Session Sheet - 2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026): Feedforward and Feedback Control in AM

2026 Annual International Solid Freeform Fabrication Symposium (SFF Symp 2026): Feedforward and Feedback Control in AM
Program Organizers: David Leigh, University of Texas at Austin

Monday 1:30 PM
August 3, 2026
Room: Comal
Location: AT&T Center

Session Chair: Chinedum Okwudire, University of Michigan; Sneha Prabha Narra, Carnegie Mellon University

1:30 PM
Adaptive Feedforward Control of Mass Flow Oscillations in Powder Blown Directed Energy Deposition : Neel Shah¹; Zachary Brunson¹; Jin Yeon Kim¹; Aaron Stebner¹; ¹Georgia Intitute of Technology
     Consistent powder mass flow is critical to part quality in blown powder Directed Energy Deposition (PB-DED), yet rotary feeders exhibit two persistent instabilities: periodic oscillations driven by hopper mechanics and stochastic parameter drift from powder compaction and settling. We demonstrate a control architecture that tightly couples feedforward and feedback strategies: a feedforward controller derived from a sinusoidal model of rotor-coupled mass flow dynamics provides proactive disturbance cancellation, while a Lyapunov-based Model Reference Adaptive Control (MRAC) layer closes the loop by continuously updating model parameters in response to stochastic powder settling dynamics. Real-time experiments on an Optomec LENS 840 with FPGA-based control demonstrate successful parameter drift tracking and a 31–64% reduction in steady-state mass flow standard deviation across a seven-step setpoint sequence compared to open-loop operation.

1:50 PM
Investigating the Effect of Disturbances on the Dimensional Quality of Parts from Directed Energy Deposition(DED): Emmanuel Bamido¹; Michael Cullinan¹; ¹University of Texas at Austin
    Metal based Additive manufacturing processes like Directed Energy Deposition experience limited industrial adaptation due to dimensional inaccuracies of the final part. One source of quality defects is the dynamic changes occurring within the process. In this research, a multiphysics model was developed to analyze the effects of wave-like disturbances on the final dimension of the part. The disturbances at varying frequencies were integrated into the substrate geometry, and two nozzles sprayed metallic powder(Inconel) on the substrate at constant laser power. The final height of the part was measured at different frequencies, and it was observed that the amplitude of the response reduced with increasing frequency.

2:10 PM
Thermal Optimization to Enable SMARTT Toolpaths for Efficient and Effective Multi-Robot DED: Cheng-Hao Chou¹; Chinedum Okwudire¹; ¹University of Michigan
    Robotic directed energy deposition (DED) enables larger work volumes and faster material deposition than other additive manufacturing processes. Its productivity can be further enhanced by coordinating multiple robots. However, DED’s higher energy input causes non-uniform temperature distributions and part distortion, while nonprinting time, including interpass cooling and air travel, compromises the overall throughput. This work proposes a Scalable Multihead Assignment and Routing using Thermal- and Time-optimized (SMARTT) toolpath algorithm for dual-robot DED systems. The framework first develops a graph-based thermal simulation model to approximate the temperature distributions for optimization. Then, toolpath optimization, considering temperature uniformity, print time minimization, and collision avoidance, is applied to optimize the toolpath sequence of the two robotic print heads. Simulation shows that the proposed SMARTT toolpath can achieve superlinear speedup and improve temperature uniformity, compared to a single-robot benchmark, demonstrating its potential to enhance productivity and quality of robotic DED.

2:30 PM
Physics-Preserving Graph Neural Network Surrogates for Thermomechanical Field Prediction: Toward Real-Time Digital Twins in Metal AM: Usman Tariq¹; Frank Liou¹; ¹Missouri University of Science and Technology
    Metal additive manufacturing holds promise for aerospace and defense, yet qualification remains slow due to the high cost of physics-based simulation. Finite element analysis of a directed energy deposition build can take hours to days, making it incompatible with real-time process control. This work develops a two-stage physics-aware graph neural network surrogate that inherits the accuracy of validated multiphysics finite element models at a fraction of their cost. The mesh is converted into a graph with physics-informed node features including laser proximity, boundary distance, and element birth timing. A DeeperGCN thermal model predicts full-field nodal temperature across deposition time steps. A recurrent graph neural network then maps the predicted thermal history to full-field residual stress and displacement, preserving the causal link between thermal and mechanical evolution. The framework generalizes to unseen built geometries without retraining. Ongoing work integrates in-situ sensor data toward real-time digital twin deployment for additive manufacturing qualification.

2:50 PM Break

3:30 PM
Sinusoidal Rosenthal Solutions for Design of AM Meltpool Feedback Control: Aidan Brooks¹; Douglas Bristow¹; ¹Missouri University of Science & Technology
    Meltpool feedback control is critical for achieving consistent deposition geometry in metal AM processes such as laser DED, LPBF, and WAAM, where chaotic process conditions would otherwise produce messy parts and inconsistent builds. This control is often realized by varying the input power and monitoring the solidus isotherm in-situ, with thermal or optical cameras acting as non-contact sensors, as a surrogate for meltpool geometry. However, the resulting closed-loop systems are typically bespoke, lack generalizability, and often depend upon extensive experiential tuning. The well-known Rosenthal point-source solutions provide a bedrock understanding of steady-state meltpool behavior in AM and welding, but do not sufficiently capture isotherm dynamics to bridge from DC behavior to closed-loop design. In this paper, we lay a foundation for such a bridge by extending the Rosenthal solution in 3D to the steady-state sinusoidal-input case and developing analytical transfer function descriptions for isotherm length, width, and size (area).

3:50 PM
Reducing Porosity and Distortion Through Coupled Multi-Scale Control of Laser Powder Bed Fusion: Nicholas Kirschbaum¹; Chinedum Okwudire¹; ¹University of Michigan
    Achieving effective closed-loop process control remains a key challenge in LPBF. Previous work has separately explored scan-vector-level modulation of laser power to control the meltpool area and layer-wise scan sequencing to reduce residual stress and distortion. However, these methods are typically used independently, and limited work has examined how scan sequencing affects porosity. In prior work, we introduced a feedforward power-to-meltpool-area controller using a lightweight thermal model coupled with an analytical meltpool model. This work integrates an improved version of that controller with SmartScan, a layer-wise scan-sequencing algorithm developed at the University of Michigan that minimizes distortion by reordering scan vectors within each layer. Initial testing showed significant increases in porosity when using SmartScan alone, but porosity decreased when coupled with power control, without a significant increase in distortion. This multi-scale control framework reduces distortion while minimizing meltpool variability through feedforward vector-level control.