基于机器学习的汽车级产品离群点检测

2023 IEEE International Reliability Physics Symposium (IRPS) Pub Date : 2023-03-01 DOI:10.1109/IRPS48203.2023.10118207

Y. L. Yang, P. Tsao, C. W. Lin, Ross Lee, Olivia Ni, T. T. Chen, Y. Ting, C. Lai, Jason Yeh, Arnold Yang, Wayne Huang, Peng Chen, Charly Tsai, Ryan Yang, Y. S. Huang, B. Hsu, M. Z. Lee, T. Lee, Michael Huang, Coming Chen, L. Chu, H. Kao, N. S. Tsai

{"title":"基于机器学习的汽车级产品离群点检测","authors":"Y. L. Yang, P. Tsao, C. W. Lin, Ross Lee, Olivia Ni, T. T. Chen, Y. Ting, C. Lai, Jason Yeh, Arnold Yang, Wayne Huang, Peng Chen, Charly Tsai, Ryan Yang, Y. S. Huang, B. Hsu, M. Z. Lee, T. Lee, Michael Huang, Coming Chen, L. Chu, H. Kao, N. S. Tsai","doi":"10.1109/IRPS48203.2023.10118207","DOIUrl":null,"url":null,"abstract":"Near-zero defective parts per million (DPPM) and returned material authorization (RMA) from customers is the goal pursued by many companies. In this paper, a machine learning based method is proposed to detect outliers in the final test stage using the XGBoost algorithm and the Mahalanobis distance. We captured the weak integrated circuits (ICs) that passed the final test but failed in the system level test (SLT) or the verification of quality engineering aging (QEA). Compared to the random sampling, the experiments showed we could recognize 2x~3x weak IC ratio in the SLT and >10x in the QEA to achieve automotive grade DPPM.","PeriodicalId":159030,"journal":{"name":"2023 IEEE International Reliability Physics Symposium (IRPS)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-03-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Performing Machine Learning Based Outlier Detection for Automotive Grade Products\",\"authors\":\"Y. L. Yang, P. Tsao, C. W. Lin, Ross Lee, Olivia Ni, T. T. Chen, Y. Ting, C. Lai, Jason Yeh, Arnold Yang, Wayne Huang, Peng Chen, Charly Tsai, Ryan Yang, Y. S. Huang, B. Hsu, M. Z. Lee, T. Lee, Michael Huang, Coming Chen, L. Chu, H. Kao, N. S. Tsai\",\"doi\":\"10.1109/IRPS48203.2023.10118207\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Near-zero defective parts per million (DPPM) and returned material authorization (RMA) from customers is the goal pursued by many companies. In this paper, a machine learning based method is proposed to detect outliers in the final test stage using the XGBoost algorithm and the Mahalanobis distance. We captured the weak integrated circuits (ICs) that passed the final test but failed in the system level test (SLT) or the verification of quality engineering aging (QEA). Compared to the random sampling, the experiments showed we could recognize 2x~3x weak IC ratio in the SLT and >10x in the QEA to achieve automotive grade DPPM.\",\"PeriodicalId\":159030,\"journal\":{\"name\":\"2023 IEEE International Reliability Physics Symposium (IRPS)\",\"volume\":\"25 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-03-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2023 IEEE International Reliability Physics Symposium (IRPS)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/IRPS48203.2023.10118207\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2023 IEEE International Reliability Physics Symposium (IRPS)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/IRPS48203.2023.10118207","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}

引用次数: 0

摘要

接近零的次品率(DPPM)和客户的退货授权(RMA)是许多公司追求的目标。本文提出了一种基于机器学习的方法，利用XGBoost算法和马氏距离来检测最终测试阶段的异常值。我们捕获了通过最终测试但未通过系统级测试(SLT)或质量工程老化验证(QEA)的弱集成电路(ic)。与随机抽样相比，我们可以在SLT中识别出2x~3x的弱集成电路比，在QEA中识别出>10x的弱集成电路比，从而实现汽车级DPPM。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

查看原文

微信好友朋友圈 QQ好友复制链接

本刊更多论文

Performing Machine Learning Based Outlier Detection for Automotive Grade Products

Near-zero defective parts per million (DPPM) and returned material authorization (RMA) from customers is the goal pursued by many companies. In this paper, a machine learning based method is proposed to detect outliers in the final test stage using the XGBoost algorithm and the Mahalanobis distance. We captured the weak integrated circuits (ICs) that passed the final test but failed in the system level test (SLT) or the verification of quality engineering aging (QEA). Compared to the random sampling, the experiments showed we could recognize 2x~3x weak IC ratio in the SLT and >10x in the QEA to achieve automotive grade DPPM.

求助全文

通过发布文献求助，成功后即可免费获取论文全文。去求助

来源期刊

2023 IEEE International Reliability Physics Symposium (IRPS)

自引率

0.00%

发文量