{"title":"Explainable artificial intelligence and multi-stage transfer learning for injection molding quality prediction","authors":"Chung-Yin Lin, Jinsu Gim, Demitri Shotwell, Mong-Tung Lin, Jia-Hau Liu, Lih-Sheng Turng","doi":"10.1007/s10845-024-02436-w","DOIUrl":null,"url":null,"abstract":"<p>High-precision optical products made of polymeric materials have been surging in recent years due to the prevalence of smartphones and their camera modules. Manufacturing fast-changing generations of high-precision optical lenses with accurately predicted qualities is a challenging task. Simulations and modern artificial intelligence (AI) techniques play crucial roles in accelerating precise process development. Coupled with computer simulation, this research employs a fusion of explainable AI (XAI) and multi-stage transfer learning (TL) approaches with artificial neural network (ANN) models to predict the surface profile variation of injection-molded polycarbonate (PC) lenses. The proposed method efficiently bridges preliminary simulations to injection molding experiments, covering a complete process development workflow from feature selection, process modeling, to experimental investigation in the same modeling domain. Only one model from scratch is required, which carries knowledge to the final quality prediction model. When compared with the conventional TL and the naïve model, the multi-stage TL approach provides better predictions with a maximum reduction of 64% and 43% in simulation and actual manufacturing data requirement, respectively. This research demonstrates a viable connection between each stage in the injection molding (IM) process development in predicting the qualities of high-precision optical lenses. Meanwhile, the combined usage of XAI and (multi-stage) TL confirms model explanations and pinpoints a potential pathway to assess future TL capabilities from the modeling perspectives.</p>","PeriodicalId":16193,"journal":{"name":"Journal of Intelligent Manufacturing","volume":"37 1","pages":""},"PeriodicalIF":5.9000,"publicationDate":"2024-05-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Intelligent Manufacturing","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1007/s10845-024-02436-w","RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0
Abstract
High-precision optical products made of polymeric materials have been surging in recent years due to the prevalence of smartphones and their camera modules. Manufacturing fast-changing generations of high-precision optical lenses with accurately predicted qualities is a challenging task. Simulations and modern artificial intelligence (AI) techniques play crucial roles in accelerating precise process development. Coupled with computer simulation, this research employs a fusion of explainable AI (XAI) and multi-stage transfer learning (TL) approaches with artificial neural network (ANN) models to predict the surface profile variation of injection-molded polycarbonate (PC) lenses. The proposed method efficiently bridges preliminary simulations to injection molding experiments, covering a complete process development workflow from feature selection, process modeling, to experimental investigation in the same modeling domain. Only one model from scratch is required, which carries knowledge to the final quality prediction model. When compared with the conventional TL and the naïve model, the multi-stage TL approach provides better predictions with a maximum reduction of 64% and 43% in simulation and actual manufacturing data requirement, respectively. This research demonstrates a viable connection between each stage in the injection molding (IM) process development in predicting the qualities of high-precision optical lenses. Meanwhile, the combined usage of XAI and (multi-stage) TL confirms model explanations and pinpoints a potential pathway to assess future TL capabilities from the modeling perspectives.
期刊介绍:
The Journal of Nonlinear Engineering aims to be a platform for sharing original research results in theoretical, experimental, practical, and applied nonlinear phenomena within engineering. It serves as a forum to exchange ideas and applications of nonlinear problems across various engineering disciplines. Articles are considered for publication if they explore nonlinearities in engineering systems, offering realistic mathematical modeling, utilizing nonlinearity for new designs, stabilizing systems, understanding system behavior through nonlinearity, optimizing systems based on nonlinear interactions, and developing algorithms to harness and leverage nonlinear elements.