Probabilistic and Global Sensitivity Analysis for 3D Concrete Printing (2026-04)¶

The high uncertainty during 3D concrete printing (3DCP) process presents significant challenges for quality control and process reliability due to the strong nonlinearity of material behavior and the high computational cost associated with conventional sampling-based approaches. This study proposes an efficient probabilistic framework that integrates point-evolution kernel density estimation (PKDE) with Fréchet-derivative-based global sensitivity analysis (Fre-GSA) to enable rapid uncertainty propagation through the 3DCP process and systematic identification of dominant risk drivers governing associated buildability. The proposed framework reconstructs high-fidelity output probability distributions using only 80–200 representative samples, achieving accuracy comparable to large-scale Monte Carlo simulations (104 ∼ 106 samples) with Kullback-Leibler divergences below 0.01. Beyond computational efficiency, the Fre-GSA yields directional and physically interpretable sensitivity indices, revealing complex nonlinearities and non-variance-dominated effects that are not captured by traditional variance-based methods. Application to straight wall and hollow cylinder geometries demonstrates that the dominant uncertainty source is geometry-dependent: elastic buckling-governed walls are dominated by the initial elastic modulus, whereas plastic collapse-governed cylinders are dominated by initial cohesion. These findings offer actionable insights for targeted quality control and robust process optimization in digital construction.