Engineering the future: AI ‐driven design tool for 3D ‐printed concrete buildings

Abstract

Digital technologies, such as 3D printing, are revolutionizing the construction industry by automating structure assembly. However, the lack of simultaneous printing and standardization of design methods hinders full automation. This study presents an AI‐based design tool that forecasts the load‐carrying capacity of 3D‐printed concrete walls under combined axial and lateral loading and automatically generates structural designs. This tool also identifies the most efficient cross‐sectional geometry, supporting the identification of material‐efficient cross‐sectional geometries that satisfy the prescribed strength requirement. Two AI models, multilayer perceptron (MLP) and convolutional neural network (CNN), were evaluated. From the point of view of mean absolute percentage error, the CNN model had a maximum test error of 5.99%, while for the MLP model, the errors were greater than 10% in most cases and even reached a maximum of 25% in a worst‐case scenario. For the present database, the CNN models provided lower prediction errors than the investigated MLP alternatives and were therefore selected as the preferred surrogate formulation for capacity‐based section screening. The pre‐trained version of this design tool has been uploaded to a permanent repository, with an included link to facilitate further research into fully automated 3D concrete printing.