Optimised Machine Learning and Statistical Modelling for Predicting the Design Strengths of Hardened 3D Printed Concrete
3D Concrete Printing (3DCP) has emerged as a transformative construction technology, offering enhanced design flexibility, reduced material waste, improved construction efficiency, and safer working environments. Despite these advantages, the absence of standardised design provisions for structural 3DCP elements remains a major obstacle to its widespread adoption. This study develops statistical and machine learning models to predict the design strengths of hardened 3D-printed concrete. A comprehensive database of experimental results published between 2012 and 2025 was compiled and analysed, encompassing the compressive, flexural, and shear strengths of hardened 3DCP materials. Regression-based statistical modelling was employed to derive design strength equations and assess their consistency with conventional concrete design approaches. In parallel, several machine learning algorithms were developed using the same dataset to benchmark predictive performance and evaluate the robustness of the proposed statistical models. The statistical models achieved good predictive accuracy for compressive and flexural strengths, with coefficients of determination (R2) of 0.83 and 0.67, respectively. By contrast, the shear strength model exhibited greater variability (R2 = 0.39), reflecting the limited availability and considerable scatter of published experimental data. Benchmarking against the optimised machine learning model demonstrated excellent agreement with the proposed statistical equations, yielding coefficients of determination of 0.93 and 0.94 for compressive and flexural strengths, respectively. The proposed modelling framework provides preliminary design strength predictions for hardened 3D-printed concrete, contributing to the development of robust structural design guidelines and facilitating the wider adoption of 3DCP in construction practice.