CC BY-NC-ND 4.0Mende, HendrikHendrikMendeLeyendecker, LarsLarsLeyendeckerUpadhyay, KashyapKashyapUpadhyayGrunert, DennisDennisGrunertSchmitt, Robert H.Robert H.Schmitt2025-06-232025-06-232025-06-16https://doi.org/10.24406/publica-4799https://publica.fraunhofer.de/handle/publica/48885910.1016/j.procir.2025.03.05410.24406/publica-4799The application of machine learning (ML) for surrogate modelling in model-based optimization (MBO) of manufacturing processes offers great potential for achieving time, cost, and resource efficiency for process development. This paper explores two main optimization strategies of MBO: sequential and non-sequential optimization. We utilize a case study from non-isothermal glass molding (NGM) in which a thin glass component is formed into a 3-dimensional shape. We apply a sequential optimization approach utilizing an ML-surrogate model with Bayesian optimization. We investigate these approaches and propose a comprehensive methodology for the application of MBO, offering valuable insights for the implementation of sequential and non-sequential optimization techniques in similar manufacturing contexts. Furthermore, the results from the application on the NGM use case are compared, highlighting their effectiveness in enhancing the NGM process’s quality.enModel-based optimizationSurrogate modellingMachine learningProcess optimizationMethodology600 Technik, Medizin, angewandte Wissenschaften::620 IngenieurwissenschaftenOptimizing Non-Isothermal Glass Molding Processes: Methodology and Comparative Analysis of Sequential and Non-Sequential Approachesjournal article