SEARCH · Engineering Papers

Results for “machine learning 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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 775 records · Page 43

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Machine learning-aided inverse design for biogas upgrading through biological CO 2 conversion

The biogas upgrading process through bioconversion of CO 2 to CH 4 by hydrogenotrophic methanogens is an attractive strategy for energy decarbonation. Many studies have optimized operational parameters to improve key performance indicators such as CH 4 % and H 2 utilization efficiency. However, inconsistent laboratory conditions make it challenging to compare results. Existing models for analyzing operating conditions can only assess the impact of individual conditions and lack the ability to simultaneously optimize multiple conditions. To address this, two XGBoost models were built with R 2 of 0.779 and 0.903 with data collected from literatures and were embedded into multi-objective partitive swarm optimization algorithm to optimal operating conditions. Predictions were compared with experimental validations under optimized conditions, revealing an 8.50% and 2.95% relative error in CH 4 % and H 2 conversion rate, respectively. This approach streamlines biogas upgrading processes, offering a data-driven solution to enhance efficiency and consistency in the pursuit of sustainable methane production.

Biogas upgrading↗

DOE OSTI2025

Agentic AI vs ML-Based Autotuning: A Comparative Study for Loop Reordering Optimization

High Performance Computing (HPC) applications rely heavily on code optimizations to achieve good performance on modern CPU and GPU architectures. Traditional Machine Learning auto-tuning approaches have demonstrated success in exploring high-dimensional spaces, but they often require expensive compile-run evaluations and lack adaptability for large HPC applications. The recent advances in Large Language Models (LLMs) and Agentic AI systems raise intriguing questions about the potential of these approaches to address specific optimization methodologies. This work aims to answer an essential question for the HPC community: “How Agentic AI Systems Compare to Traditional ML Autotuning Techniques?” To address this question, we present a comparative analysis between a traditional ML-based optimization approach and an Agentic AI system, evaluating their respective capabilities and limitations for loop-level optimization. In addition, we introduced a new Agentic AI system named LoopGen-AI using three different Large Language Models: GPT-4.1, Claude 4.0, and Gemini 2.5. A key finding is that LoopGen-AI achieves competitive per-formance with only a few program runs, the reasoning logs from the agents revealed that their decisions rely heavily on the combination of semantic understanding of the target kernel with dynamic feedback from the environment, highlighting a promising new dimension in performance tuning. In contrast, ML-based autotuners focus on statistical exploration, and require orders of magnitude more runs to reach peak performance. Additionally, our analysis shows that prompt engineering, particularly using Persona + Context Manager patterns, significantly impacts the effectiveness of Agentic AI. Our results indicate that while Agentic AI systems are not yet a complete replacement for ML-based autotuners, it can effectively complement traditional methods.

Rosas, Miguel Romero↗

DOE OSTI2025

SpectraCodec: A Hilbert curve-based method for encoding metadata in mass spectra for machine learning applications (SpectraCodec) v1

Machine learning approaches to mass spectrometry (MS) data analysis require structured metadata for optimal performance. However, current MS file formats necessitate external metadata sources, creating integration challenges that impede analytical workflows. Here, we present a novel approach for encoding metadata directly within mzML files using one-hot encoding of ASCII characters mapped via Hilbert space-filling curves. This strategy embeds metadata in the first spectrum's m/z-intensity space, ensuring persistence with the primary data, eliminating the need for external metadata files, and maintaining compatibility with existing MS software. We demonstrate that the Hilbert curve mapping efficiently utilizes the two-dimensional spectral space while maintaining robust data recovery. This method offers a practical solution for machine learning applications in mass spectrometry by ensuring metadata and spectral data remain unified through all stages of analysis.

Bowen, Benjamin [Lawrence Berkeley National Labora↗

DOE OSTI2021

Balanced k -means clustering on an adiabatic quantum computer

Adiabatic quantum computers are a promising platform for efficiently solving challenging optimization problems. Therefore, many are interested in using these computers to train computationally expensive machine learning models. We present a quantum approach to solving the balanced k-means clustering training problem on the D-Wave 2000Q adiabatic quantum computer. In order to do this, we formulate the training problem as a quadratic unconstrained binary optimization (QUBO) problem. Unlike existing classical algorithms, our QUBO formulation targets the global solution to the balanced k-means model. We test our approach on a number of small problems and observe that despite the theoretical benefits of the QUBO formulation, the clustering solution obtained by a modern quantum computer is usually inferior to the solution obtained by the best classical clustering algorithms. Nevertheless, the solutions provided by the quantum computer do exhibit some promising characteristics. We also perform a scalability study to estimate the run time of our approach on large problems using future quantum hardware. Finally, as a final proof of concept, we used the quantum approach to cluster random subsets of the Iris benchmark data set.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

The Path towards Plasma Facing Components: A Review of State-of-the-art in W-Based Refractory High-Entropy Alloys

Developing advanced materials for plasma-facing components (PFCs) in fusion reactors is a crucial aspect for achieving sustained energy production. Tungsten (W) - based refractory high-entropy alloys (RHEAs) have emerged as promising candidates due to their superior radiation tolerance and high-temperature strength. This review paper will focus on recent advancements in W-based RHEA research, particularly emphasizing the key role of modelling using machine learning (ML) in the stage of discovery by predicting properties for each composition and expediting the identification of optimal RHEA compositions with desired properties. Additionally, the application of additive manufacturing (AM) techniques for fabricating W-based RHEAs is explored, highlighting their advantages for rapid prototyping and multi-compositional sample production in a high throughput manner. The review critically evaluates the current understanding of mechanical properties relevant to PFC applications, including high-temperature strength and ductility. Furthermore, the radiation tolerance of W-based RHEAs under irradiated conditions is discussed. Finally, the validity of current AM-manufactured W-based RHEAs as PFC materials is assessed, and key challenges and opportunities for future research are identified. This review aims to provide a comprehensive overview of W-based RHEAs for fusion applications and their potential to guide the development and validation of advanced refractory high entropy alloys.

Hatler, Caleb [University of Wisconsin-Madison]↗

DOE OSTI2020

Development of neural network force fields for corrosion studies

To fully understand the chemistry and physics of corrosion, novel methods of simulation must be developed. One approach is designing machine learning (ML) algorithms integrated with density functional theory to develop adaptive force fields to gain insight into corrosion behavior namely at the surface of metal oxides. Current methods of modeling corrosion are slow due to the computational cost of resolving both reaction mechanics and mass transport processes. Machine learning methods can be implemented to obtain structure-activity relationships at both the molecular and bulk scale while still retaining the accuracy of density functional theory (DFT) and significantly decreasing the time needed for simulations of complex chemical processes in the various environments of corrosion. Multiscale models are needed for corrosion studies to fully understand its processes not only at the atomic length scale (chemical bonding, energies, and forces), but also at the nano and meso length scales (solid-state physics and material science processes). Current methods of study include DFT, molecular dynamics, and Monte Carlo. The limitation of DFT is that only a small number of atoms or molecules can be simulated at that level of theory. Density functional theory is used to study the electronic structure of atoms and molecules, and calculate the force component of each atom. However, these calculations are limited to about 1000 atoms. Custom periodic boundary conditions (PBC) can be used to describe the various environments and defects that affect the atomic forces to produce a large data set from which a training set can be derived. Machine learning can be utilized to overcome the barrier of modeling macroscopic and multi-scale processes from ab initio calculations through the development of adaptive force fields. Local environments determine the atomic forces of a given system, therefore adaptive force fields must be created to produce reliable quantum mechanical calculations. This can be achieved by developing a learning algorithm that uses the mapped atomic forces or fingerprint as an input to produce energies and magnetic moments as output. A systematic approach was used to begin to build a data set in order to accurately describe the atomic forces in various environments. In Figure 4 below, a simple PBC cell of Fe{sub 2}O{sub 3} was first optimized. A surface optimization was performed next, followed by a hydroxylated surface optimization. Once this calculation has converged, the adsorption of halide species to the hydroxylated surface will be investigated. TensorFlow is an open source platform for machine learning developed by Google. Using a high level application program interface (API) such as Keras allows for building and training ML models easily in a number of different environments and languages. For this project, a neural network was developed within Anaconda in Python. Future Work: Further development of reference data set; Refining neural network and learning algorithm; Fingerprinting atomic environment to enable mapping of atomic force components; Choosing appropriate training set from reference data; Learning from training set and enabling non-linear mapping of training set fingerprints and the atomic forces; Estimation of uncertainty to identify ranges of outside applicability; Testing and analysis of molecular dynamic simulations.

36 MATERIALS SCIENCE↗

DOE OSTI2024

Tailoring Molecular Space to Navigate Phase Complexity in Cs-Based Quasi-2D Perovskites via Gated-Gaussian-Driven High-Throughput Discovery

Cesium-based quasi-2D halide perovskites (HPs) offer promising functionalities and low-temperature manufacturability, suited to stable tandem photovoltaics. However, the chemical interplays between the molecular spacers and the inorganic building blocks during crystallization cause substantial phase complexities in the resulting matrices. To successfully optimize and implement the quasi-2D HP functionalities, a systematic understanding of spacer chemistry, along with the seamless navigation of the inherently discrete molecular space, is necessary. Herein, by utilizing high-throughput automated experimentation, the phase complexities in the molecular space of quasi-2D HPs are explored, thus identifying the chemical roles of the spacer cations on the synthesis and functionalities of the complex materials. Furthermore, a novel active machine learning algorithm leveraging a two-stage decision-making process, called gated Gaussian process Bayesian optimization is introduced, to navigate the discrete ternary chemical space defined with two distinctive spacer molecules. Through simultaneous optimization of photoluminescence intensity and stability that “tailors” the chemistry in the molecular space, a ternary-compositional quasi-2D HP film realizing excellent optoelectronic functionalities is demonstrated. Finally, this work not only provides a pathway for the rational and bespoke design of complex HP materials but also sets the stage for accelerated materials discovery in other multifunctional systems.

36 MATERIALS SCIENCE↗

DOE OSTI2024

Steam Generator Model Design Parameter Sensitivity Study Using Advanced Optimization Tools

This study focuses on design parameter sensitivity studies pertaining to a steam generator (SG) model, using both Python and machine-learning tools. The SG model is a mathematical representation (including fluid flow and heat transfer equations/models/correlations) of a steam-generating unit in a pressurized water reactor (PWR)-type small modular reactor (SMR) system. Design studies involve changing the model’s input design parameters (e.g., temperature, pressure, mass flow rate) to observe the resulting effects on the output of the system (e.g., heat transfer coefficient [HTC], Nusselt number, heat transfer performance). Sensitivity studies analyze the degree to which system output and/or desired parameters (e.g., HTC or heat transfer performance) are sensitive to changes in input parameters. By using machine-learning tools such as the Risk Analysis Virtual Environment (RAVEN) developed at Idaho National Laboratory (INL), detailed design parametric sensitivity studies and model optimization were performed. Six input parameters—namely, the pressure, temperature, and mass flow rate for the inlet of the primary-side (hot fluid) and secondary-side (cold fluid) of the SG—were randomly perturbed via RAVEN’s Monte Carlo Sampler module, using uniform distributions (±1% relative changes). The analysis results give valuable insights into SG system performance and optimization, and provide justification for researching optimized sensor placement to effectively monitor and obtain experimental data.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

DOE OSTI2024

Steam Generator Model Design Parameter Sensitivity Study Using Advanced Optimization Tools

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

DOE OSTI2024

Optimal Electrification Using Renewable Energies: Microgrid Installation Model with Combined Mixture k-Means Clustering Algorithm, Mixed Integer Linear Programming, and Onsset Method

Optimal planning and design of microgrids are priorities in the electrification of off-grid areas. Indeed, in one of the Sustainable Development Goals (SDG 7), the UN recommends universal access to electricity for all at the lowest cost. Several optimization methods with different strategies have been proposed in the literature as ways to achieve this goal. This paper proposes a microgrid installation and planning model based on a combination of several techniques. The programming language Python 3.10 was used in conjunction with machine learning techniques such as unsupervised learning based on K-means clustering and deterministic optimization methods based on mixed linear programming. These methods were complemented by the open-source spatial method for optimal electrification planning: onsset. Four levels of study were carried out. The first level consisted of simulating the model obtained with a cluster, which is considered based on the elbow and k-means clustering method as a case study. The second level involved sizing the microgrid with a capacity of 40 kW and optimizing all the resources available on site. The example of the different resources in the Togo case was considered. At the third level, the work consisted of proposing an optimal connection model for the microgrid based on voltage stability constraints and considering, above all, the capacity limit of the source substation. Finally, the fourth level involved a planning study of electrification strategies based mainly on microgrids according to the study scenario. The results of the first level of study enabled us to obtain an optimal location for the centroid of the cluster under consideration, according to the different load positions of this cluster. Then, the results of the second level of study were used to highlight the optimal resources obtained and proposed by the optimization model formulated based on the various technology costs, such as investment, maintenance, and operating costs, which were based on the technical limits of the various technologies. In these results, solar systems account for 80% of the maximum load considered, compared to 7.5% for wind systems and 12.5% for battery systems. Next, an optimal microgrid connection model was proposed based on the constraints of a voltage stability limit estimated to be 10% of the maximum voltage drop. The results obtained for the third level of study enabled us to present selective results for load nodes in relation to the source station node. Finally, the last results made it possible to plan electrification using different network technologies and systems in the short and long term. The case study of Togo was taken into account. The various results obtained from the different techniques provide the necessary leads for a feasibility study for optimal electrification of off-grid areas using microgrid systems.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2024

A deep learning-based Bayesian framework for high-resolution calibration of building energy models

Calibrating building energy models (BEMs), i.e., closing discrepancy between modeling and field measurements, is of significance to support its applications in building sustainability and resilience analysis. However, as being widely used in practice, current Bayesian calibration is mostly performed in low-resolution (annual or monthly), instead of high-resolution (hourly or sub-hourly), which is crucial to support emerging BEM applications, such as building-renewable energy integration (demand response) and smart control. This is attributable to the gaps in current Bayesian calibration process, including (1) difficulty in supporting reliable high-resolution calibration with over-parameterization and multi-solution issues, (2) inadequacy of meta-model to capture temporal building dynamics in high-resolution, and (3) excessive computational burdens of covariance matrix calculation in Bayesian inference. Therefore, to close these gaps, this research proposes a novel deep learning-based Bayesian calibration framework, involving pre-calibration mechanism, Long Short-Term Memory as surrogate models, and simplified covariance matrix calculation, to calibrate BEMs in high temporal resolution (i.e., hourly) with enhanced accuracy and computational efficiency. Finally, the case study demonstrates its effectiveness to match modeling outcomes with measurements and realize CV-RMSE of < 30 % and NMBE of < 6 % in hourly resolution, as well as a significant reduction of calibration time (by > 99 %, from > 600 h to ~ 1.5 h).

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

DOE OSTI2024

Design, tuning, and blackbox optimization of laser systems

Chirped pulse amplification (CPA) and subsequent nonlinear optical (NLO) systems constitute the backbone of myriad advancements in semiconductor manufacturing, communications, biology, defense, and beyond. Accurately and efficiently modeling CPA+NLO-based laser systems is challenging because of the complex coupled processes and diverse simulation frameworks. Our modular start-to-end model unlocks the potential for exciting new optimization and inverse design approaches reliant on data-driven machine learning methods, providing a means to create tailored CPA+NLO systems unattainable with current models. To demonstrate this new, to our knowledge, technical capability, we present a study on the LCLS-II photo-injector laser, representative of a high-power and spectro-temporally non-trivial CPA+NLO system.

47 OTHER INSTRUMENTATION↗

DOE OSTI2022

Biomaniac

Los Alamos National Laboratory researchers are developing a machine-learning based algorithm to revolutionize how biopolymers are designed and optimized. BioManIAC is a software platform that fills the gaps in data to gain insight into new potential polymer properties without making and testing them. Polymer engineers are able to use BioManIAC software to extrapolate and tune new biopolymer properties that allow the user to create small smart libraries, rather than largescale trial and error testing. This targeted design is expected to improve both time to market and R&D expenditure, as well as the resulting mechanical properties of biopolymers, enabling for more direct competition with traditional petroleum-based polymers. Los Alamos is seeking partners to further validate our software platform.

36 MATERIALS SCIENCE↗

DOE OSTI2024

Learning to Branch with Interpretable Machine Learning Models

This presentation describes an algorithm for applying machine learning to branching to speed up the solution of integer optimization problems. These problems are challenging and solved multiple times a day by power systems operators. We show that our approach speeds up a widely used open-source optimization solver.

Bayramoglu, Selin↗

DOE OSTI2024

Crystal Engineering of Hydrogen Bonding for Direct Air Capture of CO2: A Quantum Crystallography Perspective

Rising atmospheric CO2 levels demand efficient and sustainable carbon capture solutions. Direct air capture (DAC) via crystallizing hydrogen-bonded frameworks such as carbonate salts has emerged as a promising approach. This review explores the potential of crystal engineering, in tandem with advanced quantum crystallography techniques and computational modeling, to unlock the full potential of DAC materials. We examine the critical role of hydrogen bonding and other noncovalent interactions within a family of bis-guanidines that governs the formation of carbonate salts with high CO2 capture capacity and low regeneration energies for utilization. Quantum crystallography and charge density analysis prove instrumental in elucidating these interactions. A case study of a highly insoluble carbonate salt of a 2,6-pyridine-bis-(iminoguanidine) exemplifies the effectiveness of these approaches. However, challenges remain in the systematic and precise determination of hydrogen atom positions and atomic displacement parameters within DAC materials using quantum crystallography, and limitations persist in the accuracy of current energy estimation models for hydrogen bonding interactions. Future directions lie in exploring diverse functional groups, designing advanced hydrogen-bonded frameworks, and seamlessly integrating experimental and computational modeling with machine learning. This synergistic approach promises to propel the design and optimization of DAC materials, paving the way for a more sustainable future.

36 MATERIALS SCIENCE↗

DOE OSTI2025

PowerModel-AI: A First On-the-Fly Machine-Learning Predictor for AC Power Flow Solutions

The real-time creation of machine-learning models via active or on-the-fly learning has attracted considerable interest across various scientific and engineering disciplines. These algorithms enable machines to build models autonomously while remaining operational. Through a series of query strategies, the machine can evaluate whether newly encountered data fall outside the scope of the existing training set. In this study, we introduce PowerModel-AI, an end-to-end machine learning software designed to accurately predict AC power flow solutions. We present detailed justifications for our model design choices and demonstrate that selecting the right input features effectively captures load flow decoupling inherent in power flow equations. Our approach incorporates on-the-fly learning, where power flow calculations are initiated only when the machine detects a need to improve the dataset in regions where the model’s suboptimal performance is based on specific criteria. Otherwise, the existing model is used for power flow predictions. This study includes analyses of five Texas A&M synthetic power grid cases, encompassing the 14-, 30-, 37-, 200-, and 500-bus systems. The training and test datasets were generated using PowerModels.jl, an open-source power flow solver/optimizer developed at Los Alamos National Laboratory, NM, USA.

24 POWER TRANSMISSION AND DISTRIBUTION↗

DOE OSTI2025

Equation-Free Coarse Control of Distributed Parameter Systems via Local Neural Operators

The control of high-dimensional distributed parameter systems (DPS) remains a challenge when explicit coarse-grained equations are unavailable. Classical equation-free (EF) approaches rely on fine-scale simulators treated as black-box timesteppers. However, repeated simulations for steady-state computation, linearization, and control design are often computationally prohibitive, or the microscopic timestepper may not even be available, leaving us with data as the only resource. We propose a data-driven alternative that uses local neural operators, trained on spatiotemporal microscopic/mesoscopic data, to obtain efficient short-time solution operators. These surrogates are employed within Krylov subspace methods to compute coarse steady and unsteady-states, while also providing Jacobian information in a matrix-free manner. Krylov-Arnoldi iterations then approximate the dominant eigenspectrum, yielding reduced models that capture the open-loop slow dynamics without explicit Jacobian assembly. Both discrete-time Linear Quadratic Regulator (dLQR) and pole-placement (PP) controllers are based on this reduced system and lifted back to the full nonlinear dynamics, thereby closing the feedback loop.

93B52, 93C20, 47N70, 65J15, 65M32, 68T07, 68T20, 6↗

DOE OSTI2021

MADE3D: Enabling the Next-Generation High-Torque- Density Wind Generators by Additive Design and 3D Printing

Direct-drive wind turbine generators are increasing in popularity, thanks to recent project developments - especially offshore, where reliability and efficiency are major cost drivers. Yet, high capital costs are forcing many original equipment manufacturers to consider lightweight, high-torque density generators for next-generation multi-megawatt turbines that may be difficult to realize by traditional design or manufacturing methods. In this study, we present a new design framework enabled by advanced machine learning and multimaterial additive manufacturing to perform a magnetic topology optimization that maximizes the torque per rotor active mass for a 15-megawatt direct-drive permanent magnet wind generator. A comparison of the proposed approach against conventional topology optimization demonstrated a significant increase in computational efficiency and accuracy in performance predictions. Results using single and multimaterial compositions for rotor core and magnets identify a wider choice of 3D printable designs for a given specification. A hybrid combination of sintered and dysprosium-free polymer-bonded magnets shows good potential for torque performance by saving material costs up to 8.75%. More than 30% improvement in rotor torque densities is identified which can marginally improve the overall generator torque density. With the rapid evolution of multipowder deposition technologies, this study can greatly inspire a new paradigm for design-driven manufacturing with novel material compositions and lightweight, low-cost, high-strength multimaterial geometries that were previously unexplored for direct-drive generators.

3D printing↗