Smart parameter space exploration for industry process optimization
Abstract
Industrial process optimization is increasingly driven by data rather than intuition. This project explores a data-driven framework for improving the performance of industrial manufacturing processes, with a particular focus on continuously generating gear grinding. By combining process monitoring, learning-based modeling, and smart sampling strategies, we aim to improve product quality and productivity while reducing experimental costs. Gaussian Process models are used to predict part quality outcome for a given set of configuration parameters and quantify uncertainty, enabling an efficient exploration-exploitation trade-off. These results demonstrate that meaningful performance gains can be achieved with limited data, while uncovering optimal operating regions and Pareto fronts relevant to industrial decision-making. and frequency-domain analysis using Fast Foureir Transform
Publication Reference
CSEM Scientific and Technical Report 2025, p. 29–30
Year
2025