A data-driven framework for determining extrudability in 3D concrete printing: Experiments and artificial neural network modeling

Citations

WEB OF SCIENCE

0
Citations

SCOPUS

0

초록

Extrudability is a critical aspect in 3D concrete printing (3DCP), as it governs extrusion stability, geometric accuracy, and surface quality. This study proposed an artificial neural network-based data-driven framework for determining extrudability during 3DCP. The influences of printing parameters and rheological properties on extrusion performance were experimentally quantified in terms of extrusion rate, extrusion error, average filament width, average surface roughness, and tearing region. The resulting dataset was used to develop regression models to predict extrusion rate and average filament width and classification models to determine extrusion consistency and texture quality. These models exhibited robust generalization performance under k-fold crossvalidation. On the test set, the regression models exhibited mean absolute percentage error (MAPE) values of 15.76% and 8.27% for extrusion rate and average filament width, respectively, and the classification models exhibited F1-scores of 0.82 and 0.86 for texture quality and extrusion consistency, respectively. Model interpretability was enhanced using Shapley additive explanations and partial dependence plot analyses. Validation experiments confirmed stable predictive performance when evaluating combinations of material and printing characteristics that were not explicitly included in the training data, yielding an average MAPE of 11.9% and no misclassifications. By formalizing extrudability using multiple quantitative indicators and combining experimental analysis, data-driven modeling, and independent experimental validation into a unified framework, this study linked rheological properties to printing performance within the investigated material and processing domain, providing a quantitative basis for future data-driven process control and automation in 3DCP.

키워드

3D concrete printing; Extrudability; Cementitious materials; Rheological properties; Printing parameters; Artificial neural network; SELF-COMPACTING CONCRETE; RHEOLOGICAL PROPERTIES; STRENGTH; THIXOTROPY; BEHAVIOR; DESIGN
제목
A data-driven framework for determining extrudability in 3D concrete printing: Experiments and artificial neural network modeling
저자
Jung, Ju Hyeok; Oh, Sangwoo; Choi, Seongcheol
DOI
10.1016/j.conbuildmat.2026.148243
발행일
2026-10
유형
Article
저널명
Construction and Building Materials
권
544