Comparative analysis of evolutionary computational methods for predicting mechanical properties of fiber-reinforced 3D printed concrete

Abstract In this study, Gene Expression Programming (GEP) and Multi-Expression Programming (MEP) were employed to predict the compressive strength (C-S) and flexural strength (F-S) of 3D-printed fiber-reinforced concrete (FR-3DPC). A comprehensive literature database consisting of 299 data points for C-S and 200 for F-S was compiled for model training. The models were evaluated using various statistical metrics, with MEP demonstrating superior predictive performance, achieving high R 2 values of 0.96 for C-S and 0.98 for F-S. To further interpret the model predictions, SHapley Additive exPlanations (SHAP), feature contribution score (FCS), and sensitivity analysis were conducted. The results from all three methods consistently highlighted the water-to-cement ratio, fiber content, loading direction, and cement content as the most influential factors affecting strength prediction. Additionally, a user-friendly Graphical User Interface (GUI) was developed for practical application. In conclusion, predictive equations derived from evolutionary programming, along with GUI incorporating diverse fiber types and loading directions, enhance the efficiency of 3D printing processes in construction. Graphical abstract