SEARCH · Engineering Papers
Results for “surrogates”
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.
Derivative-free bound-constrained optimization for solving structured problems with surrogate models
Not provided.
Nonlinear gyrokinetic predictions of SPARC burning plasma profiles enabled by surrogate modeling
Multi-channel, nonlinear predictions of core temperature and density profiles are performed for the SPARC tokamak accounting for both kinetic neoclassical and fully nonlinear gyro-kinetic turbulent fluxes. A series of flux-tube, nonlinear, electromagnetic simulations using the CGYRO code with six gyrokinetic species are coupled to a nonlinear optimizer using Gaussian process regression techniques. The simultaneous evolution of energy sources, including alpha heat, radiation, and energy exchange, coupled with these high fidelity models and techniques, leads to a converged solution in electron temperature, ion temperature and electron density channels with a minimal number of expensive gyrokinetic simulations without compromising accuracy.
Improving surrogate model accuracy for the LCLS-II injector frontend using convolutional neural networks and transfer learning
Not Available
Transformer-powered surrogates close the ICF simulation-experiment gap with extremely limited data
Abstract Recent advances in machine learning, specifically transformer architecture, have led to significant advancements in commercial domains. These powerful models have demonstrated superior capability to learn complex relationships and often generalize better to new data and problems. This paper presents a novel transformer-powered approach for enhancing prediction accuracy in multi-modal output scenarios, where sparse experimental data is supplemented with simulation data. The proposed approach integrates transformer-based architecture with a novel graph-based hyper-parameter optimization technique. The resulting system not only effectively reduces simulation bias, but also achieves superior prediction accuracy compared to the prior method. We demonstrate the efficacy of our approach on inertial confinement fusion experiments, where only 10 shots of real-world data are available, as well as synthetic versions of these experiments.
Nonlinear dimension reduction for surrogate modeling using gradient information
We introduce a method for the nonlinear dimension reduction of a high-dimensional function $u:{\mathbb{R}}^d\rightarrow{\mathbb{R}}$, $d\gg 1$. Our objective is to identify a nonlinear feature map $g:{\mathbb{R}}^d\rightarrow{\mathbb{R}}^m$, with a prescribed intermediate dimension $m\ll d$, so that $u$ can be well approximated by $f\circ g$ for some profile function $f:{\mathbb{R}}^m\rightarrow{\mathbb{R}}$. We propose to build the feature map by aligning the Jacobian $\nabla g$ with the gradient $\nabla u$, and we theoretically analyze the properties of the resulting $g$. Once $g$ is built, we construct $f$ by solving a gradient-enhanced least squares problem. Our practical algorithm uses a sample $\{{\textbf{x}}^{(i)},u({\textbf{x}}^{(i)}),\nabla u({\textbf{x}}^{(i)})\}_{i=1}^N$ and builds both $g$ and $f$ on adaptive downward-closed polynomial spaces, using cross validation to avoid overfitting. We numerically evaluate the performance of our algorithm across different benchmarks, and explore the impact of the intermediate dimension $m$. We show that building a nonlinear feature map $g$ can permit more accurate approximation of $u$ than a linear $g$, for the same input data set.
Surrogate modelling the Baryonic Universe – I. The colour of star formation
The spectral energy distribution of a galaxy emerges from the complex interplay of many physical ingredients, including its star formation history (SFH), metallicity evolution, and dust properties. Using GALAXPY , a new galaxy spectral prediction tool, and SFHs predicted by the empirical model UNIVERSEMACHINE and the cosmological hydrodynamical simulation IllustrisTNG, we isolate the influence of SFH on optical and near-infrared colours from 320 to 1080 Å at z = 0. By carrying out a principal component analysis, we demonstrate that physically motivated SFH variations modify galaxy colours along a single direction in colour space: the SFH-direction. We find that the projection of a galaxy’s present-day colours on to the SFH-direction is almost completely regulated by the fraction of stellar mass that the galaxy formed over the last billion years. Together with cosmic downsizing, this results in galaxies becoming redder as their host halo mass increases. We additionally study the change in galaxy colours due to variations in metallicity, dust attenuation, and nebular emission lines, finding that these properties vary broad-band colours along distinct directions in colour space relative to the SFH-direction. Finally, we show that the colours of low-redshift Sloan Digital Sky Survey galaxies span an ellipsoid with significant extent along two independent dimensions, and that the SFH-direction is well-aligned with the major axis of this ellipsoid. Our analysis supports the conclusion that variations in SFH are the dominant influence on present-day galaxy colours, and that the nature of this influence is strikingly simple.
Surrogate modelling the Baryonic Universe II: On forward modelling the colours of individual and populations of galaxies
ABSTRACT Among the properties shaping the light of a galaxy, the star formation history (SFH) is one of the most challenging to model due to the variety of correlated physical processes regulating star formation. In this work, we leverage the stellar population synthesis model fsps, together with SFHs predicted by the hydrodynamical simulation IllustrisTNG and the empirical model universemachine, to study the impact of star formation variability on galaxy colours. We start by introducing a model-independent metric to quantify the burstiness of a galaxy formation model, and we use this metric to demonstrate that universemachine predicts SFHs with more burstiness relative to IllustrisTNG. Using this metric and principal component analysis, we construct families of SFH models with adjustable variability, and we show that the precision of broad-band optical and near-infrared colours degrades as the level of unresolved short-term variability increases. We use the same technique to demonstrate that variability in metallicity and dust attenuation presents a practically negligible impact on colours relative to star formation variability. We additionally provide a model-independent fitting function capturing how the level of unresolved star formation variability translates into imprecision in predictions for galaxy colours; our fitting function can be used to determine the minimal SFH model that reproduces colours with some target precision. Finally, we show that modelling the colours of individual galaxies with per cent-level precision demands resorting to complex SFH models, while producing precise colours for galaxy populations can be achieved using models with just a few degrees of freedom.
Uncertainty aware machine-learning-based surrogate models for particle accelerators: Study at the Fermilab Booster Accelerator Complex
Not Available
Multifidelity validation of digital surrogates using variable-density turbulent mixing models
In this study, ensembles of experimental data are presented and utilized to compare and validate two models used in the simulation of variable-density (Atwood = 0.22), compressible turbulent mixing. Though models of this kind (Reynolds averaged NavierStokes and large-eddy simulations) have been validated extensively with more canonical flows in previous studies, here the present approach offers novelty in the complexity of the geometry, the ensemble-based validation, and the uniformity of the computational framework on which the models are tested. Moreover, all experimental and computational tasks were completed by the authors which has led to a tightly coupled experimental configuration with its “digital twin.” The experimental divergent-shock-tube facility and its data acquisition methods are described and replicated in simulation space. A 2D Euler model which neglects the turbulent mixing at the interface is optimized to experimental data using a Gaussian process. This model then serves as the basis for both the 2D RANS and 3D LES studies that make comparisons to the mixing-layer data from the experiment. A relatively simple RANS model is shown to produce good agreement with experimental data only at late flow development times. The LES ensembles generally show good agreement with experimental data but display sensitivity to the characterization of initial conditions. Resolution-dependent behavior is also observed for certain higher-order statistics of interest. Overall, the LES model successfully captures the effects of divergent geometry, compressibility, and combined nonlinear instabilities inherent to the problem. The successful prediction of mixing width and its growth rate highlight the existence of three distinct regimes in the development of the instability, each with similarities to previously studied instabilities.
Surrogate light curve models for kilonovae with comprehensive wind ejecta outflows and parameter estimation for AT2017gfo
Not Available
Dynamic Validation of CNN-Based Surrogate Models for Inverter-Based Resources in Open-Source Solvers
Traditionally, distribution system planning has focused on steady-state analyses, with limited consideration of dynamic behavior. However, as large or medium-scale inverter-based resources (IBRs), particularly grid-following (GFL) inverters in commercial or industry buildings, become more prevalent, understanding their dynamic impact is essential for grid planning and operation. This article presents an innovative deep-learning (DL)-approach using convolutional neural networks technique to model the GFL inverters. Developed from real grid-tied commercial IBR transient data, these dynamic DL models overcome proprietary constraints by requiring minimal knowledge of internal converter physics while maintaining high accuracy and flexibility. To demonstrate their applicability, the models were incorporated into GridLAB-D, an open-source, three-phase distribution analysis tool. This integration enables dynamic simulations of large-scale distribution networks with high IBR penetration stability analysis. Rigorous testing and validation, aligned with industry standards, confirmed the reliability and efficiency of this approach, paving the way for enhanced planning and operational assessments of modern power systems.
Transfer Learning of High-Fidelity Opacity Spectra in Autoencoders and Surrogate Models
Simulations of high energy density physics are expensive, largely in part for the need to produce nonlocal thermodynamic equilibrium opacities. High-fidelity spectra may reveal new physics in the simulations not seen with low-fidelity spectra, but the cost of these simulations also scales with the level of fidelity of the opacities being used. Neural networks are capable of reproducing these spectra, but neural networks need data to train them, which limits the level of fidelity of the training data. Here this article demonstrates that it is possible to reproduce high-fidelity spectra with median errors in the realm of 3%–4% using as few as 50 samples of high-fidelity Krypton data by performing transfer learning on a neural network trained on many times more low-fidelity data.
Inverse Design of Two-Dimensional Airfoils Using Conditional Generative Models and Surrogate Log-Likelihoods
Abstract This paper shows how to use conditional generative models in two-dimensional (2D) airfoil optimization to probabilistically predict good initialization points within the vicinity of the optima given the input boundary conditions, thus warm starting and accelerating further optimization. We accommodate the possibility of multiple optimal designs corresponding to the same input boundary condition and take this inversion ambiguity into account when designing our prediction framework. To this end, we first employ the conditional formulation of our previous work BézierGAN–Conditional BézierGAN (CBGAN)—as a baseline, then introduce its sibling conditional entropic BézierGAN (CEBGAN), which is based on optimal transport regularized with entropy. Compared with CBGAN, CEBGAN overcomes mode collapse plaguing conventional GANs, improves the average lift-drag (Cl/Cd) efficiency of airfoil predictions from 80.8% of the optimal value to 95.8%, and meanwhile accelerates the training process by 30.7%. Furthermore, we investigate the unique ability of CEBGAN to produce a log-likelihood lower bound that may help select generated samples of higher performance (e.g., aerodynamic performance). In addition, we provide insights into the performance differences between these two models with low-dimensional toy problems and visualizations. These results and the probabilistic formulation of this inverse problem justify the extension of our GAN-based inverse design paradigm to other inverse design problems or broader inverse problems.
KSTAR_UEDGE_Dataset: UEDGE Simulation Dataset for Training the KSTAR DivControlNN Surrogate Model
Explore the source record for details and available documents.