A Data-Driven View on Digital Fabrication with Concrete: The Case of Interlayer Bond Strength

This paper addresses the potential of data-driven methodologies in enhancing sustainability and efficiency within digital fabrication with concrete (DFC). It emphasizes the role of real-time in-line sensors and their autonomous synchronization into a digital representation of the physical object. This digital representation enables the prediction of variables that are difficult to measure directly, such as the interlayer bond strength (IBS) in 3D concrete printing. For this purpose, a data-driven virtual sensor for the IBS is in development, utilizing machine learning algorithms or other statistical tools to predict its value based on in-line sensory data. The development of such a virtual sensor relies heavily on a dataset where sensory data is paired with the data from experiments performed to determine the IBS. The paper discusses different strategies for this pairing procedure, noting both the predictive performance and practical concerns. The study is contextualized within a 3DCP facility, focusing on a select few of the integrated sensors and showcasing their outcomes in more detail. Using data collected to explore the effects of variable manipulations on the IBS, the paper evaluates three pairing strategies – object scale, layer scale, and specimen scale – for their influence on the resulting dataset's composition. Although the exact impact of these strategies can only be determined after the virtual sensors are finalized, the paper concludes that embracing more detailed data-driven methodologies has the potential to advance process monitoring in DFC.