Real-time monitoring and feedback operation system for 3D-printed concrete based on computer vision

While 3D-printed concrete (3DPC) has been widely studied and successfully applied in various real-world engineering projects globally, challenges persist in the practical application of 3DPC, with the most significant one being quality monitoring and control [1, 2]. Concrete is batch mixed but continuously extruded in 3D printing process, resulting in instabilities in material status. Therefore, secondary near-nozzle mixing process was proposed [3]. Pre-mixed fresh concrete with a high flowability was secondarily mixed with rheological modifier in the near-nozzle mixer to form printable concrete [1]. The addition rate of admixture during secondary mixing should be matched with the time-dependent concrete fluidity. Consequently, inline monitoring was suggested for printed filament [2] while there is still a lack of research on the feedback system for quality control to realize real-time adjustment operations for 3DPC with the secondary mixing process. This study implements secondary mixing with an accelerator as the chosen strategy for 3DPC, and an intelligent real-time adjustment system based on computer vision is developed for quality control of 3DPC. Visual geometry group (VGG) network [4] was employed to classify the images of printed filament, and a feedback operation system based on this classification results was built to regulate the extrusion rate and material fluidity.