SEARCH · Search NASA
Results for “Model Optimization”
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.
Alice
SAND2025-00233O Alice is a PyTorch optimizer that helps improve the performance of machine learning models. It focuses on optimizing the process of training these models by using a special technique called the curvature of expectation. This technique allows Alice to make better decisions during the training process, leading to more accurate predictions. By using Alice, researchers and developers can potentially achieve better results and speed up the optimization of machine learning models. It's like having a smarter assistant that helps you train your models more effectively. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.
Bayesian D‐Optimal Designs for Gaussian Process Surrogate Models
Computer experiments often employ space-filling strategies to create surrogate models with strong predictive performance. The impact of model parameter estimation for Gaussian process surrogates, however, is often overlooked. Obtaining a better initial estimate of the covariance lengthscale parameter, θ, can greatly improve the resulting Gaussian process fit through more effective sequential acquisitions during active learning. In this work, we propose a novel initial design maximizing the Bayesian D-optimality criterion of the Gaussian process lengthscale parameter. Previously published results have shown the emphasis on lengthscale estimation to be promising, but relied on an empirically driven design creation process. Our Bayesian D-optimal designs are rooted in information theory and lead to more informative sequential acquisitions by improving lengthscale estimation. In many cases, these gains eventually result in better surrogates than those seeded with space-filling initial designs. Furthermore, Bayesian D-optimal designs can be tailored to either isotropic or anisotropic covariance structures, and the Bayesian framework enables the inclusion of prior knowledge in the design process, offering greater flexibility and adaptability. Through several simulation studies, we demonstrate the advantages of Bayesian D-optimal designs in terms of both lengthscale estimation accuracy and predictive performance during active learning.
Optimizing the Weather Research and Forecasting Model with OpenMP Offload and Codee
Currently, the Weather Research and Forecasting model (WRF) utilizes shared memory (OpenMP) and distributed memory (MPI) parallelisms. To take advantage of GPU resources on the Perlmutter supercomputer at NERSC, we port parts of the computationally expensive routine Fast Spectral Bin Microphysics (FSBM) to NVIDIA GPUs using OpenMP device offloading directives. To facilitate this process, we explore a workflow for optimization which uses both runtime profilers and a static code inspection tool Codee to refactor the subroutine. We observe an 2.24x overall speedup for the CONUS-12km storm test case.
Efficient First-Order Algorithms for Large-Scale, Non-Smooth Maximum Entropy Models with Application to Wildfire Science
Maximum entropy (MaxEnt) models are a class of statistical models that use the maximum entropy principle to estimate probability distributions from data. Due to the size of modern data sets, MaxEnt models need efficient optimization algorithms to scale well for big data applications. State-of-the-art algorithms for MaxEnt models, however, were not originally designed to handle big data sets; these algorithms either rely on technical devices that may yield unreliable numerical results, scale poorly, or require smoothness assumptions that many practical MaxEnt models lack. In this paper, we present novel optimization algorithms that overcome the shortcomings of state-of-the-art algorithms for training large-scale, non-smooth MaxEnt models. Our proposed first-order algorithms leverage the Kullback–Leibler divergence to train large-scale and non-smooth MaxEnt models efficiently. For MaxEnt models with discrete probability distribution of n elements built from samples, each containing m features, the stepsize parameter estimation and iterations in our algorithms scale on the order of O(mn) operations and can be trivially parallelized. Moreover, the strong ℓ1 convexity of the Kullback–Leibler divergence allows for larger stepsize parameters, thereby speeding up the convergence rate of our algorithms. To illustrate the efficiency of our novel algorithms, we consider the problem of estimating probabilities of fire occurrences as a function of ecological features in the Western US MTBS-Interagency wildfire data set. Our numerical results show that our algorithms outperform the state of the art by one order of magnitude and yield results that agree with physical models of wildfire occurrence and previous statistical analyses of wildfire drivers.
Design and Optimization of Processes for Recovering Rare Earth Elements from End-of-Life Hard Disk Drives
In this poster, we first provide motivation for why rare earth elements as rare earth permanent magnets (REPM) are increasing in demand. We then highlight some of the recent work that has been done by several national labs (National Renewable Energy Laboratory (NREL), Environmental Protection Agency (EPA), Critical Minerals Institute (CMI)) on recycling rare earth elements from end-of-life hard disk drives (EOL). Then, we mention our long-term plan to design a feedstock agnostic process to recover rare earth elements as rare earth oxides from many different EOL products at once. Next, we discuss how we quantified the rare earth elements available for recycling from EOL hard disk drives from consumer desktops and laptops. We then discuss how we used superstructure optimization to design the optimal pathway. The proposed superstructure was modeled as a MILP optimization problem, selecting the net present value as the objective function. Costing data from the literature was used to inform this model whenever possible. However, due to the novelty of this research area, data were often unavailable, thus requiring the generation of flowsheets implemented in Aspen Plus.
Optimization of a Secondary Air Injector for a Rich-Quench-Lean (RQL) Ammonia Combustor Using Computational Fluid Dynamics
Ammonia has emerged as a promising medium for moving hydrogen around the globe due to its energy density, existing infrastructure for production, transportation, and storage, and multiple applications from fertilizer to power generation. However, one of the most significant challenges with ammonia combustion for power generation is the formation of nitrogen oxides (NOx) during combustion. Several combustion strategies have been developed to minimize the formation of NOx. One such strategy, known as rich-quench-lean (RQL), is a method that combusts NH3 in a fuel-rich environment, followed by a quick mix section and a lean burnout section. Rapid mixing of secondary air before lean burnout is thought to be important to minimize the formation of NOx. A genetic algorithm (GA) is used to parametrically vary the secondary air injector design, including the diameter, count, and angles over a constrained design space. The designs are evaluated using a non-reacting OpenFOAM model, with the objective function being the uniformity index of the secondary air (modeled as a scalar). Optimization campaigns show that traditional correlation-based designs might not be adequate. After running 414 OpenFOAM models, the optimal design is 32, 1 mm diameter, angled air injectors, increasing the uniformity index at the outlet of the quick mix zone by 36.8%. The most promising designs will be manufactured and tested in a small RQL burner setup, which is expected to lead to validation and insight into the practical usage of NH3 combustion for power generation.
Incorporating corrosion design constraints in desalination process optimization: A case study in mechanical vapor compression
Corrosion is an expensive and complex challenge for desalination, yet current design approaches do not explicitly account for corrosion mechanisms in process modeling and technoeconomic analysis. Here, to address this gap, we present a workflow for incorporating corrosion design constraints directly into desalination process optimization models. We develop surrogate models for general and localized corrosion metrics as functions of temperature, pH, salinity, dissolved oxygen, and material using data from OLI Systems’ Corrosion Analyzer. We then integrate these surrogates as corrosion design constraints in a cost-optimization MVC model that minimizes the levelized cost of water (LCOW). For a case study of mechanical vapor compression (MVC) treating seawater across a range of recoveries, we find dissolved oxygen (DO) is the dominant driver of localized corrosion, and thus of cost-optimal material choice and operating conditions. Reducing the DO from 8 mg/L to 0.5 mg/L reduces the LCOW by 15-35%, informing the breakeven costs for implementing DO removal or selecting highly corrosion-resistant alloys. This framework is broadly applicable across corrosion types, materials, and components and enables desalination process design that minimizes capital costs.
Conceptual Design of Integrated Energy Systems with Market Interaction Surrogate Models
Most integrated energy system (IES) optimization frameworks employ the price-taker approximation, which ignores important interactions with market and can result in overestimated economic values. In this work, we pro-pose a machine learning surrogate-assisted optimization framework to quantify the IES/market interactions and thus go beyond price taker. We use time series clustering to generate representative IES operation profiles for the IES optimization problem and use machine learning surrogate models to predict the IES/market interaction. We quantify the accuracy of the time series clustering and surrogate models in a case study to optimally retrofit a nuclear power plant with polymer electrolyte membrane electrolyzer to co-produce electricity and hydrogen.
Conceptual Design of Integrated Energy Systems with Market Interaction Surrogate Models
Most integrated energy system (IES) optimization frameworks employ the price-taker approximation, which ignores important interactions with the market and can result in overestimated economic values. In this work, we propose a machine learning surrogate-assisted optimization framework to quantify IES/market interactions and thus go beyond price-taker. We use time series clustering to generate representative IES operation profiles for the optimization problem and use machine learning surrogate models to predict the IES/market interaction. We quantify the accuracy of the time series clustering and surrogate models in a case study to optimally retrofit a nuclear power plant with a polymer electrolyte membrane electrolyzer to co-produce electricity and hydrogen.
Optimization of direct air capture processes using reactive transport models of adsorption-desorption cycles
In this study, we develop and implement a reactive transport model in COMSOL Multiphysics® to address the challenges of direct air carbon capture. The model is validated against experimental data and used to simulate the cyclic steady state of the adsorption-desorption process. The optimization of this model is achieved through advanced trust-region methods integrated with Gaussian Processes. Key decision variables, including adsorption and desorption times, desorption temperature and pressure, input velocity, bed porosity, column length, and radius were optimized to minimize the capture cost. After optimization, a sensitivity analysis revealed the complex interplay between the decision variables and their effect on the specific energy and cost of removing the CO 2 . We optimized the capture cost while taking into account the trade-off between energy consumption and productivity. The resulting minimum capture cost was determined to be 265.2 $/t-CO 2 , which aligns with expected values reported in the literature. Numerical results suggest the effectiveness of the optimization strategies applied, and underscore the importance of simultaneous decision variable selection in improving the performance in direct air capture processes. We also extend the modeling approach to a 2D axisymmetric model to better visualize CO₂ uptake and temperature profiles, revealing significant radial gradients during the regeneration step. As a main drawback, this enhanced model comes with a computational cost approximately 40 times higher than that of the 1D model.
Inverse design of hypoeutectoid pearlite steel microstructures using a deep learning and genetic algorithm optimization framework
Goal-oriented microstructure design in metallic materials is a challenging task due to complex structure-property relationships. Traditional experimental and computational approaches are time-intensive and economically inefficient, limiting their applicability for large-scale design space exploration. Here, in this work, we propose an end-to-end framework that integrates deep learning models with genetic optimization to design microstructures with targeted mechanical properties. Deep learning models enable accurate forward design, while their integration with genetic optimization enables efficient inverse design within a few hours, compared to days or weeks using conventional finite element simulations. The framework combines experimental characterization and finite element modeling to analyze the influence of microstructural features on the mechanical behavior of hypoeutectoid steels. Data from both experiments and simulations are used to train the deep learning models. To demonstrate its effectiveness, we apply the framework to 0.63% carbon steel with proeutectoid ferrite and pearlite phases, commonly used in industrial applications. In this study, 2D microstructures were used for modeling, selected primarily for computational efficiency and to establish proof of concept. The framework successfully optimizes microstructures for targeted yield strength, ultimate strength, and stress concentration factors while significantly reducing computational time. Beyond hypoeutectoid steels, this scalable framework can be extended to other material systems and integrated with additive manufacturing, offering an efficient approach for accelerating microstructure design for specific engineering applications.
Virtual Engineering: Python framework for engineering process design
Virtual Engineering (VE) is a Python software framework designed to accelerate the research and development of engineering processes that are fundamentally defined by multiple unit operations executed in series. VE supports a wide variety of different multi-physics models and integrates them to simulate a complete end-to-end process. To automate the execution of this model sequence, VE provides (i) a robust method to communicate between models, (ii) a high-level, user-friendly interface to set model parameters and enable optimization, and (iii) an overall model-agnostic approach that allows new computational units to be swapped in and out of workflows. Although the VE framework was developed to support the biochemical conversion of biomass to fuel, we have designed each component to easily accommodate new domains and unit models.
Optimal Design of Intensified Towers for CO2 Capture with Internal, Printed Heat Exchangers
This poster discusses the modeling, performance and optimization of a solvent absorption system for CO2 capture with towers utilizing intensified packing. The packing is an alternative to traditional structured packing by incorporating cooling channel for simultaneous mass and heat transfer.
Machine learning framework for predicting uranium enrichments from M400 CZT gamma spectra
A machine learning framework was developed for predicting uranium enrichments from M400 CZT gamma spectra. This framework leverages the availability of a large amount of measured M400 gamma spectra and uses a recently updated version of Gamma Detector Response and Analysis Software (GADRAS) for gamma spectrum analysis and generation. It also leverages the existing machine learning modules in Python for gamma spectrum data processing, curation, model training, benchmarking, and optimization of the deep machine learning models. The framework is used to develop a deep learning model to analyze gamma spectra from a set of U 3 O 8 samples with enrichments ranging from 0.31 to 93.17% and UF 6 cylinders with enrichments ranging from 0.2 to 4.95%, and the model performance is tested using a set of measured spectra and the respective declared enrichment values. Results show that the model can correctly classify 99.35% of the U 3 O 8 sample enrichments, and can predict the samples’ enrichments within an average absolute error of 0.099% (in percentage points of enrichment). For the UF 6 cylinders, the average absolute error was approximately 0.03%, with an accuracy of 98% in classifying discrete enrichment values of UF 6 samples. Finally, the results also show that the model has performed significantly better in terms of predicting enrichments in UF 6 cylinders based on measured gamma spectra than the GEM code, with a standard deviation (of the relative errors) of 2.23% (compared with the 11.51% value for the GEM code) based on results from a set of test data.
Geomatchd
Geomatchd is a software tool to assist non-experts in the rapid conceptual design of geothermal district heating systems. This tool comprises three main parts: user-side heating and cooling demand profiles, district heating and cooling (DHC) network models for sizing and optimal layout selection, and geothermal system models for heat extraction calculation and cost comparison. The architecture of Geomatchd is opensource and is intended to simplify efforts by other coders to improve on existing functions, to add functions and to leverage available existing code and data – to improve the tool for public use. Geomatchd and the material supporting its development may provide a more intuitive window on district heating and the use of local shallow geothermal resources.
An optimization-based coupling of reduced order models with efficient reduced adjoint basis generation approach
Explore the source record for details and available documents.
Bioreactor Optimization through Multi-Phase Flow Models (CRADA Final Report)
Chemical manufacturing uses 29% of energy in the United States and produces 925 million metric tons of CO2 annually. Biomanufacturing offers the potential to leverage America’s rich agricultural resources to produce critical chemicals such as lubricants, pharmaceutical precursors, and components of energetic materials that today are sourced extensively from overseas. The bioreactors used in biomanufacturing applications, such as one developed by Capra Biosciences, involve multiphase flow of biofilm-coated solid support particles that are continuously circulated in a fluidized state within the reactor along with a constant supply of oxygen via an aeration mechanism. In this project, Capra Biosciences and LBNL developed a multiscale modeling framework to simulate the multiphase flows of solid particles in a liquid-gas bubble mixture that occurs in the bioreactor using the current MFIX-Exa software, an opensource multiphase flow solver developed and maintained at LBNL and NETL. By leveraging HPC capabilities, this high-fidelity multiscale model was used to inform design decisions for bioreactor architecture to make them operationally efficient.