Data-driven additive manufacturing with concrete: Enhancing in-line sensory data with domain knowledge, Part I: Geometry
Jelle Versteege, Rob Wolfs, T.A.M. Salet
Abstract
First-time-right manufacturing is an important step toward unlocking the full potential of digital fabrication with concrete (DFC), which can be advanced through data-driven approaches. Non-invasive in-line sensors can collect vast amounts of measurements during the manufacturing process. However, knowledge-driven feature engineering (KDFE) strategies are necessary to extract meaningful information, referred to as features, from the raw sensory data. This contribution, part of a two-part study, presents an approach to integrating KDFE with various in-line sensors in a 3D concrete printing (3DCP) facility, focusing on 2D laser scanning techniques to capture the ‘as-printed’ layer geometry during production. The geometric profiles are translated into features that quantify layer dimensions, cross-sectional area, and surface texture, reducing data complexity while enhancing relevancy. Real-world data is utilized to demonstrate the approach. A companion paper extends the methodology to other sensors, including those monitoring moisture and temperature, further advancing process monitoring in 3DCP. • In-line sensors enable autonomous quality assessment in 3DCP systems. • Knowledge-driven feature engineering reduces complexity in high-dimensional data. • Laser scanning techniques allow for detailed layer and surface geometry analysis. • The computation of geometric features is demonstrated using real-world data.