Katy Huff

I'm thrilled to share a new paper from Illinois NPRE PhD student Luke Seifert! It tackles a long-standing question in molten salt reactor depletion analysis. When do we truly need spatial resolution in flowing-fuel systems, and when does the scaled flux method give an accurate answer at a fraction of the computational cost? https://lnkd.in/eyus9c2a Luke compared two bounding cases for MSR depletion. One assumes perfect mixing through the primary loop while the other applies a one-dimensional advection model that assumes no mixing at all. He showed that the scaled flux method can match the spatially resolved model under most conditions. He also identified the conditions that break that agreement, including large ex-core removal rates, very slow flow, or strong power oscillations. I am particularly delighted by the technical detail in his work. It sharpens our confidence in modeling approaches for molten salt reactor safety, fuel management, and operational licensing! As always, we're grateful for the fabulous collaborators who contributed in essential ways. Madicken Munk, Shayan Shahbazi, and Scott Richards shaped this study through thoughtful discussions and rigorous review, making the findings stronger at every stage. University of Illinois Urbana-Champaign, The Grainger College of Engineering, Nuclear, Plasma & Radiological Engineering at Illinois, Argonne National Laboratory, Oregon State University

151

1 Comment

Graduate student Noah South's perspective about modeling hydrogen atom transfer (HAT) rate constants in heterogeneous catalysis was recently published in ACS Catalysis! Hydrogen atoms consist of an electron and proton, and quantum effects of each of these subatomic particles can impact reactivity. This perspective discusses various classical, semiclassical, and quantum-mechanical treatments of hydrogen and associated rate constant models in the context of heterogeneous catalysis. https://lnkd.in/gz5iQiYD

106

1 Comment

Thrilled to share a revised, updated version of our paper "A hybrid deep learning approach for annular flow regime identification in 25.4 mm diameter round pipe geometry" is now posted on open-access SSRN. Our main findings include: 🌊 New high‑void flow regimes experimental dataset: - 460 time‑resolved conductivity signals over 52 conditions in air-water flows in a 25.4 mm ID vertical pipe . - Droplet-capable conductivity probe signals resolve interfacial structures in annular, rolling‑annular, and wispy‑annular regimes.​ 🧠 Hybrid CNN–BiLSTM for three annular sub‑regimes: - A custom 1D Convolutional NN extracts local temporal features from the DCCP‑2 voltage signals, while a BiLSTM learns the long‑range dynamics of the interface. - A softmax layer outputs probabilistic labels for three annular classes.​ 📈 Promising and balanced performance: The model reaches 87.2% validation accuracy with high, balanced precision/recall and F1 across all three regimes, with most misclassifications occurring only between neighboring annular states (which is exactly where the physics is fuzzy).​ ✅ Beats conventional ML baselines: Compared to linear regression, random forest, and k‑NN, the CNN–BiLSTM significantly reduces large misclassification errors (lower RMSE), showing the value of leveraging temporal structure instead of static features alone.​ 🧩 From static maps to probabilistic, time‑resolved regimes: Instead of relying on hard boundaries in flow‑regime maps, we use the network’s probability outputs to quantify ambiguity in transitional states—directly relevant for CHF/dryout prediction, uncertainty reduction, and online diagnostics in BWRs and advanced multiphase systems.​ If you’re working on two‑phase flow, CHF, or ML for regime identification, we'd love to hear your thoughts and critiques. #twophaseflow #thermalhydraulics #annularflow #deepLearning #CNN #LSTM #multiphaseflow https://lnkd.in/gN9ZGrm6

26

1 Comment

🔬 Open-Source Nuclear Physics Simulation Packages from LLNL ☢️ Lawrence Livermore National Laboratory (LLNL) has released three powerful simulation tools capable of modeling various physical phenomena and integrating with Monte Carlo transport codes such as MCNP, Geant4, and others: CRY 🌌 Simulates cosmic ray particle showers at three different altitudes. Produces muons, neutrons, protons, electrons, photons, and pions. 🔗 Use case: Serves as an input generator for detector simulations and transport codes. Downloads(version 1.7) : https://lnkd.in/gRJ56-Q2 RadSrc ☢️ Computes gamma spectra from mixtures of radioisotopes. 🧪 Considers isotope concentrations aged through Bateman decay chain solutions. Downloads(version 1.6) : https://lnkd.in/gT8fkgwb Fission (with FREYA) ⚛️ Simulates event-by-event fission of heavy nuclei including neutron and gamma emissions. 📈 Provides accurate outputs for isotopes like U-233, U-235, and Pu-239. Downloads(from source) :https://lnkd.in/gv4rA4f7 🛠️ All tools can be used as callable libraries from C, C++, or Fortran. 📎 More info and download links: LLNL Physics Simulation Packages⤵️ https://lnkd.in/gpedw6yM Credit Link:https://lnkd.in/g5sSKUbC #nuclearphysics #simulation #MonteCarlo #LLNL #MCNP #Geant4 #CRY #RadSrc #FREYA #fission #radiation

125

10 Comments

I’m thrilled to share two incredible new publications from University of Michigan College of Engineering students in my research group! Their work perfectly highlights how bridging Physics, Materials Science, Nuclear Engineering and Radiological Sciences—University of Michigan, and computational methods can advance nuclear energy. 🧠 Alex Birmingham (Undergrad, LSA Physics) created a new computational tool to analyze uncertainties in predicting how reactor materials degrade over time. Read it here: https://lnkd.in/gy-MQC9M The Impact: Alex proved that minor chemical impurities and computational methods drastically affect predicted helium generation rates, establishing critical new guidelines for predicting and reporting helium generation rates of nuclear materials in fission and fusion systems. 🔬 Txai Sibley (Grad Student, Materials Science) adapted Meta’s Segment Anything Model (SAM) for niche nuclear microscopy. Read it here: https://lnkd.in/gK6XdWQc The Impact: Txai developed domain-informed filters to overcome severe data shortages, unlocking zero-shot AI foundation models for complex microstructural analysis. This synergy of physics-driven modeling and AI-enhanced materials characterization is exactly what we need to build the next generation of resilient nuclear systems. Huge congrats to Alex and Txai! 👏 #MaterialsScience #Physics #NuclearEngineering #MachineLearning #StudentSuccess

57

https://lnkd.in/eeBSA3N6 Nuclear physics, with its current $866 million budget, might feel the brunt of any cuts, frets Rosi Reed, a nuclear physicist at Lehigh University. “I worry that one community or the other—and most likely it would be nuclear physics, since we’re smaller—would get sort of overshadowed.”

6

2 Comments