Tailoring 3D printed concrete through explainable artificial intelligence

Advances on the construction front continue to rise as the next industrial revolution (Construction 4.0) nears. One promising front revolves around additively fabricated or simply 3D printed concrete. The growing number of ongoing parallel research programs has now made it possible to collect a large amount of data on such concrete as, up to this point, the open literature lacks a comprehensive database. Thus, this paper presents the largest database spanning over 300 experiments on 3D printed concrete. This database is then examined via multilinear regression as well as two explainable artificial intelligence (XAI) algorithms, namely, Random Forest and XGBoost, to arrive at a working model capable of predicting the compressive strength property for 3D concrete mixtures that incorporate the following seven features: age of specimens, as well as the magnitude of cement, water, fly ash, silica fume, fine aggregate, and superplasticizer. Findings from this work infer the superiority of XAI models in predicting the strength property of 3D printed concrete. Our analysis identifies two features, namely, the age of specimens and the quantity of fine aggregate, as the most important features that can accurately predict the compressive strength property. Finally, the deployed explainability methods successfully quantified the highly nonlinear relations between the selected features and compressive strength, and this newly acquired knowledge can help tailor functional concrete mixtures.