{"title":"极端冷弯:几何考虑和形状预测与机器学习","authors":"Keyan Rahimzadeh, Evan Levelle, J. Douglas","doi":"10.47982/cgc.8.460","DOIUrl":null,"url":null,"abstract":"Cold-bent glass is seeing increasing adoption in construction projects with non-planar geometries. This paper presents work undergone for a set of four high-rise towers, featuring 11,136 unique cold-bent panels, hundreds of which are pushed beyond 250mm. The panels are all unique, non-rectangular, and in some cases, slightly curved. The challenging geometry complicates the prediction of the final panel shape, which is an essential step for producing fabrication drawings of a panel’s flat shape prior to bending. While Machine Learning is still a nascent technology in the AEC industry, prediction is a class of problems for which many Machine Learning techniques are ideal, especially when dealing with a large quantity of data, or in this case, panels. The paper discusses the geometric characteristics of highly bent glass, a methodology for the shape prediction of the panels, and the use of Machine Learning in its implementation. The methodology was deployed for over 3,500 pieces of installed architectural glass, and was shown to reduce geometric deviations as much as 75%, down to sub-millimetre tolerances.","PeriodicalId":332145,"journal":{"name":"Challenging Glass Conference Proceedings","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2022-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Extreme Cold-Bending: Geometric Considerations and Shape Prediction with Machine Learning\",\"authors\":\"Keyan Rahimzadeh, Evan Levelle, J. Douglas\",\"doi\":\"10.47982/cgc.8.460\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Cold-bent glass is seeing increasing adoption in construction projects with non-planar geometries. This paper presents work undergone for a set of four high-rise towers, featuring 11,136 unique cold-bent panels, hundreds of which are pushed beyond 250mm. The panels are all unique, non-rectangular, and in some cases, slightly curved. The challenging geometry complicates the prediction of the final panel shape, which is an essential step for producing fabrication drawings of a panel’s flat shape prior to bending. While Machine Learning is still a nascent technology in the AEC industry, prediction is a class of problems for which many Machine Learning techniques are ideal, especially when dealing with a large quantity of data, or in this case, panels. The paper discusses the geometric characteristics of highly bent glass, a methodology for the shape prediction of the panels, and the use of Machine Learning in its implementation. The methodology was deployed for over 3,500 pieces of installed architectural glass, and was shown to reduce geometric deviations as much as 75%, down to sub-millimetre tolerances.\",\"PeriodicalId\":332145,\"journal\":{\"name\":\"Challenging Glass Conference Proceedings\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-06-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Challenging Glass Conference Proceedings\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.47982/cgc.8.460\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Challenging Glass Conference Proceedings","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.47982/cgc.8.460","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Extreme Cold-Bending: Geometric Considerations and Shape Prediction with Machine Learning
Cold-bent glass is seeing increasing adoption in construction projects with non-planar geometries. This paper presents work undergone for a set of four high-rise towers, featuring 11,136 unique cold-bent panels, hundreds of which are pushed beyond 250mm. The panels are all unique, non-rectangular, and in some cases, slightly curved. The challenging geometry complicates the prediction of the final panel shape, which is an essential step for producing fabrication drawings of a panel’s flat shape prior to bending. While Machine Learning is still a nascent technology in the AEC industry, prediction is a class of problems for which many Machine Learning techniques are ideal, especially when dealing with a large quantity of data, or in this case, panels. The paper discusses the geometric characteristics of highly bent glass, a methodology for the shape prediction of the panels, and the use of Machine Learning in its implementation. The methodology was deployed for over 3,500 pieces of installed architectural glass, and was shown to reduce geometric deviations as much as 75%, down to sub-millimetre tolerances.