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.

SourceResearch connectionFrom yearThrough year

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

At least 505 records · Page 28

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Enhancing predictive capabilities in fusion burning plasmas through surrogate-based optimization in core transport solvers

Abstract This work presents the PORTALS framework (Rodriguez-Fernandez et al 2022 Nucl. Fusion 62 076036), which leverages surrogate modeling and optimization techniques to enable the prediction of core plasma profiles and performance with nonlinear gyrokinetic simulations at significantly reduced cost, with no loss of accuracy. The efficiency of PORTALS is benchmarked against standard methods, and its full potential is demonstrated on a unique, simultaneous 5-channel (electron temperature, ion temperature, electron density, impurity density and angular rotation) prediction of steady-state profiles in a DIII-D ITER Similar Shape plasma with GPU-accelerated, nonlinear CGYRO (Candy et al 2016 J. Comput. Phys. 324 73–93). This paper also provides general guidelines for accurate performance predictions in burning plasmas and the impact of transport modeling in fusion pilot plants studies.

Physics↗

DOE OSTI2024

Impact of various DIII-D diagnostics on the accuracy of neural network surrogates for kinetic EFIT reconstructions

Abstract Kinetic equilibrium reconstructions make use of profile information such as particle density and temperature measurements in addition to magnetics data to compute a self-consistent equilibrium. They are used in a multitude of physics-based modeling. This work develops a multi-layer perceptron (MLP) neural network (NN) model as a surrogate for kinetic Equilibrium Fitting (EFITs) and trains on the 2019 DIII-D discharge campaign database of kinetic equilibrium reconstructions. We investigate the impact of including various diagnostic data and machine actuator controls as input into the NN. When giving various categories of data as input into NN models that have been trained using those same categories of data, the predictions on multiple equilibrium reconstruction solutions (poloidal magnetic flux, global scalars, pressure profile, current profile) are highly accurate. When comparing different models with different diagnostics as input, the magnetics-only model outputs accurate kinetic profiles and the inclusion of additional data does not significantly impact the accuracy. When the NN is tasked with inferring only a single target such as the EFIT pressure profile or EFIT current profile, we see a large increase in the accuracy of the prediction of the kinetic profiles as more data is included. These results indicate that certain MLP NN configurations can be reasonably robust to different burning-plasma-relevant diagnostics depending on the accuracy requirements for equilibrium reconstruction tasks.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

DOE OSTI2024

Calibration of the 2-Phase Bubble Tracking Model for Liquid Mercury Target Simulation with Machine Learning Surrogate Models

The Spallation Neutron Source (SNS) at Oak Ridge National Laboratory is one of the most powerful accelerator-driven neutron sources in the world. The intense protons strike on SNS's mercury target to provide bright neutron beams, which also leads to severe fluid-structure interactions inside the target. Prediction of resultant loading on the target is difficult particularly when helium gas is injected into mercury to reduce the loading and mitigate the pitting damage on vessel walls. A 2-phase material model that incorporates the Rayleigh-Plesset (R-P) model is expected to address this multi-physics problem. However, several uncertain parameters in the R-P model require intensive simulations to determine their optimal values. With the help of machine learning and the measured target strain, we have studied the major uncertain parameters in this R-P model and developed a framework to identify optimal parameters that significantly reduce the discrepancy between simulations and experimental strains. The preliminary results show the possibility of using this mercury/helium mixture and surrogate models to predict a better match of target strain response when the helium gas is injected.

Lin, Lianshan↗

DOE OSTI2024

Machine learning surrogate for charged particle beam dynamics with space charge based on a recurrent neural network with aleatoric uncertainty

In this work, we develop a machine learning (ML) model with aleatoric uncertainty for the low energy beam transport (LEBT) region of the LANSCE linear accelerator in which we model the transport of a space-charge-dominated 750 keV proton beam through a lattice of 22 quadrupole magnets. Our ML model is developed based on data generated by a Kapchinsky–Vladimirsky (KV) envelope model of beam transport. We show that a recurrent neural network can be used as a dynamical surrogate model for fast prediction of the LEBT beam envelope. Furthermore, we endow the model with the prediction of aleatoric uncertainty and compare three different approaches. We demonstrate that the ML-based uncertainty quantification models are well calibrated and produce good estimates of the regions where the model is less certain about its predictions. This ML framework is a necessary step in the development of a real-time virtual diagnostic tool with uncertainty quantification that can be integrated into more complex downstream tasks (e.g., adaptive control or learning flexible control policies via reinforcement learning) for improved efficiency in beam operations. In future work, we plan to expand on this preliminary study by considering more realistic envelope models that include longitudinal momentum spread and dispersive effects in bending magnets, as well as particle tracking codes with 3D space charge (such as and ). Published by the American Physical Society 2024

43 PARTICLE ACCELERATORS↗

DOE OSTI2019

Determining the average prompt-fission-neutron multiplicity for Pu 239 ( n , f ) via a Pu 240 ( α , α ' f ) surrogate reaction

The average prompt-fission-neutron multiplicity ν ¯ is of significance in the areas of nuclear theory, nuclear nonproliferation, and nuclear energy. In this work, the surrogate-reaction method has been used for the first time to indirectly determine ν ¯ for Pu 239 ( n , f ) via Pu 240 ( α , α ' f ) reactions. A Pu 240 target was bombarded with a beam of 53.9-MeV α particles. Scattered α particles, fission products, and neutrons were measured with the NeutronSTARS detector array. Values of ν ¯ were obtained for a continuous range of equivalent incident neutron energies between 0.25 and 26.25 MeV, and the results agree well with direct neutron measurements.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2014

Pu 236 ( n , f ) , Pu 237 ( n , f ) , and Pu 238 ( n , f ) cross sections deduced from ( p , t ) , ( p , d ) , and ( p , p ' ) surrogate reactions

The Pu 236 (n,f), Pu 237 (n,f) and Pu 238 (n,f) cross sections have been inferred by utilizing the surrogate ratio method. Targets of Pu 239 and U 235 were bombarded with 28.5-MeV protons, and the light ion recoils, as well as fission fragments, were detected using the STARS detector array at the K150 Cyclotron at the Texas A&M cyclotron facility. The (p, tf) reaction on Pu 239 and U 235 targets was used to deduce the σ (Pu 236 (n,f))/σ(U 232 (n,f)) ratio, and the Pu 236 (n,f) cross section was subsequently determined for En=0.5–7.5 MeV. Similarly, the (p,df) reaction on the same two targets was used to deduce the σ(Pu 237 (n,f))/σ(U 233 (n,f)) ratio, and the Pu 237 (n,f) cross section was extracted in the energy range En=0.5–7 MeV. The Pu 238 (n,f) cross section was also deduced by utilizing the (p,p') reaction channel on the same targets. There is good agreement with the recent ENDF/B-VII.1 evaluated cross section data for Pu 238 (n,f) in the range En=0.5–10.5 MeV and for Pu 237 (n,f) in the range En=0.5–7 MeV; however, the Pu 236 (n,f) cross section deduced in the present work is higher than the evaluation between 2 and 7 MeV.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2018

Constraining Neutron Capture Cross Sections for Unstable Nuclei with Surrogate Reaction Data and Theory

Obtaining reliable data for nuclear reactions on unstable isotopes remains an extremely important task and a formidable challenge. Neutron capture cross sections—crucial ingredients for models of astrophysical processes, national security applications, and simulations of nuclear energy generation—are particularly elusive, as both projectile and target in the reaction are unstable. We demonstrate a new method for determining cross sections for neutron capture on unstable isotopes, using 87 Y (n , γ) as a prototype. To validate the method, a benchmark experiment is carried out to obtain the known 90 Zr (n , γ) cross section analogously. Our approach, which employs an indirect (“surrogate”) measurement combined with theory, can be generalized to a larger class of nuclear reactions. It can be used both with traditional stable-beam experiments and in inverse kinematics at rare-isotope facilities.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2025

Surrogate models for linear response

Linear response theory is a well-established method in physics and chemistry for exploring excitations of many-body systems. In particular, the quasiparticle random-phase approximation (QRPA) provides a powerful microscopic framework by building excitations on top of the mean-field vacuum; however, its high computational cost limits model calibration and uncertainty quantification studies. Here, we present two complementary QRPA surrogate models and apply them to study response functions of finite nuclei. One is a reduced-order model that exploits the underlying QRPA structure, while the other utilizes the recently developed parametric matrix model algorithm to construct a map between the system’s Hamiltonian and observables. Our benchmark applications, the calculation of the electric dipole polarizability of 180 Yb and the 𝛽-decay half-life of 80 Ni, show that both emulators can achieve 0.1%–1% accuracy while offering a 6–7 orders of magnitude speedup compared to state-of-the-art QRPA solvers. These results demonstrate that the developed QRPA emulators are well positioned to enable Bayesian calibration and large-scale studies of computationally expensive physics models describing the properties of many-body systems.

Beta decay↗

DOE OSTI2022

Strategies for Integrating Deep Learning Surrogate Models with HPC Simulation Applications

The emerging trend of the convergence of high performance computing (HPC), machine learning/deep learning (ML/DL), and big data analytics presents a host of challenges for large-scale computing campaigns that seek best practices to interleave traditional scientific simulation-based workloads with ML/DL models. A portfolio of systematic approaches to incorporate deep learning into modeling and simulation serves a vital need when we support AI for science at a computing facility. In this paper, we evaluate several strategies for deploying deep learning surrogate models in a representative physics application on supercomputers at the Oak Ridge Leadership Computing Facility (OLCF). We discuss a set of recommended deployment architectures and implementation approaches. We analyze and evaluate these alternatives and show their performance and scalability up to 1000 GPUs on two mainstream platforms equipped with different deep learning hardware and software stacks.

Yin, Junqi↗

DOE OSTI2021

HYPPO: A Surrogate-Based Multi-Level Parallelism Tool for Hyperparameter Optimization

We present a new software, HYPPO, that enables the automatic tuning of hyperparameters of various deep learning (DL) models. Unlike other hyperparameter optimization (HPO) methods, HYPPO uses adaptive surrogate models and directly accounts for uncertainty in model predictions to find accurate and reliable models that make robust predictions. Using asynchronous nested parallelism, we are able to significantly alleviate the computational burden of training complex architectures and quantifying the uncertainty. HYPPO is implemented in Python and can be used with both TensorFlow and PyTorch libraries. We demonstrate various software features on time-series prediction and image classification problems as well as a scientific application in computed tomography image reconstruction. Finally, we show that (1) we can reduce by an order of magnitude the number of evaluations necessary to find the most optimal region in the hyperparameter space and (2) we can reduce by two orders of magnitude the throughput for such HPO process to complete.

adaptation models↗

DOE OSTI2026

Stellarator Design Exploration Using Symbolic-Regression Neutronics Surrogates

Systems codes require fast, simplified models to rapidly evaluate fusion power plant concepts, but neutronics analyses are often a computational bottleneck. Here, to address this, surrogate models for key neutronics responses have been developed using 3-D neutronics-ready models built with the open-source code ParaStell from a database of stellarator equilibria. Neutronics responses such as tritium breeding ratio (TBR), nuclear heating, and neutron-induced radiation damage displacements per atom (dpa) were simulated using OpenMC. Through sensitivity analysis and symbolic regression (SR), simple power-law formulas were derived connecting these neutronics responses to global stellarator parameters, including fusion power, plasma surface area, and plasma elongation. Validation shows these formulas can predict the simulation results with low error, enabling quick and accurate assessment of neutronics requirements in stellarator design exploration activities with systems codes.

Modeling↗

DOE OSTI2022

Redox chemistry of plutonium and plutonium surrogates in vitrified nuclear wastes

Lanthanide borosilicate glasses containing Pu and the analog element, Ce, were subjects of an X-ray absorption spectroscopy investigation to quantify the +3/+4 ratio of the redox-sensitive element. The data show that the dominant oxidation states are +4 for Pu and +3 for Ce. The data also indicate that the reductive potential of glasses can be quantified, although allowances must be made for glass composition, from a solution chemistry method adapted to glass chemistry. These data can therefore be used to formulate glass compositions and processing schedules that lead to knowledge of the oxidation state of Pu in melts. Furthermore, the data show that Ce is a poor analog for Pu behavior in melt and that the suitability of surrogates can be evaluated by this method. The methods demonstrated in this paper can be used to estimate the oxidation states of a range of multi-valent elements as functions of temperature and composition with data from only a single redox couple.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗

DOE OSTI2022

Thermal properties of HfN-MoW surrogate cermet fuel for nuclear thermal propulsion

Here, the thermal properties of MoW-HfN, a surrogate cermet for MoW-UN nuclear thermal propulsion fuel, were characterized over a wide range of elevated temperatures. Thermal diffusivity, coefficient of thermal expansion (CTE), and heat capacity were measured. Optical and scanning electron microscopy were performed to characterize the microstructure and draw structure–property correlations. The thermal diffusivity was obtained using the laser flash method. Diffusivity values ranged from about 0.18 cm 2 /s at 200°C to 0.15 cm 2 /s at 1800°C. The CTE was measured using push-rod dilatometry up to 1600°C, giving values between 6.0 and 9.0 μm/m. A scientific rationalization of the effective material properties is made using the rule-of-mixtures and other effective properties models.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

DOE OSTI2015

ASME V\&V challenge problem: Surrogate-based V&V

The process of verification and validation can be resource intensive. From the computational model perspective, the resource demand typically arises from long simulation run times on multiple cores coupled with the need to characterize and propagate uncertainties. In addition, predictive computations performed for safety and reliability analyses have similar resource requirements. For this reason, there is a tradeoff between the time required to complete the requisite studies and the fidelity or accuracy of the results that can be obtained. At a high level, our approach is cast within a validation hierarchy that provides a framework in which we perform sensitivity analysis, model calibration, model validation, and prediction. The evidence gathered as part of these activities is mapped into the Predictive Capability Maturity Model to assess credibility of the model used for the reliability predictions. With regard to specific technical aspects of our analysis, we employ surrogate-based methods, primarily based on polynomial chaos expansions and Gaussian processes, for model calibration, sensitivity analysis, and uncertainty quantification in order to reduce the number of simulations that must be done. The goal is to tip the tradeoff balance to improving accuracy without increasing the computational demands.

42 ENGINEERING↗

DOE OSTI2019

A Kriging Surrogate Model for Computing Gas Mixture Equations of State

Accurate simulation of the complex flow following the detonation of an explosive material is a challenging problem. In these flows, the detonation products of the explosive must be treated as a real gas while the surrounding air is treated as an ideal gas. As the detonation process unfolds and the blast wave moves into the surrounding ambient air, the products of detonation expand outward and interact with the air creating a mixture region. In this region, both of the state equations for air and the products must be satisfied. One of the most accurate, yet computationally expensive, methods to handle this problem is an algorithm that iterates between the equations of state until both pressure and temperature reach an equilibrium inside of a computational cell. Since this mixture region moves and grows over time, this algorithm must be performed millions, or even billions, of times in a typical detonation simulation. As such, these calculations can account for a large percentage of the overall solution time. This work aims to use a kriging surrogate model to replace this process. The iterative method solves a nonlinear system of equations created from the gas mixture density, internal energy, and composition using a Broyden iterative solver to obtain an output pressure and temperature. Kriging is used to produce curve fits which interpolate selected pressures and temperatures from this solver from appropriate ranges of the mixture input quantities. Using a finite volume hydrocode, the performance of the model with respect to the iterative solver is demonstrated in the simulation of a pentaerythritol tetranitrate (PETN) charge detonation. The model's computational speed and accuracy are quantified as a function of the choice of sampling points in order to try optimize the combination as well as to show the benefits of this novel approach.

Engineering↗

DOE OSTI2014

TH-E-17A-01: Internal Respiratory Surrogate for 4D CT Using Fourier Transform and Anatomical Features

Purpose: To develop a novel algorithm to generate internal respiratory signals for sorting of four-dimensional (4D) computed tomography (CT) images. Methods: The proposed algorithm extracted multiple time resolved features as potential respiratory signals. These features were taken from the 4D CT images and its Fourier transformed space. Several low-frequency locations in the Fourier space and selected anatomical features from the images were used as potential respiratory signals. A clustering algorithm was then used to search for the group of appropriate potential respiratory signals. The chosen signals were then normalized and averaged to form the final internal respiratory signal. Performance of the algorithm was tested in 50 4D CT data sets and results were compared with external signals from the real-time position management (RPM) system. Results: In almost all cases, the proposed algorithm generated internal respiratory signals that visibly matched the external respiratory signals from the RPM system. On average, the end inspiration times calculated by the proposed algorithm were within 0.1 s of those given by the RPM system. Less than 3% of the calculated end inspiration times were more than one time frame away from those given by the RPM system. In 3 out of the 50 cases, the proposed algorithm generated internal respiratory signals that were significantly smoother than the RPM signals. In these cases, images sorted using the internal respiratory signals showed fewer artifacts in locations corresponding to the discrepancy in the internal and external respiratory signals. Conclusion: We developed a robust algorithm that generates internal respiratory signals from 4D CT images. In some cases, it even showed the potential to outperform the RPM system. The proposed algorithm is completely automatic and generally takes less than 2 min to process. It can be easily implemented into the clinic and can potentially replace the use of external surrogates.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

DOE OSTI2015

SU-D-210-06: Feasibility for Monitoring the Head of the Pancreas Motion Through a Surrogate Using Ultrasound During Radiation Therapy Delivery

Purpose: Substantial target motion during the delivery of radiation therapy (RT) for pancreatic cancer is well recognized as a major limiting factor on RT effectiveness. The aim of this work is to monitor intra-fractional motion of the pancreas using ultrasound during RT delivery. Methods: Transabdominal Ultrasound B-mode images were collected from 5 volunteers using a research version of the Clarity Autoscan System (Elekta). The autoscan transducer with center frequency of 5 MHz was utilized for the scans. Imaging parameters were adjusted to acquire images at the desired depth with good contrast and a wide sweep angle. Since well-defined boundaries of the pancreas can be difficult to find on ultrasound B-mode images, the portal vein was selected as a surrogate for motion estimation of the head of the pancreas. The selection was due to its anatomical location posterior to the neck of the pancreas and close proximity to the pancreas head. The portal vein was contoured on the ultrasound images acquired during simulation using the Clarity Research AFC Workstation software. Volunteers were set up in a similar manner to the simulation for their monitoring session and the ultrasound transducer was mounted on an arm fixed to the couch. A video segment of the portal vein motion was captured. Results: The portal vein was visualized and segmented. Successful monitoring sessions of the portal vein were observed. In addition, our results showed that the ultrasound transducer itself reduces breathing related motion. This is analogous to the use of a compression plate to suppress respiration motion during thorax or abdominal irradiation. Conclusion: We demonstrate the feasibility of tracking the pancreas through the localization of the portal vein using abdominal ultrasound. This will allow for real-time tracking of the intra-fractional motion to justify PTV-margin and to account for unusual motions, thus, improving normal tissue sparing. This research was funding in part by Elekta Inc.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2016

Image segmentation with a novel regularized composite shape prior based on surrogate study

Purpose: Incorporating training into image segmentation is a good approach to achieve additional robustness. This work aims to develop an effective strategy to utilize shape prior knowledge, so that the segmentation label evolution can be driven toward the desired global optimum. Methods: In the variational image segmentation framework, a regularization for the composite shape prior is designed to incorporate the geometric relevance of individual training data to the target, which is inferred by an image-based surrogate relevance metric. Specifically, this regularization is imposed on the linear weights of composite shapes and serves as a hyperprior. The overall problem is formulated in a unified optimization setting and a variational block-descent algorithm is derived. Results: The performance of the proposed scheme is assessed in both corpus callosum segmentation from an MR image set and clavicle segmentation based on CT images. The resulted shape composition provides a proper preference for the geometrically relevant training data. A paired Wilcoxon signed rank test demonstrates statistically significant improvement of image segmentation accuracy, when compared to multiatlas label fusion method and three other benchmark active contour schemes. Conclusions: This work has developed a novel composite shape prior regularization, which achieves superior segmentation performance than typical benchmark schemes.

60 APPLIED LIFE SCIENCES↗