Towards Adaptive Additive Manufacturing with Semi-autonomous Robotic Binder Jet 3D Printing of Concrete

Abstract This research proposes an integration of image-based feedback monitoring into a process of additive manufacturing with the aim of expanding 3D concrete printing (3DCP) capabilities toward more diverse material feedstock. Building upon our ongoing research focused on current binder-jetting techniques in 3D concrete printing, we address sequential quality control of printing with coarse powders that advance the adaptation of this technology for sustainable construction with a focus on material reuse. By incorporating an image-based monitoring system into binder jetting, we aim to enhance the adaptability and robustness of 3DCP towards using more variable and irregular powder feedstock from waste streams. A comprehensive dataset, collected over a year and comprising approximately 400 samples, was leveraged to formulate an image-based flaw-detection algorithm. Deep learning training was employed to ensure precise segmentation of printed patterns. The adoption of a dual-resolution imaging strategy further accentuated the quality of data collection and interpretation. Central to our flaw-detection mechanism is a feedback loop framework, designed to discern discrepancies between desired and printed binder deposition pathways. This study proposes an integrated approach of binder jetting with flaw-detection feedback and error rectification within 3DCP, holding a transformative potential for the broader construction technology landscape and material reuse.