J. A. Torres, Ivan Kissiov, M. Essam, C. Hartig, Richard Gardner, Ken Jantzen, Stefan Schueler, M. Niehoff
{"title":"机器学习辅助新产品设置","authors":"J. A. Torres, Ivan Kissiov, M. Essam, C. Hartig, Richard Gardner, Ken Jantzen, Stefan Schueler, M. Niehoff","doi":"10.1109/ASMC49169.2020.9185215","DOIUrl":null,"url":null,"abstract":"In the past, concepts like critical area analysis, have been successfully implemented to predict random and systematic layout induced effects. This has enabled semiconductor companies to have an initial estimate as to how a fixed process will respond to a variety of different designs. However, as the number of individual products increases, along with a reduction in the total number of wafers per product, it becomes increasingly difficult to determine which process parameters will lead to the highest possible yield for each individual product. We have outlined a methodology using machine learning that combines process and design data to greatly reduce the time needed for setting up new products. We have shown that similar designs (based on our feature extraction) behave similarly in the fab, thus allowing us to construct models that can eventually be used to find the optimal process conditions for a given design. Due to the nature or process optimization, this methodology also explores the use of SHAPley additive explanations (SHAP) in order to “interface” with existing human and physical explanations of the observations, thus providing a mechanism to assess the quality and reliability of the numerically derived models.","PeriodicalId":6771,"journal":{"name":"2020 31st Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC)","volume":"26 1","pages":"1-5"},"PeriodicalIF":0.0000,"publicationDate":"2020-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Machine Learning Assisted New Product Setup\",\"authors\":\"J. A. Torres, Ivan Kissiov, M. Essam, C. Hartig, Richard Gardner, Ken Jantzen, Stefan Schueler, M. Niehoff\",\"doi\":\"10.1109/ASMC49169.2020.9185215\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"In the past, concepts like critical area analysis, have been successfully implemented to predict random and systematic layout induced effects. This has enabled semiconductor companies to have an initial estimate as to how a fixed process will respond to a variety of different designs. However, as the number of individual products increases, along with a reduction in the total number of wafers per product, it becomes increasingly difficult to determine which process parameters will lead to the highest possible yield for each individual product. We have outlined a methodology using machine learning that combines process and design data to greatly reduce the time needed for setting up new products. We have shown that similar designs (based on our feature extraction) behave similarly in the fab, thus allowing us to construct models that can eventually be used to find the optimal process conditions for a given design. Due to the nature or process optimization, this methodology also explores the use of SHAPley additive explanations (SHAP) in order to “interface” with existing human and physical explanations of the observations, thus providing a mechanism to assess the quality and reliability of the numerically derived models.\",\"PeriodicalId\":6771,\"journal\":{\"name\":\"2020 31st Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC)\",\"volume\":\"26 1\",\"pages\":\"1-5\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-08-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 31st Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ASMC49169.2020.9185215\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 31st Annual SEMI Advanced Semiconductor Manufacturing Conference (ASMC)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ASMC49169.2020.9185215","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
In the past, concepts like critical area analysis, have been successfully implemented to predict random and systematic layout induced effects. This has enabled semiconductor companies to have an initial estimate as to how a fixed process will respond to a variety of different designs. However, as the number of individual products increases, along with a reduction in the total number of wafers per product, it becomes increasingly difficult to determine which process parameters will lead to the highest possible yield for each individual product. We have outlined a methodology using machine learning that combines process and design data to greatly reduce the time needed for setting up new products. We have shown that similar designs (based on our feature extraction) behave similarly in the fab, thus allowing us to construct models that can eventually be used to find the optimal process conditions for a given design. Due to the nature or process optimization, this methodology also explores the use of SHAPley additive explanations (SHAP) in order to “interface” with existing human and physical explanations of the observations, thus providing a mechanism to assess the quality and reliability of the numerically derived models.