| Abstract Scope |
Ensuring consistent binder deposition remains a critical challenge in Binder Jetting Technologyand related additive manufacturing processes. This study presents a computer vision and machinelearning framework for real-time monitoring and predictive modeling of binder behavior, with aparticular focus on wetted region development during printing. Captured images of the print bedare processed using advanced segmentation algorithms to isolate binder-affected zones and enablecontinuous, high-resolution tracking of deposition patterns. These data streams are used to trainmachine learning models that capture binder flow dynamics, spread rates, and depositionuniformity across successive layers. The system functions as a non-invasive, automated diagnostictool capable of identifying process deviations such as under-deposition, excessive wetting, oruneven binder distribution—all of which directly impact part integrity and dimensional precision.Beyond defect detection, the predictive modeling supports adaptive feedback control, enablingimproved consistency and reduced variability in additive manufacturing builds. Experimental results validate the ability of this approach to characterize binder deposition in real time while accurately modeling key parameters that govern print quality. Insights generated from this method pave the way for closed-loop control strategies, reduced waste, and improved process repeatability. By integrating machine learning with vision-based monitoring, this work demonstrates how data- driven methods can accelerate the transition toward intelligent, autonomous additive manufacturing systems. |