Explainable Data-Driven Analysis of Uncertainty Propagation in 3D Concrete Printing via Adaptive Polynomial Chaos Expansion (2025-07)¶

Understanding buildability uncertainty during the printing process remains a major challenge in 3D concrete printing (3DCP). This paper investigates how underlying uncertainty sources influence buildability in 3DCP. An explainable data-driven framework is proposed by integrating adaptive polynomial chaos expansion (aPCE) with Shapley additive explanation (SHAP) to efficiently analyze uncertainty propagation in 3DCP and interpret its underlying mechanisms. Applied to two diverse 3DCP scenarios (a straight wall and a hollow cylinder), the framework achieves accuracy comparable to Monte Carlo simulations with 90% less computational cost, while effectively capturing uncertainty propagation patterns and identifying dominant drivers of buildability uncertainty among potential uncertainty sources. The findings are valuable for 3DCP practitioners and can provide them with actionable insights for mitigating buildability uncertainty, thereby improving the reliability of the printing process. These insights also motivate future research on controlling variability in dominant factors and developing uncertainty-aware process optimization strategies for 3DCP.