Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “constrained learning”

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.

At least 505 records · Page 28
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Solving multiphysics-based inverse problems with learned surrogates and constraints

Abstract Solving multiphysics-based inverse problems for geological carbon storage monitoring can be challenging when multimodal time-lapse data are expensive to collect and costly to simulate numerically. We overcome these challenges by combining computationally cheap learned surrogates with learned constraints. Not only does this combination lead to vastly improved inversions for the important fluid-flow property, permeability, it also provides a natural platform for inverting multimodal data including well measurements and active-source time-lapse seismic data. By adding a learned constraint, we arrive at a computationally feasible inversion approach that remains accurate. This is accomplished by including a trained deep neural network, known as a normalizing flow, which forces the model iterates to remain in-distribution, thereby safeguarding the accuracy of trained Fourier neural operators that act as surrogates for the computationally expensive multiphase flow simulations involving partial differential equation solves. By means of carefully selected experiments, centered around the problem of geological carbon storage, we demonstrate the efficacy of the proposed constrained optimization method on two different data modalities, namely time-lapse well and time-lapse seismic data. While permeability inversions from both these two modalities have their pluses and minuses, their joint inversion benefits from either, yielding valuable superior permeability inversions and CO 2 plume predictions near, and far away, from the monitoring wells.

Yin, Ziyi (ORCID:0000000250248771)↗
DOE OSTI2020

Conditional Generative Adversarial Networks for Solving Heat Transfer Problems

Generative Adversarial Networks (GANs) have been used as a deep learning approach to solving physics and engineering problems. Using deep learning for these problems is attractive in that reasonably accurate models can be inferred from only raw data, eliminating the need to define the exact physical equations governing a problem. We expand on previous work using GANs to generate steady-state solutions to the two-dimensional heat equation. Using a basic conditional GAN (cGAN), we generate accurate solutions for rectangular domains conditioned on four edge boundary conditions (MAE < 0.5%). For finding steady-state solutions over arbitrary two-dimensional domains (not constrained to rectangles), we use a cGAN designed for image-to-image translation. We train this GAN on various types of geometric domains (circles, squares, triangles, shapes with one circular or rectangular hole), achieving accurate results on test data made up of geometries similar to those in training (MAE < 1%). For both of these GANs, we experiment with different loss function terms, showing that a term using the gradients of solution images significantly improves the basic cGAN but not the image-to-image GAN. Lastly, we show that the image-to-image GAN performs poorly when applied to two-dimensional geometries that vary in structure from training data (MAE < 8% for shapes with multiple holes or different shaped holes). This demonstrates the cGAN's lack of generalizability. While the cGAN is an accurate and computationally efficient method when trained and tested on similarly structured data, it is a much less reliable method when applied to data that is slightly different in structure from the training data.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

AI and extreme scale computing to learn and infer the physics of higher order gravitational wave modes of quasi-circular, spinning, non-precessing black hole mergers

We use artificial intelligence (AI) to learn and infer the physics of higher order gravitational wave modes of quasi-circular, spinning, non precessing binary black hole mergers. We trained AI models using 14 million waveforms, produced with the surrogate model NRHybSur3dq8, that include modes up to $\ell$ ≤ 4 and (5,5), except for (4,0) and (4,1), that describe binaries with mass-ratios $\textit{q}$ ≤ 8, individual spins $s^z_{\{1,2\}} \in$[–0.8,0.8], and inclination angle $θ \in$ [0,π]. Our probabilistic AI surrogates can accurately constrain the mass-ratio, individual spins, effective spin, and inclination angle of numerical relativity waveforms that describe such signal manifold. We compared the predictions of our AI models with Gaussian process regression, random forest, k-nearest neighbors, and linear regression, and with traditional Bayesian inference methods through the PyCBC Inference toolkit, finding that AI outperforms all these approaches in terms of accuracy, and are between three to four orders of magnitude faster than traditional Bayesian inference methods. Our AI surrogates were trained within 3.4 hours using distributed training on 1,536 NVIDIA V100 GPUs in the Summit supercomputer.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Solving the Orszag–Tang vortex magnetohydrodynamics problem with physics-constrained convolutional neural networks

We study the 2D Orszag–Tang vortex magnetohydrodynamics (MHD) problem through the use of physics-constrained convolutional neural networks (PCNNs) for forecasting the density, ρ, and the magnetic field, B, as well as the prediction of B given the velocity field v of the fluid. In addition to translation equivariance from the convolutional architecture, other physics constraints were embedded: absence of magnetic monopoles, non-negativity of ρ, use of only relevant variables, and the periodic boundary conditions of the problem. The use of only relevant variables and the hard constraint of non-negative ρ were found to facilitate learning greatly. The divergenceless condition ∇·B=0 was implemented as a hard constraint up to machine precision through the use of a magnetic potential to define B=∇×A. Residual networks and data augmentation were also used to improve performance. This allowed for some of the residual models to function as surrogate models and provide reasonably accurate simulations. For the prediction task, the PCNNs were evaluated against a physics-informed neural network, which had the ideal MHD induction equation as a soft constraint. Several models were able to generate highly accurate fields, which are visually almost indistinguishable and have low mean squared error. Only methods with built-in hard constraints produced physical fields with ∇·B=0. The use of PCNNs for MHD has the potential to produce physically consistent real-time simulations to serve as virtual diagnostics in cases where inferences must be made with limited observables.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2026

Measurement of muon antineutrino charged current - 0 meson scattering, using the NOvA Near Detector

Antineutrino interaction cross sections are, at present, poorly constrained, particularly regarding the role of multi-nucleon processes such as 2-particle 2-hole (2p2h) interactions. The associated crosssection systematic uncertainties represent a significant challenge for precision oscillation measurements, especially for the next generation of neutrino experiments such as DUNE. We present a new measurement of the muon antineutrino charged-current cross section without mesons in the final state, using the high-statistics data set of the NOvA Near Detector. The analysis employs a cut-based selection enhanced by machine learning techniques to isolate a high-purity sample dominated by quasielastic (QE) and 2p2h interactions. We present the cross section as a function of the kinetic energy and scattering angle of the outgoing muon. We also present measurements of more model-dependent kinematic variables such as the neutrino energy and momentum transfer, to better probe the underlying nuclear physics. The results are compared against various neutrino event generators to test the robustness of current interaction models.

Vockerodt, Kevin John [Ohio State U.; Queen Mary, ↗
DOE OSTI2022

Co-optimized machine-learned manifold models for large eddy simulation of turbulent combustion

Many modeling approaches in large eddy simulation (LES) of turbulent combustion employ a projection of the thermochemical state onto a low-dimensional manifold within state space to reduce the number of transported variables and hence computational cost. Flamelet-generated manifolds (FGM) is an example of a well-established, physics-based approach, but increasingly, principal component analysis (PCA) is being used as a data-driven method for generating manifold models. For both approaches, the nonlinear relationship between the location on the predefined manifold and the outputs of interest, such as reaction rates, can be tabulated or encoded in a neural network. This work proposes a new approach for manifold modeling that extends these existing approaches. A modified neural network structure simultaneously encodes the definition of the manifold variables, the nonlinear mapping, and the subfilter closure for LES. This allows all three of these aspects of the model to be co-optimized, generating a model from any source of combustion thermochemical state data. The manifold parameterizing variables are constrained to be linear combinations of species, as in FGM and PCA-based models, to aid in interpretability and implementation. For LES, subfilter variances of the manifold variables are also included as inputs. Two types of a priori analysis are performed to evaluate the new approach. In the first, the model is trained on data from one-dimensional premixed flames. In this case, the approach recovers the behavior of flamelet-based manifold approaches, and in fact slightly improves performance by identifying an optimized progress variable. The approach is also applied to data from direct numerical simulations of spherical ignition kernels in isotropic turbulence. For any specified manifold dimensionality, the new approach provides substantially lower prediction errors than a PCA-based model developed from the same data set. Additionally, the LES formulation of the new approach can provide accurate predictions for filtered reaction rates across a variety of filter widths.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Physics vs structure: A systematic benchmark of learning strategies for multi-zone building thermal dynamics

Recent advances in physics-informed and data-driven machine learning promise improved thermal models for advanced building control, yet there is limited quantitative evidence on when added physics structure and architectural complexity are beneficial. Here, this work presents a systematic benchmark of five representative system identification methods for modeling multi-zone building thermal dynamics: linear state-space models, multi-layer perceptrons, neural state-space models, neural ordinary differential equations, and physically-consistent neural networks. The methods are evaluated across multiple data regimes and zone coupling strategies. Using a high-fidelity multi-zone commercial building emulator, we examine short-term and long-term prediction accuracy, computational efficiency, and ease of development. Our results reveal critical trade-offs between prediction performance, model complexity, and physical consistency. We demonstrate that decoupled, nonlinear black-box models consistently outperform coupled physics-constrained architectures in both predictive accuracy and out-of-distribution robustness in majority of the test cases for the building type considered in the study. Our findings quantify the cost of complexity in building thermal modeling and provide concrete, actionable, scenario-based guidelines for selecting model classes for control-oriented applications.

Building thermal modeling↗
DOE OSTI2024

Dynamically Learning Incentives for Load Control

As electrical generation becomes more distributed and volatile, and loads become more uncertain, controllability of distributed energy resources (DERs), regardless of their ownership status, will be necessary for grid reliability. Grid operators lack direct control over end-users' grid interactions, such as energy usage, but incentives can influence behavior -- for example, an end-user that receives a grid-driven incentive may adjust their consumption or expose relevant control variables in response. A key challenge in studying such incentives is the lack of data about human behavior, which usually motivates strong assumptions, such as distributional assumptions on compliance or rational utility-maximization. In this paper, we propose a general incentive mechanism in the form of a constrained optimization problem -- our approach is distinguished from prior work by modeling human behavior (e.g., reactions to an incentive) as an arbitrary unknown function. We propose feedback-based optimization algorithms to solve this problem that each leverage different amounts of information and/or measurements. We show that each converges to an asymptotically stable incentive with (near)-optimality guarantees given mild assumptions on the problem. Finally, we evaluate our proposed techniques in voltage regulation simulations on standard test beds. We test a variety of settings, including those that break assumptions required for theoretical convergence (e.g., convexity, smoothness) to capture realistic settings. In this evaluation, our proposed algorithms are able to find near-optimal incentives even when the reaction to an incentive is modeled by a theoretically difficult (yet realistic) function.

demand response↗
DOE OSTI2025

Data efficiency assessment of generative adversarial networks in energy applications

This study investigates the data requirements of generative artificial intelligence (AI), particularly generative adversarial networks (GANs), for reliable data augmentation in energy applications. Generative AI, though seen as a solution to data limitations, requires substantial data to learn meaningful distributions—a challenge often overlooked. This study addresses the challenge through synthetic data generation for critical heat flux (CHF) and power grid demand, focusing on renewable and nuclear energy. Two variants of GAN employed are conditional GAN (cGAN) and Wasserstein GAN (wGAN). Our findings include the strong dependency of GAN on data size, with performance declining on smaller datasets and varying performance when generalizing to unseen experiments. Mass flux and heated length significantly influence CHF predictions. wGAN is more robust to feature exclusion, making it suitable for constrained synthetic data generation. In energy demand forecasting, wGAN performed well for solar, wind, and load predictions. Longer lookback hours and larger datasets improved predictions, especially for load power. Seasonal variations posed challenges, with wGAN achieving a relatively high error of Root Mean Squared Error (RMSE) of 0.32 for load power prediction, compared to RMSE of 0.07 under same-season conditions. Feature exclusions impacted cGAN the most, while wGAN showed greater robustness. This study concludes that, while generative AI is effective for data augmentation, it requires substantial data and careful training to generate realistic synthetic data and generalize to new experiments in engineering applications.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

Transformation rate maps of dissolved organic carbon in the contiguous US

Riverine dissolved organic carbon (DOC) plays a vital role in regional and global carbon cycles. However, the processes of DOC conversion from soil organic carbon (SOC) and leaching into rivers are insufficiently understood, inconsistently represented, and poorly parameterized, particularly in land surface and Earth system models. As a first attempt to fill this gap, we propose a generic formula that directly connects SOC concentration with DOC concentration in headwater streams, where a single parameter, the transformation rate from SOC in the soil to DOC leaching flux (P r ), accounts for the overall processes governing SOC conversion to DOC and leaching from soils (along with runoff) into headwater streams. We then derive high-resolution P r maps over the contiguous US (CONUS) using SOC data from two different sources: the Harmonized World Soil Database v1.2 (HWSD) and SoilGrids 2.0. Both maps are developed following the same five major steps: (1) selecting independent catchments where observed riverine DOC data are available with reasonable quality; (2) estimating catchment-average SOC for the independent catchments; (3) estimating the P r values for these catchments based on the generic formula and catchment-average SOC; (4) developing a predictive model of P r with machine learning (ML) techniques and catchment-scale climate, hydrology, geology, and other attributes; and (5) deriving a national map of P r based on the ML model. For evaluation, we compare the DOC concentration derived using the P r map and the observed DOC concentration values at evaluation catchments. The resulting mean absolute scaled error and coefficient of determination are 0.73 and 0.47 for the HWSD-based model and 0.58 and 0.72 for the SoilGrids-based model, respectively, suggesting the effectiveness of the overall methodology. Efforts to constrain uncertainty and evaluate sensitivity of P r to different factors are discussed. To illustrate the use of such maps, we derive a riverine DOC concentration reanalysis dataset over CONUS. The two P r maps, robustly derived and empirically validated, lay a critical cornerstone for better simulating the terrestrial carbon cycle in land surface and Earth system models. Our findings not only set a foundation for improving our predictive understanding of the terrestrial carbon cycle at the regional and global scales, but also hold promises for informing policy decisions related to decarbonization and climate change mitigation. The data presented in this study are publicly available at https://doi.org/10.5281/zenodo.14563816 (Li et al., 2024).

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

A randomized sketching trust-region secant method for low-memory dynamic optimization

The numerical solution of dynamic optimization problems is often limited by the memory required to store the state trajectory, which is used to evaluate the objective function and its derivatives. Recently, [R. Muthukumar et al., SIAM Journal on Optimization 31(2), pp. 1242–1275 (2021)] introduced a trust-region method for dynamic optimization that employs randomized sketching to compress the state trajectory, resulting in inexact derivative computations. By adaptively learning the sketch rank, the trust-region algorithm achieves rigorous convergence guarantees. Here, we extend this approach to use secant Hessian approximations. Due to the randomness introduced by the sketch, the traditional secant update formulae can produce poor Hessian approximations. In particular, the difference of two gradients, computed from two different sketches, may be inconsistent. To overcome this, we employ a sketched approximation of the Hessian application, in lieu of computing the gradient difference. We numerically demonstrate the improved stability of this approach on an example from PDE-constrained optimization.

dynamic optimization↗
NASA NTRS1992

Symmetry as Bias: Rediscovering Special Relativity

This paper describes a rational reconstruction of Einstein's discovery of special relativity, validated through an implementation: the Erlanger program. Einstein's discovery of special relativity revolutionized both the content of physics and the research strategy used by theoretical physicists. This research strategy entails a mutual bootstrapping process between a hypothesis space for biases, defined through different postulated symmetries of the universe, and a hypothesis space for physical theories. The invariance principle mutually constrains these two spaces. The invariance principle enables detecting when an evolving physical theory becomes inconsistent with its bias, and also when the biases for theories describing different phenomena are inconsistent. Structural properties of the invariance principle facilitate generating a new bias when an inconsistency is detected. After a new bias is generated. this principle facilitates reformulating the old, inconsistent theory by treating the latter as a limiting approximation. The structural properties of the invariance principle can be suitably generalized to other types of biases to enable primal-dual learning.

Lowry, Michael R.↗
DOE OSTI2025

Fast jet tagging with MLP-Mixers on FPGAs

We explore the innovative use of MLP-Mixer models for real-time jet tagging and establish their feasibility on resource-constrained hardware like FPGAs. MLP-Mixers excel in processing sequences of jet constituents, achieving state-of-the-art performance on datasets mimicking Large Hadron Collider conditions. By using advanced optimization techniques such as High-Granularity Quantization and Distributed Arithmetic, we achieve unprecedented efficiency. These models match or surpass the accuracy of previous architectures, reduce hardware resource usage by up to 97%, double the throughput, and half the latency. Additionally, non-permutation-invariant architectures enable smart feature prioritization and efficient FPGA deployment, setting a new benchmark for machine learning in real-time data processing at particle colliders.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
NASA NTRS2018

Testing and Maturing a Mass Translating Mechanism for a Deep Space CubeSat

Near Earth Asteroid (NEA) Scout is a deep space satellite set to launch aboard NASA's Exploration Mission 1. The spacecraft fits within a CubeSat standard 6U (about 300 x 200 x 100 mm) and is designed to travel 1 AU over a 2.5 year mission to observe NEA VG 1991. The spacecraft will use an 86 sq.m solar sail to maneuver from lunar orbit to the NEA. One of the critical mechanisms aboard NEA Scout, the Active Mass Translator (AMT), has gone through rigorous design and test cycles since its conception in July of 2015. The AMT is a two-axis translation table required to balance the spacecraft's center of mass (CM) and solar sail center of pressure (CP) while also trimming disturbance torque created by off-nominal sail conditions. The AMT has very limited mass and volume requirements, but is still required to deliver a large translation range-about 160 x 68 mm-at sub mm accuracy and precision. The system is constrained to operate in complete exposure to space with limited power and data budgets for mechanical and thermal needs. The NEA Scout team developed and carried out a rigorous test suite for the prototype and engineering development unit (EDU). These tests uncovered numerous design failures and led to many failure investigations and iteration cycles. A paper was previously presented at the 43rd Aerospace Mechanisms Symposia entitled, "Development of a High Performance, Low Profile Translation Table with Wire Feedthrough for a Deep Space CubeSat". This paper will make note of specific lessons learned: manufacturing philosophy, testing ideologies for high-risk missions, thermal mitigation design for small, motor-driven mechanisms.

Few, Alex↗
DOE OSTI2020

Constrained Bayesian Optimization of Criticality Experiments at LLNL [Slides]

The design of criticality experiments is typically an iterative process that employs a Monte Carlo transport code. The goal is to find a design that optimizes some variable, like the sensitivity of a response to a cross section, while simultaneously ensuring criticality. The high fidelity of the Monte Carlo code is a great asset, but it makes exploring the design space computationally expensive. Herein, we present how a constrained Bayesian optimization algorithm can be used to efficiently design a criticality experiment. It uses Gaussian processes as a surrogate model to probe the design space and to reduce the number of code executions that are needed to find the optimum. We demonstrate constrained Bayesian optimization with a Pu-239/polyethylene solution system and a TEX experiment that is designed for critical ity safety validation of a nuclear waste model at the Hanford Site. For both systems, a global optimum was found within 75 Monte Carlo simulations.

42 ENGINEERING↗
DOE OSTI2020

Constrained Bayesian Optimization of Criticality Experiments [Slides]

The design of criticality experiments is typically an iterative process that employs a Monte Carlo transport code. The goal is to find a design that optimizes some variable, like the sensitivity of a response to a cross section, while simultaneously ensuring criticality. The high fidelity of the Monte Carlo code is a great asset, but it makes exploring the design space computationally expensive. Herein, we present how a constrained Bayesian optimization algorithm can be used to efficiently design a criticality experiment. It uses Gaussian processes as a surrogate model to probe the design space and to reduce the number of code executions that are needed to find the optimum. We demonstrate constrained Bayesian optimization with a Pu-239/polyethylene solution system and a TEX experiment that is designed for criticality safety validation of a nuclear waste model at the Hanford Site. For both systems, a global optimum was found within 75 Monte Carlo simulations.

42 ENGINEERING↗
NASA NTRS2002

The Right Amount of Glue: Technologies and Standards Relevant to a Future Solar-Terrestrial Data Environment

In order to meet the challenge of developing a new system science, we will need to employ technology that enables researchers to access data from fields with which they are at least initially unfamiliar as well as from sources they use more regularly. At the same time, the quantity of data to be obtained by missions such as the Solar Dynamics Observatory demands ease and simplicity of data access. These competing demands must in turn fit within severely constrained funding for data analysis in such projects. Based on experience in only a single discipline but with a diversity of data types and sources, we will give examples of technology that have made a significant difference in the way people do science. Similarly, we will show how adoption of a well-documented data format has made it easier for one community to search, reduce, and analyze data. We will also describe a community-supported data reduction and analysis software tree with useful features. We will attempt to generalize the lessons learned in these instances to features the broader, solar-terrestrial community might find compelling, while avoiding overdesign of a common data environment.

Gurman, J. B.↗
NASA NTRS1992

Resource constrained design of artificial neural networks using comparator neural network

We present a systematic design method executed under resource constraints for automating the design of artificial neural networks using the back error propagation algorithm. Our system aims at finding the best possible configuration for solving the given application with proper tradeoff between the training time and the network complexity. The design of such a system is hampered by three related problems. First, there are infinitely many possible network configurations, each may take an exceedingly long time to train; hence, it is impossible to enumerate and train all of them to completion within fixed time, space, and resource constraints. Second, expert knowledge on predicting good network configurations is heuristic in nature and is application dependent, rendering it difficult to characterize fully in the design process. A learning procedure that refines this knowledge based on examples on training neural networks for various applications is, therefore, essential. Third, the objective of the network to be designed is ill-defined, as it is based on a subjective tradeoff between the training time and the network cost. A design process that proposes alternate configurations under different cost-performance tradeoff is important. We have developed a Design System which schedules the available time, divided into quanta, for testing alternative network configurations. Its goal is to select/generate and test alternative network configurations in each quantum, and find the best network when time is expended. Since time is limited, a dynamic schedule that determines the network configuration to be tested in each quantum is developed. The schedule is based on relative comparison of predicted training times of alternative network configurations using comparator network paradigm. The comparator network has been trained to compare training times for a large variety of traces of TSSE-versus-time collected during back-propagation learning of various applications.

Wah, Benjamin W.↗