CC BY 3.0 UnportedVu, Anh TuanAnh TuanVuParia, HamidrezaHamidrezaPariaGrunwald, TimTimGrunwaldBergs, ThomasThomasBergs2024-05-212024-05-212024-04-24https://publica.fraunhofer.de/handle/publica/468549https://doi.org/10.24406/publica-308010.24406/publica-308010.21741/9781644903131-197This study addresses the ever-increasing demands on glass optics for LiDAR systems in autonomous vehicles, highlighting the pivotal role of the recently developed Nonisothermal Glass Molding (NGM) in enabling the mass production of complex and precise glass optics. While NGM promises a significant advancement in cost- and energy-efficient solutions, achieving the requisite shape and form accuracy for high-precision optics remains a persistent challenge. The research focuses on expediting the development phase, presenting a methodology that strategically utilizes a sparse dataset for determining optimized molding parameters with a minimized number of experimental trials. Importantly, our method highlights the exceptional ability of a robust surrogate model to precisely predict the accuracy outputs of glass optics, strongly influenced by numerous input molding parameters of the NGM process. This significance emphasizes the surrogate model, which emerges as a promising alternative to inefficient traditional methods, such as time-consuming experiments or computation-intensive simulations, particularly in the realm of high-precision production for LiDAR glass optics. In contributing to optics manufacturing advancements, this study also aligns with contemporary trends in digitalization and Industry 4.0 within modern optics production, thereby fostering innovation in the automotive industry.enNonisothermal glass moldingGlass opticsLiDARSurrogate modelingBayesian optimizationIndustry 4.0DDC::600 Technik, Medizin, angewandte Wissenschaften::620 Ingenieurwissenschaftenconference paper