From Qualitative Diagnosis to Adaptive Compensation for Filament Imperfections During On-Site Robotic 3D Printing (2025-09)¶

Robotic 3D concrete printing in construction (R3DCP) is a challenging task due to the complexity of the environment and the properties of the material. In this case study, we employed a pattern recognition-based approach to diagnose operational modes during the printing stage. This approach involves learning the characteristic patterns associated with different classes of operational modes within the multidimensional feature space, which aids in more informed decision-making for real-time control and fault diagnosis. A set of models was trained to classify operational points under various environmental conditions (10, 20, and 30°C). The results indicate that the material exhibits distinct flowability in different environmental settings, which increases the transition flexibility between different operational modes and contributes to a higher occurrence of deposition imperfections. To ensure that the system can execute rapid and robust compensatory actions to maintain print quality, an adaptive robot velocity compensation strategy was proposed. The experimental results demonstrated that the integrated diagnosis-compensation strategy reduced the percentage of imperfections by more than half in the deposition of the material.