2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026): Process Development In-Situ Monitoring
Program Organizers: David Leigh, University of Texas at Austin
Wednesday 8:00 AM
August 5, 2026
Room: Comal
Location: AT&T Center
8:00 AM
Smoke, Mirrors, and Melt Pools: An Assessment of Commercial In-Situ Monitoring Solutions for Metal Laser Powder Bed Fusion: Abdalla Nassar1; Liam Adler-Pollock1; Christopher Apple1; 1ASTRO America
In the context of Laser Powder Bed Fusion (PBF-LB), the ability to detect and correct process drifts and anomalies has long been the "Holy Grail" of additive manufacturing. In-situ inspection promises near-real-time identification of flaws and non-conformances during production. Indeed, lab-scale and commercial systems have been developed and demonstrated to identify some process and part flaws. To assess the state-of-the-art of in-situ inspection, five commercially viable solutions were demonstrated within an ASTRO-ASTM-InSPIRE sponsored challenge. These companies — Addiguru, Additive Assurance, Applied Optimization, JENTEK Sensors, and Phase3D — demonstrated their solution on a commercial PBF-LB system. Here, we overview the technologies demonstrated along with insights on their current utility. Furthermore, we demonstrate that combining high-resolution melt pool imaging with near-infrared long-exposure imaging enables the detection of flaws on the order of tens of microns in thin-walled lattice structures. Applications include quality assurance for heat exchangers and other thin-walled structural components.
8:20 AM
Validation of Vision-Based Detection of Powder Spreading Defects in PBF-LB Using Light Scanning: Yuvraj Upadhyay1; Brian Johnstone1; Nicole Van Handel1; Maegan Lenertz1; Thomas Kurfess1; Kyle Saleeby1; 1Georgia Institute of Technology
Layerwise vision systems are common process control tools in additive manufacturing due to easy setup and high information density. In laser powder bed fusion (PBF-LB), optical imaging reveals spreading and melt defects, but quantification depends heavily on the acquisition system and ambient conditions. This work describes a method for validating detection of powder spreading defects, replicable for assessing optical camera capability via structured light scanning. As a case study, optical camera and laser profilometer captures of recoater "hopping" impressions are compared to establish camera- and system-specific detection limits. The correlation between super-elevation in lased regions and hopping in powder regions is also examined, alongside computer vision methods to quantify powder bed quality. Optical imaging and light scanning detect minimum hopping depths of 0.1 mm and 0.015 mm, respectively.
8:40 AM
From Simulation-Based Optimization to Volumetric Parts: Effects of Computer-Generated Beam Profiles on Melt-Pool Geometry and Part Density in PBF-LB/M of IN718: Richard Off1; Vijaya Holla2; Jonas Grünewald1; Katrin Wudy1; 1Professorship of Laser-based Additive Manufacturing, Technical University of Munich, Germany; 2Chair of Computing in Civil and Building Engineering, Technical University of Munich, Germany
Laser beam shaping can improve productivity and process stability in laser-based powder bed fusion of metals (PBF-LB/M). However, the influence of numerically optimized, computer-generated beam profiles on the manufacturing of volumetric parts remains largely unexplored. This study investigates three beam profiles for PBF-LB/M of IN718, targeting a brick-like melt-pool cross-section, a homogeneous temperature field, and a Gaussian-like melt-pool cross-section with limited peak temperature. Cube specimens were manufactured using varied laser power, scan speed, and hatch distance. Relative density was determined by Archimedean density measurements, and melt-pool geometry was evaluated by metallographic cross-section analysis. The investigated beam profiles achieve relative densities > 99.5 %, increase the build rate by up to 300 %, and thereby enlarge the process window compared with a Gaussian reference process. The results demonstrate the potential of computer-generated beam profiles for high-productivity PBF-LB/M of IN718, while revealing remaining challenges related to top-surface waviness and geometrical accuracy.
9:00 AM
In-Situ Characterization of Balling in Laser Powder Bed Fusion from Multi-Modalities Monitoring Data: Zhuo Yang1; Ho Yeung1; 1NIST
Balling is a defect in LPBF process and is primarily caused by melt pool instability. It leads to discontinuous melting, increased porosity, and poor recoating conditions. This study investigates an in-situ approach to characterize the balling effect using a high-speed coaxial camera and a laser line profiler. A total of 252 single tracks were produced on an SS316 substrate by varying laser power from 60 to 460 W and scan speed from 350 to 2000 mm/s. Balling was observed predominantly under combinations of high laser power and high scan speed. The results indicate that both sensing modalities are directly correlated with balling; however, coaxial melt pool imaging is highly sensitive to balling characterized by excessive surface area, whereas laser profiler measurements are more sensitive to balling with excessive vertical height. Neither modality alone can reliably identify all balling events. In contrast, combining both modalities enables accurate prediction of balling locations.
9:20 AM
Segment-Resolved Probabilistic Defect Quantification in PBF-LB/M Single Scan Tracks Using µCT-Referenced Photodiode Features: Jonathan Utsch1; Jana Harbig1; David Zentgraf1; Matthias Weigold1; Holger Merschroth1; 1Technical University of Darmstadt, Institute for Production Management, Technology and Machine Tools
Reliable quality assessment in powder bed fusion – laser beam of metals (PBF-LB/M) requires not only defect detection but also quantitative estimation of defect occurrence and severity. This study presents an experimental validation of a probabilistic defect quantification approach based on synchronized laser scanner position data, on-axis photodiode signals and micro-computed tomography. To isolate fundamental correlations between defect formation and monitoring signal features single melt tracks generated from a systematic process parameter variation are analyzed. Due to the limited spatial-temporal resolution of monitoring signals and the stochastic nature of melt pool dynamics, defect-specific assignment is challenging, making segment-wise probabilistic estimation more robust. Therefore, scan tracks are discretized into fixed-length segments, enabling localized correlation of process signatures with µCT-based defect count and size. Process-physically motivated signal features are subsequently used as informative priors for Bayesian inference of defect probability and severity.
9:40 AM Break
10:00 AM
Multimodal Non-Destructive Evaluation of Leak Behavior in LPBF 316L Stainless Steel Pipe Structures: Jorge Avila1; 1The University of Texas at El Paso
This study investigates leak detection in 316L stainless steel pipe components fabricated via Laser Powder Bed Fusion (LPBF). A modular sensing platform integrating nondestructive evaluation (NDE) techniques (thermal imaging and ultrasonic monitoring) was developed to detect and characterize leaks in additively manufactured structures under medium-pressure gas conditions. The system enables real-time monitoring without interrupting flow and can be adapted for deployment on unmanned aerial vehicles (UAVs) for remote inspection. A custom pneumatic test rig with temperature and pressure instrumentation was used to evaluate LPBF-printed 316L tee pipe specimens. Experimental results and simulations confirmed the platform’s capability to identify leakage and assess structural integrity. Additionally, a Python-based image analysis workflow using metallography and optical microscopy quantified porosity and identified potential internal leak pathways. Computed tomography (CT) in 2D and 3D validated defect presence. This integrated methodology enhances leak detection and quality control in additively manufactured components.
10:20 AM
Development of an Eddy Current In-situ Diagnostic for Directed Energy Deposition: Michael Juhasz1; Saptarshi Mukherjee1; Lei Peng2; Sapana Shirsat2; Yiming Deng2; Joseph Tringe1; Ethan Rosenberg1; 1Lawrence Livermore National Laboratory; 2Michigan State University
Eddy current sensing offers a potential path for direct, real-time observation of melt pool behavior during directed energy deposition (DED). While prior additive manufacturing applications have focused mainly on layer-wise defect detection, this work explores the feasibility of adapting eddy current sensing for real-time, in-situ use in the challenging DED environment. The long-term goal is to support both defect detection and improved understanding of subsurface process dynamics. This presentation summarizes early validation efforts aimed at identifying technical barriers and establishing a foundation for real-time implementation. We describe the development of progressively more representative test cases to assess sensor performance, signal response, and integration considerations under controlled conditions relevant to DED. Initial results indicate that eddy current sensing may be a viable approach for monitoring subsurface phenomena during processing, while also revealing practical challenges for industrial deployment. Prepared by LLNL under Contract DE-AC52-07NA27344.
10:40 AM
A Simplified Vision Framework for Feedstock Monitoring and Flow-Related Defect Prevention in Pellet and Flake Extrusion Systems: Daniel Bañuelos Chacon1; Eric MacDonald1; Andrei Alexandru Popa2; 1University of Texas at El Paso; 2University of Southern Denmark
Stable material extrusion is essential for reliable additive manufacturing, yet it is often disrupted by over- or under-extrusion, flow discontinuities, and clogging. This challenge is critical in extrusion-based systems handling nonhomogeneous flakes and low-quality feedstocks, such as recycled polymers, where maintaining consistent flow is difficult. While computer vision has been explored, many solutions rely on complex sensor setups. In contrast, this work proposes a simplistic stereo-vision system to monitor the extrudate at or below the nozzle and apply real-time G-code adjustments to stabilize flow.Particularly useful for clog prevention in auger-based systems with large deposition nozzles, the approach scales to process parameter tuning and flow stabilization. Using flow consistency as a proxy for process health, system supports reliable operation with heterogeneous materials. Demonstrated on an auger-based flake-extruding setup, the method limits extrudate variation at the nozzle to within 10% and applies to filament manufacturing, where consistent diameter is critical.
11:00 AM
In-Situ Displacement Monitoring of Roller-Powder Interactions in Binder Jetting: Tugrul Yaylali1; 1BYU
Binder jetting process monitoring largely depends on simple optical sensing methods. These methods lack the ability of capturing mechanical roller-powder interactions. This study presents a novel in-situ monitoring approach which transforms the roller into a diagnostic probe by measuring roller displacement during powder spreading with micron resolution. Process parameters such as layer thickness, ambient humidity, roller speed, and saturation produce unique displacement signals. These changes can be monitored to gain insights into the process variable relationships. In this work, the displacement response is used to identify the thresholds at which spreading behavior is no longer influenced by underlying powder layers. The relationship between roller displacement and packing fraction of green parts is measured. The results are used to evaluate the impact of bound powder on the density of subsequent spread layers.
11:20 AM
G-Code-Aware Segmentation for Quality Monitoring in Direct Ink Writing: Jesus Diaz1; Syed Bashar1; Satyajayant Misra1; Chaitanya Mahajan1; 1New Mexico State University
Direct Ink Writing (DIW) is gaining momentum for its capacity to deposit a wide range of materials; however, the inherent rheological behavior of these substances often leads to expansion and shrinkage, compromising structural quality. To mitigate these geometric inaccuracies, we present GG-Net, a U-Net-derived segmentation model designed for real-time quality monitoring. GG-Net processes in-situ imagery at each layer, comparing the printed output directly against the intended toolpath mapping. By effectively distinguishing the top-most layer from underlying structures, the framework enables a precise, layer-by-layer evaluation of the printed part’s fidelity. This automated comparison enables immediate identification of deviations between the physical print and its digital counterpart, thereby minimizing waste and improving the efficiency of the material extrusion process.
11:40 AM
Agentic Process Optimization for Selective Laser Sintering: Peter Pak1; Amir Barati Farimani1; 1Carnegie Mellon University
Within the additive manufacturing process there are many operational and environmental factors that can affect the build quality or build feasibility of the final part. Mitigation of these effects is often achieved using feedforward process control where model based approaches are applied to anticipate problematic conditions that can lead to issues during the build process. Feedback control addresses these issues during the build process utilizing visual, thermal, or depth information obtained through various sensors to dynamically adjust parameters and toolpath trajectories in an effort to resolve potential build defects. In this work an agentic system enables the intelligent automation of parameter optimization and in-situ process monitoring during the selective laser sintering process. The agentic system is evaluated on a range of polymers and polymer based composites with the mechanical properties tested to the ASTM 638-22 standard.