Data-driven reliability-oriented buildability analysis of 3D concrete printed curved wall

The inherent uncertainties, particularly material uncertainties, significantly impact the buildability of 3D concrete-printed curved walls, leading to substantial variations that complicate quality control. To address this, a data-driven stochastic analysis framework is proposed for reliability-oriented buildability evaluation. Material uncertainties are quantified using a maximum likelihood-based stochastic parameter estimation method and considered as the uncertainty sources. Subsequently, a data-driven model, namely sparse Gaussian process regression (SGPR) model, is trained and combined with Monte Carlo simulation to assess the stochastic behavior of curved wall buildability. The influences of print speed, layer height, and horizontal curvature on buildability are analyzed under varying reliability levels. Additionally, an empirical model is proposed for the rapid evaluation of maximum buildability at specified horizontal curvature and reliability levels, providing significant practical value for 3D concrete printing designers. The impact of other uncertainty sources including the model error on reliability-oriented buildability is also discussed. These sources exhibit negligible influence when their intensities are less than 30 % of that caused by material uncertainty. Furthermore, the feasibility of the data-driven reliability-oriented buildability analysis for more complex geometry is also demonstrated.