Machine-learning networks to predict the ultimate axial load and displacement capacity of 3D printed concrete walls with different section geometries

Abstract

This paper presents details on the machine learning (ML) models for predicting the ultimate axial load capacity and ultimate displacement capacity of 3D printed concrete (3DPC) walls. The large database required for training and testing procedures of ML models is generated using a validated finite element (FE) model, verified using experimental specimens from the literature. Then, a wide range of physical and mechanical properties is selected to form a large database of 3DPC walls. To this end, 61800 3DPC walls with five different cross-sections and with various geometries were analyzed. The ultimate axial load capacity and ultimate displacement capacity of each wall are determined using explicit dynamic analysis. In conclusion, ML algorithms provide accurate predictions with coefficient of determination values of 0.95 and above. It should be noted that the prediction of the maximum axial load capacities was better than that of the ultimate displacement capacities.