Machine Learning-Based Real-Time Tracking and Modeling of Binder Deposition in Additive Manufacturing

Wednesday, October 22, 2025: 12:50 PM
Dr. Navin Manjooran, Ph.D., MBA, CEng. FASM, FACerS, FIIM, FIE, FIMMM, FIIE, FAEM, HoF-VTAEE , Solve, Windermere, FL
Shahjahan Hossain , University of Central Florida, Orlando, FL
Pranta Sarkar , University of Central Florida, Orlando, FL
Ranajay Ghosh , University of Central Florida, Orlando, FL
Dr. Ramesh Subramanian , Siemens Energy Inc., Orlando, FL
Dr. Gary R. Pickrell , Virginia Polytechnic Institute and State University, Blacksburg, VA
Accurate, real-time monitoring of binder deposition is critical for enhancing quality, consistency, and efficiency in Binder Jetting Technology (BJT) and related additive manufacturing (AM) processes. This work introduces an image-based machine learning framework to segment, track, and model binder deposition dynamics—specifically focusing on wetted area evolution during the process.

The approach employs computer vision algorithms for image segmentation to identify and isolate wetted regions from captured print bed images. By distinguishing binder-affected zones, the system enables continuous tracking of deposition patterns with high spatial and temporal resolution. The segmented image data is then used to train predictive machine learning models that characterize binder flow behavior, spread rates, and deposition uniformity across layers.

This method provides a non-invasive, automated monitoring tool that can detect deviations such as over-saturation, under-deposition, or inconsistencies in binder distribution—critical factors affecting part integrity and dimensional accuracy. The integration of data-driven modeling not only facilitates in-process quality assurance but also supports closed-loop feedback control and process optimization.

Experimental validation confirms the system’s ability to track binder behavior in real time and model key parameters influencing print quality. The insights derived from this analysis can inform adaptive control strategies, reduce material waste, and improve repeatability in AM workflows.

In summary, this work demonstrates how combining image processing with machine learning delivers a powerful platform for real-time analysis and predictive modeling of binder deposition. It supports smarter, more autonomous manufacturing environments and lays the groundwork for future advancements in precision AM process control.