Intelligent real-time quality control for 3D-printed concrete with near-nozzle secondary mixing

An intelligent monitoring method coupled with a feedback adjustment system was developed to accomplish real-time quality control of 3D printed concrete (3DPC). To achieve instant concrete rheology modification, a liquid accelerator was added in 3DPC before extrusion employing the near-nozzle secondary mixing strategy. Effects of accelerator contents on fresh and mechanical properties of 3DPC were explored, providing valuable insights for real-time regulation of material fluidity. Images of filaments captured during concrete 3D printing were categorized into 15 classes according to geometry and material characteristics. Subsequently, classification models were developed based on a lightweight modified Inception-ResNet, and its superiority was confirmed through a comparison with traditional VGG network. The classification ability, confusion harmful effects and the generalization performance of the models were evaluated. Eventually, guided by AI-aided quality assessment, real-time automated adjustments to extrusion speed and accelerator injection rate were realized, achieving in-situ quality control for automatic construction of 3DPC structures.