SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Code for plots, models, data generation and other utilities relating to the paper "wa-hls4ml: A Benchmark and Surrogate Models for hls4ml Resource and Latency Estimation" https://arxiv.org/abs/2511.05615 [FERMILAB-PUB-25-0359-CSAID]
The focus of this project was to observe and simulate the solidification of uranium metal at the crystallographic level from its molten state. Melting experiments were conducted at two different scales to observe microstructural evolution using either a laboratory-scale induction furnace (hundreds of grams of metal) or a microscope heating stage (hundreds of milligrams of metal), respectively. Experimental parameters and characterization data were then used to inform a phase field model of gamma-U crystal growth as dendrites with or without secondary phase impurities in the form of uranium carbide particles. Finally, training datasets were generated by the phase field model as inputs to a neural network, developed with the aim of providing a faster, cheaper surrogate model for microstructural simulations within a given parameter space. Progress is reported herein for each of these task areas. Ultimately, 1) an optical microscope heating stage capability has been stood-up for uranium metal solidification studies, 2) a phase field model was advanced to simulate multiple uranium grains growing in the presence of carbide impurity particles and 3) a neural network was constructed and optimized to predict the microstructure features of individually growing uranium crystals.
This paper describes new modeling capabilities for predicting H-mode pedestal profiles in spherical tokamaks. Temperature profiles for NSTX discharges 132543 and 132588 are modeled by coupling the \textsc{astra} transport solver with neoclassical transport and gyrokinetic-based reduced models for electron temperature gradient (ETG) and kinetic ballooning mode (KBM) instabilities. A quasi-linear surrogate model for ion-scale transport is developed using linear \textsc{gene} simulations, requiring only a single free parameter calibrated to one discharge. Time-evolving the temperatures with fixed density yields good agreement with experiments for both discharges. Systematic analysis of the transport mechanisms reveals that neoclassical transport is huge across the entire pedestal region for the ion channel. ETG turbulence is large in the plasma edge and low density gradient region, contributing substantially to the electron channel. However, KBM/MHD-like modes also drive significant transport in both the ion and electron thermal channels, making them essential for accurate pedestal modeling. Further refinements, including explicit E×B shear suppression and scaled ETG transport, produce quantitative but not qualitative improvements. This work lays the foundation for predictive modeling of future devices. This paper is on arxiv and has been submitted to Nuclear Fusion.
To establish a technical basis for self-regulating microreactors, a model predictive control (MPC) system is investigated to proactively respond to anomalies and disturbances in anticipation of potential deviations from operating setpoints. Due to the difficulty of developing a physics-based surrogate model that can accurately match plant data in various operating conditions, machine learning algorithms are used in MPC, which allow for learning from both simulation and operation data, thus efficiently describing the targeted transient with arbitrary accuracy. However, one of the biggest concerns in applying ML algorithms like artificial neural networks (ANNs) is that the predictive capabilities of ANN are limited by training data. If there are gaps between the training and target domain, the accuracy of an ANN can degrade significantly when it is used to predict unseen data. To improve the predictive capability of ANN and enable a confident use of data-driven MPCs outside the training data, this study proposes an adaptive data-driven MPC framework. The system will monitor the discrepancy between plant responses and surrogate predictions, fine-tune the ANN-based surrogate when a large discrepancy is detected, and continue MPC operation with updated surrogates. The framework is demonstrated on a point kinetic model for microreactors. The hyperparameters of the update strategy, including layers to update, error thresholds, learning rate discount, and number of data points used for fine-tuning, are optimized so the simulated microreactor is able to follow changes in setpoint with the smallest of deviations.
To establish a technical basis for self-regulating microreactors, a model predictive control (MPC) system is investigated to proactively respond to anomalies and disturbances in anticipation of potential deviations from operating setpoints. Due to the difficulty of developing a physics-based surrogate model that can accurately match plant data in various operating conditions, machine learning algorithms are used in MPC, which allow for learning from both simulation and operation data, thus efficiently describing the targeted transient with arbitrary accuracy. However, one of the biggest concerns in applying ML algorithms like artificial neural networks (ANNs) is that the predictive capabilities of ANN are limited by training data. If there are gaps between the training and target domain, the accuracy of an ANN can degrade significantly when it is used to predict unseen data. To improve the predictive capability of ANN and enable a confident use of data-driven MPCs outside the training data, this study proposes an adaptive data-driven MPC framework. The system will monitor the discrepancy between plant responses and surrogate predictions, fine-tune the ANN-based surrogate when a large discrepancy is detected, and continue MPC operation with updated surrogates. The framework is demonstrated on a point kinetic model for microreactors. The hyperparameters of the update strategy, including layers to update, error thresholds, learning rate discount, and number of data points used for fine-tuning, are optimized so the simulated microreactor is able to follow changes in setpoint with the smallest of deviations.
The use of machine learning (ML)-based surrogate models is a promising technique to significantly accelerate simulation-driven design optimization of internal combustion (IC) engines, due to the high computational cost of running computational fluid dynamics (CFD) simulations. However, training the ML models requires hyperparameter selection, which is often done using trial-and-error and domain expertise. Another challenge is that the data required to train these models are often unknown a priori. In this work, we present an automated hyperparameter selection technique coupled with an active learning approach to address these challenges. The technique presented in this study involves the use of a Bayesian approach to optimize the hyperparameters of the base learners that make up a super learner model. In addition to performing hyperparameter optimization (HPO), an active learning approach is employed, where the process of data generation using simulations, ML training, and surrogate optimization is performed repeatedly to refine the solution in the vicinity of the predicted optimum. The proposed approach is applied to the optimization of a compression ignition engine with control parameters relating to fuel injection, in-cylinder flow, and thermodynamic conditions. It is demonstrated that by automatically selecting the best values of the hyperparameters, a 1.6% improvement in merit value is obtained, compared to an improvement of 1.0% with default hyperparameters. Overall, the framework introduced in this study reduces the need for technical expertise in training ML models for optimization while also reducing the number of simulations needed for performing surrogate-based design optimization.
Here, we developed a PyTorch-based architecture called HydraGNN that implements graph convolutional neural networks (GCNNs) to predict the formation energy and the bulk modulus for models of solid solution alloys for various atomic crystal structures and relaxed volumes. We trained the GCNN surrogate model on a dataset for nickel–niobium (NiNb) generated by the embedded atom model (EAM) empirical interatomic potential for demonstration purposes. The dataset was generated by calculating the formation energy and the bulk modulus as a prototypical elastic property for optimized geometries starting from initial body-centered cubic (BCC), face-centered cubic (FCC), and hexagonal compact packed (HCP) crystal structures, with configurations spanning the possible compositional range for each of the three types of initial crystal structures. Numerical results show that the GCNN model effectively predicts both the formation energy and the bulk modulus as function of the optimized crystal structure, relaxed volume, and configurational entropy of the model structures for solid solution alloys.
A potential technology for the CO 2 absorption process is utilizing intensified structured packing with embedded cooling/heating channels for continuous heat exchange, which can overcome limitations of discrete methods, such as discrete intercooling and centralized reboilers, to aid in reducing energy consumption and decreasing costs. This work investigates the modeling of intensified packing (IP) for the stripper tower, extending on previous work for the absorber, which distributes heat internally within the column, improving the thermodynamics for the solvent regeneration process. The model includes submodels for steam turbine extraction to produce steam at various qualities as well as a surrogate model for calculating steam enthalpy. A cost model for a plant-scale absorption capture process was developed, allowing for the design of the plant to be optimized, subject to minimizing capture cost using two different power plant flue gas sources. In this optimization, the placement of IP in both towers is optimized to balance the trade-off between enhanced heat transfer and reduced mass transfer volume. For natural gas combined cycle flue gas, the standard process configuration had a minimum cost of $\$$65.40/tonne CO 2 , and considering IP, the minimum capture cost is reduced to $\$$62.73/tonne, with utilization in the stripper column, which reduces yearly costs by up to $\$$2.67 MM/yr. Cooling the absorber through IP, or intercoolers, was only found to be beneficial at higher capture rates, with IP in both towers having a cost of capture of $\$$68.08/tonne at 99.9% capture, a reduction of $\$$12.64/tonne when using only intercoolers at the same capture rate. When capturing from pulverized-coal power plants, the minimum cost of capture when using IP in both towers is $\$$44.18/tonne (at 97% capture), while the standard configuration with and without intercoolers was $\$$45.69 and $\$$47.22 per tonne, respectively. This results in a reduction in yearly costs of $\$$16.98 MM/yr from the base-case configuration. At this higher CO 2 concentration, cooling in the absorber from the IP becomes extremely beneficial, reducing energy consumption by up to 6%.
The Organization for Economic Cooperation and Development (OECD) Working Party on Nucelar Criticality Safety (WPNCS) has proposed a benchmark exercise representative of neutronic behavior in criticality experiments. Here, the goal is to develop confidence in data assimilation techniques used to adjust nuclear data. Participants are given synthetic experimental models with associated measured data and asked to estimate the model parameters given the model and measurements as well as provide predictions for separate application models. In this work, we performed data assimilation using Bayesian inverse Uncertainty Quantification (UQ) with machine learning surrogate models to produce posterior parameter distributions for the requested parameters and posterior predictive distributions for the requested responses. Several experimental models are shown to insufficiently inform the posterior parameter distributions for the applications involved. However, given sufficient experimental data, posterior parameter estimates yielded reduced uncertainty in the response predictions of interest while covering the experimental data.
In this work, we present the development of an overarching software framework and supporting multiphysics models to simulate the end-to-end process of biomass conversion. This virtual engineering (VE) software is designed with the goal of accelerating research and development and reducing risk for market-relevant biomass conversion processes. We currently support multiple models, computing paradigms, and fidelities representing the steps of feedstock pretreatment, enzymatic hydrolysis, and bioconversion. Although this VE approach was developed to support a biomass workflow, we have designed each component in a way that allows us to easily support new domains, unit models, and feedstocks. We begin by presenting the user-facing aspects of the VE software and highlight how simulated elements are defined and linked by VE functions. We then present an overview of the high-fidelity computational fluid dynamics (CFD) models developed to support our target domain before segueing into our efforts to develop accurate and fast surrogate models, capturing the salient outcomes from the CFD simulations in a significantly less computationally demanding manner. We then present how VE calculations interface with a commercial techno-economic analysis software, Aspen Plus. We conclude by presenting VE case studies that leverage these methods and discuss how our methods can be extended to support a wide variety of accelerated biofuel commercialization pathways in the future.
Benders decomposition is a widely used method for solving large and structured optimization problems, but its performance is affected by the repeated solution of subproblems. We propose a flexible and modular algorithmic framework for accelerating Benders decomposition. Specifically, we express the problem structure by using a graph-theoretic modeling abstraction in which nodes represent optimization subproblems and edges represent connectivity between subproblems. A key innovation of our approach is that we embed multiparametric programming (mp) surrogates for node subproblems, which maps the exact analytical map of the subproblem solution space. The use of mp surrogates allows us to replace subproblem solves with fast look-ups and function evaluations for primal and dual variables during the iterative Benders process. We formally show the equivalence between classical Benders cuts and those derived from the mp solution. We implement our framework in the open-source PlasmoBenders.jl software package. To demonstrate the capabilities of the proposed framework, we apply it to a two-stage stochastic programming problem, which aims to make optimal capacity expansion decisions under market uncertainty. We evaluate both single-cut and multicut variants of Benders decomposition and show that the use of mp surrogates achieves substantial speedups in subproblem solve time, while preserving the convergence guarantees of Benders decomposition. We highlight advantages in solution analysis and interpretability that is enabled by mp critical region tracking; specifically, we show that these reveal how decisions evolve geometrically across the Benders search. Our results aim to demonstrate that combining surrogate modeling with graph modeling offers a promising and extensible foundation for structure-exploiting decomposition. In addition, by decomposing the problem into more tractable subproblems, the proposed approach also aims to overcome scalability issues of mp. Finally, the use of mp surrogates provides a unifying and modular optimization framework that enables the representation of heterogeneous node subproblems as modeling objects with a homogeneous structure.
The neural ordinary differential equation (ODE) framework has shown considerable promise in recent years in developing highly accelerated surrogate models for complex physical systems characterized by partial differential equations (PDEs). For PDE-based systems, state-of-the-art neural ODE strategies leverage a two-step procedure to achieve this acceleration: a nonlinear dimensionality reduction step provided by an autoencoder, and a time integration step provided by a neural-network based model for the resultant latent space dynamics (the neural ODE). This work explores the applicability of such autoencoder-based neural ODE strategies for PDEs in which advection terms play a critical role. More specifically, alongside predictive demonstrations, physical insight into the sources of model acceleration (i.e., how the neural ODE achieves its acceleration) is the scope of the current study. Such investigations are performed by quantifying the effects of both autoencoder and neural ODE components on latent system time-scales using eigenvalue analysis of dynamical system Jacobians. To this end, the sensitivity of various critical training parameters – de-coupled versus end-to-end training, latent space dimensionality, and the role of training trajectory length, for example – to both model accuracy and the discovered latent system timescales is quantified. Furthermore, this work specifically uncovers the key role played by the training trajectory length (the number of rollout steps in the loss function during training) on the latent system timescales: larger trajectory lengths correlate with an increase in limiting neural ODE time-scales, and optimal neural ODEs are found to recover the largest time-scales of the full-order (ground-truth) system. Demonstrations are performed across fundamentally different unsteady fluid dynamics configurations influenced by advection: (1) the Kuramoto–Sivashinsky equations (2) Hydrogen-Air channel detonations (the compressible reacting Navier–Stokes equations with detailed chemistry), and (3) 2D Atmospheric flow.
Heterogeneous Differential Privacy (HDP) in Federated Learning (FL) allows clients to select individual privacy budgets () according to institutional policies and data sensitivity. In practice, many HDP-FL systems employ -aware server aggregation to improve model utility by re-weighting client updates according to their declared privacy budgets. However, gradient updates in FL retain structural patterns induced by non-independent and identically-distributed (non-IID) data, and these additional signals exposed by -aware aggregation create new opportunities for inference by an honest-but-curious server. In this work, we first show that a server equipped with gradient denoising and surrogate modeling can mount a Privacy Inference Attack that infers distributional attributes of clients and links updates from the same client across training rounds, measured via surrogate inference accuracy and linkage success, under realistic knowledge constraints. The Shuffle-Model has been widely studied as a defense against such inference risks by anonymizing update sources, but it is fundamentally incompatible with HDP-FL -aware aggregation. To address this challenge, we propose IntraShuffler, a middleware defense framework designed for HDP-FL systems. IntraShuffler introduces a privacy-aware shuffling mechanism that groups clients into privacy-compatible buckets and performs parameter-level shuffling within each bucket to disrupt persistent gradient structure while preserving -aware aggregation. Experiments across four different datasets show that IntraShuffler reduces gradient recoverability by over 60% and decreases surrogate inference accuracy from 0.78 to 0.33 while maintaining comparable model utility across multiple FL aggregation rules.
Agent-based models (ABM) provide an excellent framework for modeling outbreaks and interventions in epidemiology by explicitly accounting for diverse individual interactions and environments. However, these models are usually stochastic and highly parametrized, requiring precise calibration for predictive performance. When considering realistic numbers of agents and properly accounting for stochasticity, this high-dimensional calibration can be computationally prohibitive. This paper presents a random forest-based surrogate modeling technique to accelerate the evaluation of ABMs and demonstrates its use to calibrate an epidemiological ABM named CityCOVID via Markov chain Monte Carlo (MCMC). The technique is first outlined in the context of CityCOVID's quantities of interest, namely hospitalizations and deaths, by exploring dimensionality reduction via temporal decomposition with principal component analysis (PCA) and via sensitivity analysis. The calibration problem is then presented, and samples are generated to best match COVID-19 hospitalization and death numbers in Chicago from March to June in 2020. Further, these results are compared with previous approximate Bayesian calibration (IMABC) results, and their predictive performance is analyzed, showing improved performance with a reduction in computation.
Abstract The work function is the key surface property that determines the energy required to extract an electron from the surface of a material. This property is crucial for thermionic energy conversion, band alignment in heterostructures, and electron emission devices. This work presents a high‐throughput workflow using density functional theory (DFT) to calculate the work function and cleavage energy of 33,631 slabs (58,332 work functions) that are created from 3,716 bulk materials. The number of calculated surface properties surpasses the previously largest database by a factor of ≈27. Several surfaces with an ultra‐low (<2 eV) and ultra‐high (>7 eV) work function are identified. Specifically, the (100)‐Ba‐O surface of BaMoO 3 and the (001)‐F surface of Ag 2 F have record‐low (1.25 eV) and record‐high (9.06 eV) steady‐state work functions. Based on this database a physics‐based approach to featurize surfaces is utilized to predict the work function. The random forest model achieves a test mean absolute error (MAE) of 0.09 eV, comparable to the accuracy of DFT. This surrogate model enables rapid predictions of the work function (≈ 10 5 faster than DFT) across a vast chemical space and facilitates the discovery of material surfaces with extreme work functions for energy conversion and electronic device applications.
Predicting the structural properties of water and simple fluids confined in nanometer scale pores and channels is essential in, for example, energy storage and biomolecular systems. Classical continuum theories fail to accurately capture the interfacial structure of fluids. In this work, we develop a deep learning-based quasi-continuum theory (DL-QT) to predict the concentration and potential profiles of a Lennard-Jones (LJ) fluid and water confined in a nanochannel. The deep learning model is built based on a convolutional encoder–decoder network (CED) and is applied for high-dimensional surrogate modeling to relate the fluid properties to the fluid–fluid potential. The CED model is then combined with the interatomic potential-based continuum theory to determine the concentration profiles of a confined LJ fluid and confined water. Further, we show that the DL-QT model exhibits robust predictive performance for a confined LJ fluid under various thermodynamic states and for water confined in a nanochannel of different widths. The DL-QT model seamlessly connects molecular physics at the nanoscale with continuum theory by using a deep learning model.
To support experiments modeling and transient analysis of the NASA-sponsored SIRIUS experiments in the Transient Reactor Test Facility (TREAT), an innovative predictive transient model has been developed to simulate various transient experiments. The developed model is based on a new approach that utilizes steady-state Monte Carlo calculations along with a surrogate model based on polynomial regression to determine the core reactivity. The reactivity is supplied to an exact point kinetics model that determines the reactor power, and it is coupled to an adiabatic feedback model to determine the fuel average temperature. In this work, the newly developed model is introduced along with validation test results considering SIRIUS-1 experiments at different power levels. The initial test results show a very good agreement with experimental data at different power levels, which indicates that the model can be used to predict the power of other transient tests.
In this paper, we present our ongoing work aimed at accelerating a particle-resolved direct numerical simulation model designed to study aerosol–cloud–turbulence interactions. The dynamical model consists of two main components—a set of fluid dynamics equations for air velocity, temperature, and humidity, coupled with a set of equations for particle (i.e., cloud droplet) tracing. Rather than attempting to replace the original numerical solution method in its entirety with a machine learning (ML) method, we consider developing a hybrid approach. We exploit the potential of neural operator learning to yield fast and accurate surrogate models and, in this study, develop such surrogates for the velocity and vorticity fields. We discuss results from numerical experiments designed to assess the performance of ML architectures under consideration as well as their suitability for capturing the behavior of relevant dynamical systems.