D]. 浙江: 浙江大学, 2022.
	SUN Yu. Data processing and database design of power equipment monitoring[D]. Zhejiang: Zhejiang University, 2022.
[2]	尚文利, 石贺, 赵剑明, 等. 基于SAE-LSTM的工艺数据异常检测方法[J]. 电子学报, 2021, 49(8): 1561-1568. doi: 10.12263/DZXB.20180015
	SHANG Wenli, SHI He, ZHAO Jianming, et al. An anomaly detection method of process data based on SAE-LSTM[J]. Acta Electronica Sinica, 2021, 49(8): 1561-1568. doi: 10.12263/DZXB.20180015
[3]	夏英, 韩星雨. 融合统计方法和双向卷积LSTM的多维时序数据异常检测[J]. 计算机应用研究, 2022, 39(5): 1362-1367.
	XIA Ying, HAN Xingyu. Multi-dimensional time series data anomaly detection fusing statistical methods and bidirectional convolutional LSTM[J]. Application Research of Computers, 2022, 39(5): 1362-1367.
[4]	孙滢涛, 张锋明, 陈水标, 等. 基于多域特征提取的电力数据异常检测方法[J]. 电力系统及其自动化学报, 2022, 34(6): 105-113.
	SUN Yingtao, ZHANG Fengming, CHEN Shuibiao, et al. Power data anomaly detection algorithm based on multi-domain feature extraction[J]. Proceedings of the CSU-EPSA, 2022, 34(6): 105-113.
[5]	傅世元, 高欣, 张浩, 等. 基于元学习动态选择集成的电力调度数据异常检测方法[J]. 电网技术, 2022, 46(8): 3248-3261.
	FU Shiyuan, GAO Xin, ZHANG Hao, et al. Anomaly detection for power dispatching data based on meta-learning dynamic ensemble selection[J]. Power System Technology, 2022, 46(8): 3248-3261.
[6]	刘鑫. 无监督异常检测方法研究及其应用[D]. 成都: 电子科技大学, 2018.
	LIU Xin. Research on unsupervised anomaly detection algorithm and application[D]. Chengdu: University of Electronic Science and Technology of China, 2018.
[7]	DU H, ZHAO S, ZHANG D, et al. Novel clustering-based approach for local outlier detection[C]//International Conference on Computer Communications Workshops. San Francisco, CA, USA: IEEE, 2016: 802-811.
[8]	吴蕊, 张安勤, 田秀霞, 等. 基于改进K-means的电力数据异常检测算法[J]. 华东师范大学学报(自然科学版), 2020(4): 79-87.
	WU Rui, ZHANG Anqin, TIAN Xiuxia, et al. Anomaly detection algorithm based on improved K-means for electric power data[J]. Journal of East China Normal University (Natural Science), 2020(4): 79-87.
[9]	吴金娥, 王若愚, 段倩倩, 等. 基于反向k近邻过滤异常的群数据异常检测[J]. 上海交通大学学报, 2021, 55(5): 598-606.
	WU Jin’e, WANG Ruoyu, DUAN Qianqian, et al. Collective data anomaly detection based on reverse k-nearest neighbor filtering[J]. Journal of Shanghai Jiao Tong University, 2021, 55(5): 598-606.
[10]	陈砚桥, 孙彤, 张侨禹. 基于DBSCAN的智能机舱多源数据异常检测方法[J]. 舰船科学技术, 2021, 43(17): 156-160.
	CHEN Yanqiao, SUN Tong, ZHANG Qiaoyu. Intelligent engine room multi-source data detecting method based on DBSCAN cluster algorithm[J]. Ship Science and Technology, 2021, 43(17): 156-160.
[11]	宋丽娜, 刘淼, 秦韬, 等. 基于LOF与CEEMD的城镇取用水监测数据异常值识别[J]. 水利信息化, 2022(2): 33-40.
	SONG Lina, LIU Miao, QIN Tao, et al. Outlier identification of urban water intake monitoring data based on LOF and CEEMD[J]. Water Resources Informatization, 2022(2): 33-40.
[12]	王锋, 高欣, 贾欣, 等. 一种基于对数区间隔离森林的电力调度数据异常检测集成算法[J]. 电网技术, 2021, 45(12): 4818-4827.
	WANG Feng, GAO Xin, JIA Xin, et al. An anomaly detection ensemble algorithm for power dispatching data based on log-interval isolation[J]. Power System Technology, 2021, 45(12): 4818-4827.
[13]	王燕晋, 易忠林, 郑思达, 等. 基于孤立森林算法的电力用户数据异常快速识别研究[J]. 电子设计工程, 2022, 30(3): 11-14.
	WANG Yanjin, YI Zhonglin, ZHENG Sida, et al. Research on fast identification of power user data abnormal based on isolation forest algorithm[J]. Electronic Design Engineering, 2022, 30(3): 11-14.
[14]	卓琳, 赵厚宇, 詹思延. 异常检测方法及其应用综述[J]. 计算机应用研究, 2020, 37(Sup.1): 9-15.
	ZHUO Lin, ZHAO Houyu, ZHAN Siyan. Anomaly detection and its application[J]. Application Research of Computers, 2020, 37(Sup.1): 9-15.
[15]	BENGIO Y, LOURADOUR J, COLLOBERT R, et al. Curriculum learning[C]//Proceedings of the 26th Annual International Conference on Machine Learning. New York, NY, USA: Association for Computing Machinery, 2009: 41-48.
[16]	KUMAR M P, PACKER B, KOLLER D. Self-paced learning for latent variable models[C]//Proceedings of the 23rd International Conference on Neural Information Processing Systems-Volume 1. Red Hook, NY, USA: Curran Associates Inc., 2010: 1189-1197.
[17]	王艺玮, 邓蕾, 郑联语, 等. 基于多通道融合及贝叶斯理论的刀具剩余寿命预测方法[J]. 机械工程学报, 2021, 57(13): 214-224. doi: 10.3901/JME.2021.13.214
	WANG Yiwei, DENG Lei, ZHENG Lianyu, et al. A multi-channel signal fusion and Bayesian theory based method for tool remaining useful life prediction[J]. Journal of Mechanical Engineering, 2021, 57(13): 214-224. doi: 10.3901/JME.2021.13.214
[18]	吴浩, 卢楠, 邹进贵, 等. GNSS变形监测时间序列的改进型3σ粗差探测方法[J]. 武汉大学学报(信息科学版), 2019, 44(9): 1282-1288.
	WU Hao, LU Nan, ZOU Jingui, et al. An improved 3σ gross error detection method for GNSS deformation monitoring time series[J]. Geomatics and Information Science of Wuhan University, 2019, 44(9): 1282-1288.
[19]	徐洪钟, 吴中如, 李雪红, 等. 基于小波分析的大坝变形观测数据的趋势分量提取[J]. 武汉大学学报(工学版), 2003(6): 5-8.
	XU Hongzhong, WU Zhongru, LI Xuehong, et al. Abstracting trend component of dam observation data based on wavelet analysis[J]. Engineering Journal of Wuhan University, 2003(6): 5-8.
[20]	党英, 吉卫喜, 陆家辉, 等. 基于深度学习的铣刀剩余寿命预测方法研究[J]. 现代制造工程, 2021(12): 79-87.
	DANG Ying, JI Weixi, LU Jiahui, et al. Research on prediction method of remaining useful life of milling cutter based on deep learning[J]. Modern Manufacturing Engineering, 2021(12): 79-87.

铣刀编号	轮槽编号	铣刀编号	轮槽编号	铣刀编号	轮槽编号
1	1~13	6	65~75	11	119~128
2	14~26	7	76~86	12	129~139
3	27~38	8	87~96	13	140~149
4	39~50	9	97~107	14	150~159
5	51~64	10	108~118	15	160~170

实验编号	β=1.5		β=2.0		β=2.5		β=3.0		β=3.5
实验编号	MAE	RSME	MAE	RSME	MAE	RSME	MAE	RSME	MAE	RSME
1	1.386	1.608	1.138	1.334	1.169	1.417	1.013	1.223	1.315	1.572
2	1.125	1.303	1.286	1.533	1.214	1.391	1.122	1.365	1.424	1.630
3	1.170	1.410	1.333	1.513	1.431	1.738	1.122	1.365	1.424	1.630
4	1.170	1.410	1.214	1.391	1.431	1.738	1.122	1.365	1.459	1.674
5	1.232	1.388	1.305	1.679	1.214	1.391	0.965	1.202	1.424	1.630
平均值	1.216 6	1.423 8	1.255 2	1.490 0	1.291 8	1.535 0	1.068 9	1.304 0	1.409 2	1.627 2

实验编号	MAE	RMSE	异常样本数量/个	异常样本编号
1	1.013	1.223	3	10、43、107
2	1.122	1.365	3	9、10、11
3	1.122	1.365	3	9、10、11
4	1.122	1.365	3	9、10、11
5	0.965	1.202	3	10、11、12
平均值	1.069	1.304
标准差	0.067	0.075
优化比/%	26.284	28.194

方法	MAE	RMSE	MAE’/ %	RMSE’/ %
未筛选	1.45	1.816	—	—
本文方法	1.069±0.067	1.304±0.075	26.28	28.19
LOF	1.219	1.486	15.93	18.17
DBSCAN	1.693	2.137	-16.76	-17.68
孤立森林	1.263	1.527	12.90	15.91
K均值	1.844	2.215	-27.17	-21.97
One-Class SVM	1.28	1.605	11.72	11.62
随机采样	1.485±0.327	1.877±0.456	-2.41	-3.37

基于自步学习的刀具加工过程监测数据异常检测方法

Abnormal Detection Method of Tool Machining Monitoring Data Based on Self-Paced Learning

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 20

相关文章 4

编辑推荐

Metrics

本文评价

[1]	许勇, 蔡云泽, 宋林. 基于数据驱动的核电设备状态评估研究综述[J]. 上海交通大学学报, 2022, 56(3): 267-278.
[2]	武光利, 郭振洲, 李雷霆, 王成祥. 融合FCN和LSTM的视频异常事件检测[J]. 上海交通大学学报, 2021, 55(5): 607-614.
[3]	吴金娥, 王若愚, 段倩倩, 李国强, 琚长江. 基于反向k近邻过滤异常的群数据异常检测[J]. 上海交通大学学报, 2021, 55(5): 598-606.
[4]	郝圣桥，许黎明，沈伟，王建楼. 电液伺服阀状态在线特征提取和异常检测方法 [J]. 上海交通大学学报（自然版）, 2010, 44(12): 1747-1752.