International Conferences A Case Study: Moving Window Optimization for Real-time Thickness Prediction Using Semiconductor Virtual Metrology Data with Screened Features
페이지 정보
조회 447회 작성일 26-01-15 22:11
본문
| Conference | ICIEA-EU 2026 (Milano, Italy) |
|---|---|
| Name | Donghyeon Koo, Seokkyu Pyo, Yuna Song, Beomseok Kim, Jihoon Hong, Kuhyun Lee, Suhwan Park, Mihee Lee, Dong-Hee Lee |
| Year | 2026 |
[Abstract]
In this paper, a comparative case study is conducted on determination of the optimal online model for predicting layer thickness response with respect to virtual metrology (VM). Although VM systematically stores and predicts large scale data online, the predictive performance of its embedded models may not always be guaranteed due to factors including complexity, high dimensionality and feature drifts of the operational data. Therefore, it is crucial for process engineers to consistently monitor and optimize the prediction models with respect to screened process variables of interest. ML-based predictors are combined with the moving window method and several online adaptive approaches for predictive performance comparison, the best of which is selected as the optimal model for online prediction of the response.