Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Surrogate Models”

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 1,549 records · Page 86

Fitting local, low-dimensional parameterizations of optical turbulence modeled from optimal transport velocity vectors

This work exploits a connection between optimal transport theory and the physics of image propagation to yield a locally low-dimensional model of turbulence-corrupted imagery. Optimal transport produces an invertible, pixel-wise linear trajectories to approximate the globally nonlinear turbulence between a clean and turbulence corrupted image pair. We use the low-dimensional model to fit subsets of the optimal transport vector fields and stitch the local models into a surrogate for the global map to be used for image cleaning. Experiments are performed on laboratory generated data of beam propagation using different values of the Fried parameter (a scale measuring turbulence coherence) as well as a toy data set. The results suggest this is a fruitful direction, and first step, towards using multiple realizations of turbulence corrupted images to learn a blind surrogate for the optimal transport vector field for image cleaning.

47 OTHER INSTRUMENTATION↗

Computational and Experimental Study for the Denitrification of Biomass-Derived Hydrothermal Liquefaction Oil

Hydrothermal liquefaction (HTL) is a promising method for processing wet biomass and waste feedstock to produce biofuels. During the HTL process, proteins and other biomolecules in certain feedstock get converted into nitrogenous compounds in produced biocrude, which represents a major challenge to further upgrading them into fuels. One promising approach is to separate nitrogenous compounds from the biocrude using polymeric resins. In this study, experiments were conducted to down-select sorbent and resin systems using a nitrogen-compound-containing surrogate biocrude. We model the binding interactions between an Amberlyst polymeric resin with various compounds present in the biocrude mixture such as nitrogenous compounds like pyrrole, pyridine, hexanamide, and representative co-existing compounds like phenol and dodecanoic acid. To ascertain the efficiency of various resins in the denitrogenation process, we have developed a quantitative structure–function model for the interacting components in the mixture. Our results suggest that the Amberlyst resin is a viable candidate for efficient removal of target nitrogen-containing compounds (such as pyridine) from the biocrude as a result of favorable interactions. The computational studies provide some insight into how and why the identified resin (Amberlyst) works in selective extraction of nitrogenous compound(s).

09 BIOMASS FUELS↗

Machine learning in materials informatics: recent applications and prospects

Propelled partly by the Materials Genome Initiative, and partly by the algorithmic developments and the resounding successes of data-driven efforts in other domains, informatics strategies are beginning to take shape within materials science. These approaches lead to surrogate machine learning models that enable rapid predictions based purely on past data rather than by direct experimentation or by computations/simulations in which fundamental equations are explicitly solved. Data-centric informatics methods are becoming useful to determine material properties that are hard to measure or compute using traditional methods—due to the cost, time or effort involved—but for which reliable data either already exists or can be generated for at least a subset of the critical cases. Predictions are typically interpolative, involving fingerprinting a material numerically first, and then following a mapping (established via a learning algorithm) between the fingerprint and the property of interest. Fingerprints, also referred to as “descriptors”, may be of many types and scales, as dictated by the application domain and needs. Predictions may also be extrapolative—extending into new materials spaces—provided prediction uncertainties are properly taken into account. This article attempts to provide an overview of some of the recent successful data-driven “materials informatics” strategies undertaken in the last decade, with particular emphasis on the fingerprint or descriptor choices. The review also identifies some challenges the community is facing and those that should be overcome in the near future.

36 MATERIALS SCIENCE↗

Genetic algorithms for computational materials discovery accelerated by machine learning

Materials discovery is increasingly being impelled by machine learning methods that rely on pre-existing datasets. Where datasets are lacking, unbiased data generation can be achieved with genetic algorithms. Here a machine learning model is trained on-the-fly as a computationally inexpensive energy predictor before analyzing how to augment convergence in genetic algorithm-based approaches by using the model as a surrogate. This leads to a machine learning accelerated genetic algorithm combining robust qualities of the genetic algorithm with rapid machine learning. The approach is used to search for stable, compositionally variant, geometrically similar nanoparticle alloys to illustrate its capability for accelerated materials discovery, e.g., nanoalloy catalysts. The machine learning accelerated approach, in this case, yields a 50-fold reduction in the number of required energy calculations compared to a traditional “brute force” genetic algorithm. This makes searching through the space of all homotops and compositions of a binary alloy particle in a given structure feasible, using density functional theory calculations.

36 MATERIALS SCIENCE↗

Fast and accurate machine learning prediction of phonon scattering rates and lattice thermal conductivity

Abstract Lattice thermal conductivity is important for many applications, but experimental measurements or first principles calculations including three-phonon and four-phonon scattering are expensive or even unaffordable. Machine learning approaches that can achieve similar accuracy have been a long-standing open question. Despite recent progress, machine learning models using structural information as descriptors fall short of experimental or first principles accuracy. This study presents a machine learning approach that predicts phonon scattering rates and thermal conductivity with experimental and first principles accuracy. The success of our approach is enabled by mitigating computational challenges associated with the high skewness of phonon scattering rates and their complex contributions to the total thermal resistance. Transfer learning between different orders of phonon scattering can further improve the model performance. Our surrogates offer up to two orders of magnitude acceleration compared to first principles calculations and would enable large-scale thermal transport informatics.

36 MATERIALS SCIENCE↗

A Multi-Objective Bayesian Optimization Approach Using the Weighted Tchebycheff Method

Abstract Bayesian optimization (BO) is a low-cost global optimization tool for expensive black-box objective functions, where we learn from prior evaluated designs, update a posterior surrogate Gaussian process model, and select new designs for future evaluation using an acquisition function. This research focuses upon developing a BO model with multiple black-box objective functions. In the standard multi-objective (MO) optimization problem, the weighted Tchebycheff method is efficiently used to find both convex and non-convex Pareto frontiers. This approach requires knowledge of utopia values before we start optimization. However, in the BO framework, since the functions are expensive to evaluate, it is very expensive to obtain the utopia values as a prior knowledge. Therefore, in this paper, we develop a MO-BO framework where we calibrate with multiple linear regression (MLR) models to estimate the utopia value for each objective as a function of design input variables; the models are updated iteratively with sampled training data from the proposed MO-BO. These iteratively estimated mean utopia values are used to formulate the weighted Tchebycheff MO acquisition function. The proposed approach is implemented in two numerical test examples and one engineering design problem of optimizing thin tube geometries under constant loading of temperature and pressure, with minimizing the risk of creep-fatigue failure and design cost, along with risk-based and manufacturing constraints. Finally, the model accuracy with frequentist, Bayesian and without MLR-based calibration are compared to true Pareto solutions.

Engineering↗

Alternative Methods for Tritium Recovery from the Life Lithium Blanket: FY13 Final Report

One of the critical components for the Laser Inertial Fusion Energy (LIFE) power plant is the recovery of tritium from the reactor lithium cooling blanket. In the process proposed by Maroni, high temperature halide salts extract tritium from the lithium using a high-speed centrifugal contactor. Because of high-temperature fluid mixture is very corrosive, this system is not practical. The LIFE team has proposed an alternative process that utilizes a halide-free salt (e.g., LiOH) with a non-centrifuge separation process to extract tritium from the lithium cooling blanket, thus eliminating the need for high-temperature, halide corrosion-resistant materials. To facilitate the design of the contactor and separator, Colorado State University (CSU) was subcontracted by LLNL to conduct the following initial steps in the design process during FY13: • Develop preliminary designs for the separator and Plexiglas models for flow visualization and computational model development, • Select surrogate fluids and flow conditions for the separator, and • Select an appropriate flow visualization fluorescent dye for the surrogate tritium solute.

42 ENGINEERING↗

Accident Progression and Source Term Demonstration Calculations for the Advanced Burner Test Reactor (ABTR)

As advanced reactor designers approach broad deployment of new reactor technologies and the associated applications, a critical milestone on this path includes licensing or authorization of these designs in the current regulatory environment. Commercial licensing and DOE authorization both consist of numerous aspects to not only certify the designed plant will operate as intended, but also related to confirmation of the safe operation of these facilities that comply with specific standards and requirements. To that end, domestic industry and regulators require toolsets and methods which can be used to confirm that advanced reactor designs not only align with expected performance during normal operation, but also a series of transient scenarios. These codes require a satisfactory validation basis, and in the absence of such a basis, code-to-code comparisons are often utilized as a surrogate to confirm model adequacy. To support this need for fast reactor technologies, a series of analyses using SAS4A/SASSYS- 1 and the Simplified Radionuclide Transport (SRT) code have been conducted for the Advanced Burner Test Reactor (ABTR). These analyses are expected to not only provide a baseline for transient and source term analyses of ABTR, but to also provide context for the key physics that determine evolution of specific transient scenarios in a sodium-cooled fast reactor (SFR). These codes are utilized for this work given their adoption in the SFR industry and by key SFR stakeholders.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

A graph neural network (GNN) approach to basin-scale river network learning: the role of physics-based connectivity and data fusion

Abstract. Rivers and river habitats around the world are under sustained pressure from human activities and the changing global environment. Our ability to quantify and manage the river states in a timely manner is critical for protecting the public safety and natural resources. In recent years, vector-based river network models have enabled modeling of large river basins at increasingly fine resolutions, but are computationally demanding. This work presents a multistage, physics-guided, graph neural network (GNN) approach for basin-scale river network learning and streamflow forecasting. During training, we train a GNN model to approximate outputs of a high-resolution vector-based river network model; we then fine-tune the pretrained GNN model with streamflow observations. We further apply a graph-based, data-fusion step to correct prediction biases. The GNN-based framework is first demonstrated over a snow-dominated watershed in the western United States. A series of experiments are performed to test different training and imputation strategies. Results show that the trained GNN model can effectively serve as a surrogate of the process-based model with high accuracy, with median Kling–Gupta efficiency (KGE) greater than 0.97. Application of the graph-based data fusion further reduces mismatch between the GNN model and observations, with as much as 50 % KGE improvement over some cross-validation gages. To improve scalability, a graph-coarsening procedure is introduced and is demonstrated over a much larger basin. Results show that graph coarsening achieves comparable prediction skills at only a fraction of training cost, thus providing important insights into the degree of physical realism needed for developing large-scale GNN-based river network models.

54 ENVIRONMENTAL SCIENCES↗

Human factors with nonhumans - Factors that affect computer-task performance

There are two general strategies that may be employed for 'doing human factors research with nonhuman animals'. First, one may use the methods of traditional human factors investigations to examine the nonhuman animal-to-machine interface. Alternatively, one might use performance by nonhuman animals as a surrogate for or model of performance by a human operator. Each of these approaches is illustrated with data in the present review. Chronic ambient noise was found to have a significant but inconsequential effect on computer-task performance by rhesus monkeys (Macaca mulatta). Additional data supported the generality of findings such as these to humans, showing that rhesus monkeys are appropriate models of human psychomotor performance. It is argued that ultimately the interface between comparative psychology and technology will depend on the coordinated use of both strategies of investigation.

Washburn, David A.↗

The rhesus measurement system: A new instrument for space research

The Rhesus Research Facility (RRF) is a research environment designed to study the effects of microgravity using rhesus primates as human surrogates. This experimental model allows investigators to study numerous aspects of microgravity exposure without compromising crew member activities. Currently, the RRF is slated for two missions to collect its data, the first mission is SLS-3, due to fly in late 1995. The RRF is a joint effort between the United States and France. The science and hardware portions of the project are being shared between the National Aeronautics and Space Administration (NASA) and France's Centre National D'Etudes Spatiales (CNES). The RRF is composed of many different subsystems in order to acquire data, provide life support, environmental enrichment, computer facilities and measurement capabilities for two rhesus primates aboard a nominal sixteen day mission. One of these subsystems is the Rhesus Measurement System (RMS). The RMS is designed to obtain in-flight physiological measurements from sensors interfaced with the subject. The RMS will acquire, preprocess, and transfer the physiologic data to the Flight Data System (FDS) for relay to the ground during flight. The measurements which will be taken by the RMS during the first flight will be respiration, measured at two different sites; electromyogram (EMG) at three different sites; electroencephalogram (EEG); electrocardiogram (ECG); and body temperature. These measurements taken by the RMS will assist the research team in meeting the science objectives of the RRF project.

Schonfeld, Julie E.↗

Climate Change Detection and Attribution of Infrared Spectrum Measurements

Climate change occurs when the Earth's energy budget changes due to natural or possibly anthropogenic forcings. These forcings cause the climate system to adjust resulting in a new climate state that is warmer or cooler than the original. The key question is how to detect and attribute climate change. The inference of infrared spectral signatures of climate change has been discussed in the literature for nearly 30 years. Pioneering work in the 1980s noted that distinct spectral signatures would be evident in changes in the infrared radiance emitted by the Earth and its atmosphere, and that these could be observed from orbiting satellites. Since then, a number of other studies have advanced the concepts of spectral signatures of climate change. Today the concept of using spectral signatures to identify and attribute atmospheric composition change is firmly accepted and is the foundation of the Climate Absolute Radiance and Refractivity Observatory (CLARREO) satellite mission being developed at NASA. In this work, we will present an overview of the current climate change detection concept using climate model calculations as surrogates for climate change. Any future research work improving the methodology to achieve this concept will be valuable to our society.

Phojanamongkolkij, Nipa↗

2015 Occupant Protection Standing Review Panel

The 2015 Occupant Protection (OP) Risk Standing Review Panel (from here on referred to as the SRP) participated in a WebEx/teleconference with members of the Space Human Factors and Habitability (SHFH) Element, representatives from the Human Research Program (HRP), NASA Headquarters, and NASA Research and Education Support Services on November 3, 2015 (list of participants is in Section VII of this report). The SRP reviewed the updated research plans for the Risk of Injury from Dynamic Loads (OP Risk). The SRP agrees that the Gaps are relevant and appropriate to mitigate the injury risk. All the appropriate and relevant Tasks have been identified to fill the Gaps. Depending upon the findings, additional tasks may need to be identified or modified. Excellent progress has been made since the 2014 SRP meeting. Publications in peer-reviewed journals validate the scientific merit of the research findings. As detailed in this report, the SRP has specific comments, guidance, and information in the following areas: human finite element modeling, human vs. surrogate dynamic responses, chest injury risk curves, matched pair testing of Test device for Human Occupant Restraint (THOR) and Hybrid III, and disc herniation risk analysis.

Steinberg, Susan↗

Uncertainty propagation of p-boxes using sparse polynomial chaos expansions

In modern engineering, physical processes are modelled and analysed using advanced computer simulations, such as finite element models. Furthermore, concepts of reliability analysis and robust design are becoming popular, hence, making efficient quantification and propagation of uncertainties an important aspect. In this context, a typical workflow includes the characterization of the uncertainty in the input variables. In this paper, input variables are modelled by probability-boxes (p-boxes), accounting for both aleatory and epistemic uncertainty. The propagation of p-boxes leads to p-boxes of the output of the computational model. A two-level meta-modelling approach is proposed using non-intrusive sparse polynomial chaos expansions to surrogate the exact computational model and, hence, to facilitate the uncertainty quantification analysis. The capabilities of the proposed approach are illustrated through applications using a benchmark analytical function and two realistic engineering problem settings. They show that the proposed two-level approach allows for an accurate estimation of the statistics of the response quantity of interest using a small number of evaluations of the exact computational model. This is crucial in cases where the computational costs are dominated by the runs of high-fidelity computational models.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Denoising diffusion models with geometry adaptation for high fidelity calorimeter simulation

Simulation is crucial for all aspects of collider data analysis, but the available computing budget in the High Luminosity LHC era will be severely constrained. Generative machine learning models may act as surrogates to replace physics-based full simulation of particle detectors, and diffusion models have recently emerged as the state of the art for other generative tasks. Here, we introduce CaloDiffusion, a denoising diffusion model trained on the public CaloChallenge datasets to generate calorimeter showers. Our algorithm employs 3D cylindrical convolutions, which take advantage of symmetries of the underlying data representation. To handle irregular detector geometries, we augment the diffusion model with a new geometry latent mapping (GLaM) layer to learn forward and reverse transformations to a regular geometry that is suitable for cylindrical convolutions. The showers generated by our approach are nearly indistinguishable from the full simulation, as measured by several different metrics.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

REV-INR: Regularized Evidential Implicit Neural Representation for Uncertainty-Aware Volume Visualization

Applications of Implicit Neural Representations (INRs) have emerged as a promising deep learning approach for compactly representing large volumetric datasets. These models can act as surrogates for volume data, enabling efficient storage and on-demand reconstruction via model predictions. However, conventional deterministic INRs only provide value predictions without insights into the model’s prediction uncertainty or the impact of inherent noisiness in the data. This limitation can lead to unreliable data interpretation and visualization due to prediction inaccuracies in the reconstructed volume. Identifying erroneous results extracted from model-predicted data may be infeasible, as raw data may be unavailable due to its large size. To address this challenge, we introduce REV-INR, Regularized Evidential Implicit Neural Representation, which learns to predict data values accurately along with the associated coordinate-level data uncertainty and model uncertainty using only a single forward pass of the trained REV-INR during inference. By comprehensively comparing and contrasting REV-INR with existing well-established deep uncertainty estimation methods, we show that REV-INR achieves the best volume reconstruction quality with robust data (aleatoric) and model (epistemic) uncertainty estimates using the fastest inference time. Consequently, we demonstrate that REV-INR facilitates assessment of the reliability and trustworthiness of the extracted isosurfaces and volume visualization results, enabling analyses to be solely driven by model-predicted data.

Saklani, Shanu [Indian Institute of Technology, Ka↗

Neural Posterior Estimation for Scalable and Accurate Inverse Parameter Inference in Li-Ion Batteries

Diagnosing the internal state of Li-ion batteries is critical for battery research, operation of real-world systems, and prognostic evaluation of remaining lifetime. By using physics-based models to perform probabilistic parameter estimation via Bayesian calibration, diagnostics can account for the uncertainty due to model fitness, data noise, and the observability of any given parameter. However, Bayesian calibration in Li-ion batteries using electrochemical data is computationally intensive even when using a fast surrogate in place of physics-based models, requiring many thousands of model evaluations. A fully amortized alternative is neural posterior estimation (NPE). NPE shifts the computational burden from the parameter estimation step to data generation and model training, reducing the parameter estimation time from minutes to milliseconds, enabling real-time applications. The present work shows that NPE can infer parameters equally or more accurately than Bayesian calibration, even if it leads to higher voltage reconstruction errors. We also demonstrate that the higher computational costs for data generation are tractable even in high-dimensional cases (ranging from 6 to 27 estimated parameters). The NPE method also offers several interpretability advantages over Bayesian calibration, such as local parameter sensitivity to specific regions of the voltage curve. The NPE method is demonstrated using an experimental fast charge dataset, with parameter estimates validated against measurements of loss of lithium inventory and loss of active material. The implementation is made available in a companion repository (https://github.com/NatLabRockies/BatFIT).

25 ENERGY STORAGE↗

Detailed chemical kinetic mechanism for the oxidation of biodiesel fuels blend surrogate.

Detailed chemical kinetic mechanisms were developed and used to study the oxidation of two large unsaturated esters: methyl-5-decenoate and methyl-9-decenoate. These models were built from a previous methyl decanoate mechanism and were compared with rapeseed oil methyl esters oxidation experiments in a jet stirred reactor. A comparative study of the reactivity of these three oxygenated compounds was performed and the differences in the distribution of the products of the reaction were highlighted showing the influence of the presence and the position of a double bond in the chain. Blend surrogates, containing methyl decanoate, methyl-5-decenoate, methyl-9-decenoate and n-alkanes, were tested against rapeseed oil methyl esters and methyl palmitate/n-decane experiments. These surrogate models are realistic kinetic tools allowing the study of the combustion of biodiesel fuels in diesel and homogeneous charge compression ignition engines.

30 DIRECT ENERGY CONVERSION↗