Automatic Detection and Isolation of Filament Width Deviation During 3-D Printing of Recycled Construction Material

In this work, we deal with the problem of quality control in the continuous deposition of construction material (concrete), printed in 3-D by a robotic device, in order to produce shapes with complex geometry. Thus, a methodology for online quality monitoring of 3-D concrete printing has been developed. This methodology enables the automatic and real-time detection of deviations in the filament width during material deposition. It also allows to locate the origin of the deviation of the filament by establishing a correlation between the curvature of the shape to be printed, the speed of the robot-nozzle, and the flow rate of the printing material. Filament width estimation is performed using a deep learning instance segmentation model and a corresponding binary mask, then a morphology-based approach is applied to the mask to calculate the overall filament width. Finally, a fault detection and isolation-based method is then applied to monitor the residual signals from the printing process under the guidance of the generated adaptive thresholds to detect and isolate the filament width deviation geometric defect. The experimental results show that this printing quality control methodology improves the process of 3-D printing construction materials in real time.