Data-Driven Parameter Calibration in Additive Manufacturing for Construction (2025-05)¶

This paper introduces the Learning by Printing framework, designed to enhance performance and robustness in extrusion-based Additive Manufacturing in Construction. Leveraging Fabrication Information Modeling (FIM) as a digital backbone, the framework integrates evaluation, prediction, and calibration stages into a closed fabrication-learning loop. An experimental study on a clay extrusion setup demonstrates the framework’s ability to optimize structural performance through data-driven parameter calibration. The Gaussian Process prediction model achieves over 95% accuracy, while calibration shows to improve system performance. Future work will scale the framework to larger systems and integrate online learning for real-time control, advancing Learning by Printing toward a predictive and adaptive approach to digital fabrication.