Machine Learning Approximation of Water Transport in 3D-Printable Composites via Karsten Tube (2025-05)¶

Non-destructive testing methods based on water penetration into porous materials through a defined surface area are widely used for characterizing construction materials such as concrete, alkali-activated mortars, and soil-based composites. Among these methods, the Karsten tube test stands out due to its practicality and costeffectiveness, allowing applications in laboratory and field conditions. In this study, water ingress measurements were performed on 3D-printable composites namely, cement based, alkali-activated, and soil/clay based materials using the karsten tube. With the increasing use of 3D-printed structural composites produced by layered manufacturing techniques, the need for rapid and reliable determination of their water transport mechanisms has become critical. In this context, the amount of water transported through the materials over different durations was measured and classified based on material types. The experimental time series data obtained were analyzed using machine learning algorithms to develop predictive models. The performance of the models was evaluated through supervised learning performance metrics, and different algorithms were comparatively assessed. As a result, artificial intelligence-based models with high predictive accuracy were developed to characterize the water transport mechanisms of 3D-printable composite materials.