SEARCH · Engineering Papers
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.
SABATH provides benchmarking infrastructure for evaluating scientific ML/AI models. It contains support for scientific machine learning surrogates from external repositories. The software dependences are explicitly exposed in the surrogate model definition, which allows the use of advanced optimization, communication, and hardware features.
Tensor datatypes representing field variables like stress, displacement, velocity, etc., have increasingly become a common occurrence in data-driven modeling and analysis of simulations. Numerous methods [such as convolutional neural networks (CNNs)] exist to address the meta-modeling of field data from simulations. As the complexity of the simulation increases, so does the cost of acquisition, leading to limited data scenarios. Modeling of tensor datatypes under limited data scenarios remains a hindrance for engineering applications. Here, in this article, we introduce a direct image-to-image modeling framework of convolutional autoencoders enhanced by information bottleneck loss function to tackle the tensor data types with limited data. The information bottleneck method penalizes the nuisance information in the latent space while maximizing relevant information making it robust for limited data scenarios. The entire neural network framework is further combined with robust hyperparameter optimization. We perform numerical studies to compare the predictive performance of the proposed method with a dimensionality reduction-based surrogate modeling framework on a representative linear elastic ellipsoidal void problem with uniaxial loading. The data structure focuses on the low-data regime (fewer than 100 data points) and includes the parameterized geometry of the ellipsoidal void as the input and the predicted stress field as the output. The results of the numerical studies show that the information bottleneck approach yields improved overall accuracy and more precise prediction of the extremes of the stress field. Additionally, an in-depth analysis is carried out to elucidate the information compression behavior of the proposed framework.
Groundwater model structural error is ubiquitous, due to simplification and/or misrepresentation of real aquifer systems. During model calibration, the basic hydrogeological parameters may be adjusted to compensate for structural error. This may result in biased predictions when such calibrated models are used to forecast aquifer responses to new forcing. Here, we investigate the impact of model structural error on calibration and prediction of a real-world groundwater flow model, using a Bayesian method with a data-driven error model to explicitly account for model structural error. The error-explicit Bayesian method jointly infers model parameters and structural error and thereby reduces parameter compensation. In this study, Bayesian inference is facilitated using high performance computing and fast surrogate models (based on machine learning techniques) as a substitute for the computationally expensive groundwater model. We demonstrate that with explicit treatment of model structural error, the Bayesian method yields parameter posterior distributions that are substantially different from those derived using classical Bayesian calibration that does not account for model structural error. We also found that the error-explicit Bayesian method gives signficantly more accurate prediction along with reasonable credible intervals. Finally, through variance decomposition, we provide a comprehensive assessment of prediction uncertainty contributed from parameter, model structure, and measurement uncertainty. The results suggest that the error-explicit Bayesian approach provides a solution to real-world modeling applications for which data support the presence of model structural error, yet model deficiency cannot be specifically identified or corrected.
Estimating marginal likelihood is of central importance to Bayesian model selection and/or model averaging. The nested sampling method has been recently used together with the Metropolis-Hasting (M-H) sampling algorithm for estimating marginal likelihood. This study develops a new implementation of nested sampling by using the DiffeRential Evolution Adaptive Metropolis (DREAMzs) sampling algorithm. The two implementations of nested sampling are evaluated for four models of a synthetic groundwater flow modeling. The DREAMzs-based nested sampling outperforms the M-H-based nested sampling in terms of their accuracy, convergence, efficiency, and stability, which is attributed to the fact that DREAMzs is more robust than M-H for parameter sampling. The nested sampling method is also compared with four other methods (arithmetic mean, harmonic mean, stabilized harmonic mean, and thermodynamic integration) commonly used for estimating marginal likelihood and posterior probability of the four groundwater models. The comparative study requires hundreds of millions of model executions, which would not be possible without using surrogate models to replace the original models. Using the arithmetic mean estimator as the reference, the comparison reveals that thermodynamic integration outperforms nested sampling in terms of accuracy and stability, whereas nested sampling is more computationally efficient to reach to convergence. Here, the harmonic mean and stabilized harmonic mean methods give the worst marginal likelihood estimation. Accurate marginal likelihood estimation is important for accurate estimation of posterior model probability and better predictive performance of Bayesian model averaging.
Control of both the magnitude and the shape of tokamak profiles will be necessary to achieve stable, high-performance plasmas. In order to reject disturbances in real time, feedback-control algorithms rely on accurate real-time knowledge of the plasma state. When diagnostics alone are insufficient, either because they are limited in number or their measurements are too noisy, observers can be used to combine diagnostic data with a response model to provide a better estimation of different plasma properties. An observer has been developed to estimate the electron temperature profile in real time using both diagnostic data from the Thomson scattering system and a model based on the electron heat transport equation describing the evolution of the electron temperature profile. Here, neural network surrogate models are leveraged to help improve the overall model prediction while staying within computation time constraints for real-time use. The observer algorithm is shown in offline tests to produce smooth profiles that are consistent with both the diagnostic data and the electron heat transport equation. When implemented into the real-time plasma control system, this observer will provide valuable information on the electron temperature profile to many potential feedback-control applications.
Surrogate nuclear explosive debris was synthesized and characterized for major, minor, and trace elemental composition as well as uranium isotopics. The samples consisted of an urban glass matrix, equal masses soda lime and cement, doped with 500 ppm uranium with varying enrichments. The surface and cross section morphology were measured with SEM, and the major elemental composition was determined by XPS. LA-ICP-MS was used to measure the uranium isotopic abundance comparing different sampling techniques. Furthermore, the results provide an example of the utility of LA-ICP-MS for forensics applications.
Prompt analysis of elemental and isotopic information from post-detonation nuclear debris is critical for rapid attribution analysis. The capabilities of dissolution methods that use ammonium bifluoride (ABF, NH 4 HF 2 ) and pressurized microwave digestion with HF acid are reported for NIST Surrogate Post-detonation Urban Debris (SPUD). Elemental concentration and isotope ratios were measured using inductively coupled plasma mass spectrometry. The SPUD material was also analyzed for U concentration using instrumental neutron activation analysis. Finally, the ABF dissolution method with a 10:1 ABF to sample ratio and the pressurized microwave digestion methods resulted in comparable results.
In this work, we present two novel approaches for predicting the power evolution and control rod height of the Transient Reactor Test Facility (TREAT) to support experiment modeling; specifically, transient analysis of the NASA-sponsored Sirius series of experiments. These approaches utilize steady-state Monte Carlo model, point kinetics model, and surrogate models to predict power evolution and control rod axial position during transient experiments. Both approaches were tested and validated against several Sirius experiments that were performed in the TREAT facility at different power levels. The validation test results show very good agreement with the experimental data, and the models were able to accurately predict the power evolution and the axial control rod position with an average error within 3.0%. Finally, this indicates that these approaches will help the reactor engineering team of the TREAT facility in preparing and predicting the power and temperature of the experiment.
Recent characterization of radioactive particles indicate that a large percentage of the radioactivity observed during the Fukushima Daiichi nuclear meltdown was insoluble 137 Cs bound within silica microparticles. Therefore, much of the decontamination research performed prior to the Fukushima incident that used either soluble radionuclides deposited onto wet surfaces or large (~100 μm) particles characteristic of nuclear weapons fallout do not accurately represent the characteristics of potential contamination. Thus, the common practice of extrapolating radioactive decontamination methods generically to all radioactive release events is, at best, suspect. In response, a method to produce chemically-inert, radiolabeled silica particles was developed. Binding 152 Eu within a sodium silicate coating required proper temperature control and ethanol was beneficial as a volatile dispersant to limit residues. In the end, a step-wise method, which first deposited 152 Eu or 241 Am as a nitrate salt, decomposed the salt to a sesquioxide, and finally coated the surface with sodium silicate led to dispersed particles of the desired 2 or 0.5 μm diameters. Dynamic light scattering and scanning election microscopy confirmed the particle size was unchanged. Leaching studies into several common decontaminants were performed to ensure particle inertness. In this paper, our approach allows for substitution of other radionuclides making it a robust, simple, and novel method to produce inert particle surrogates for a release event that allows direct comparison of decontamination techniques and contaminant fate studies, greatly aiding the development of response and recovery plans.
Widespread adoption of high-temperature polymer electrolyte membrane electrochemical systems, such as fuel cells (HT-PEMFCs), requires models and computational tools for accurate optimization and guiding new materials for enhancing performance and durability. In this contribution, knowledge-based modelling and data-driven modelling are combined using Few-Shot Learning and implementing an Automated Machine Learning framework for the generation of Machine Learning-based surrogate models. Applicability of the resulting model for derivative-free optimization is demonstrated. Additionally, a way of considering extrapolation in the optimization task is presented. Results show that although extrapolation is needed to achieve better solutions during optimization, it can be monitored and managed. As a result, tuning the electrode ionomer binder's properties, such as ionic conductivity, in the fuel cell represents a promising pathway for improving HT-PEMFC performance.
PyApprox is a Python-based one-stop-shop for probabilistic analysis of numerical models such as those used in the earth, environmental and engineering sciences. Easy to use and extendable tools are provided for constructing surrogates, sensitivity analysis, Bayesian inference, experimental design, and forward uncertainty quantification. The algorithms implemented represent a wide range of methods for model analysis developed over the past two decades, including recent advances in multi-fidelity approaches that use multiple model discretizations and/or simplified physics to significantly reduce the computational cost of various types of analyses. An extensive set of Benchmarks from the literature is also provided to facilitate the easy comparison of new or existing algorithms for a wide range of model analyses. Here, this paper introduces PyApprox and its various features, and presents results demonstrating the utility of PyApprox on a benchmark problem modeling the advection of a tracer in groundwater.
High-temperature Arrhenius ignition delay time correlations are useful for revealing the underlying parameter dependencies of combustion models, for simplifying and optimizing combustion mechanisms for use in engine simulations, for scaling experimental data to new conditions for comparison purposes, and for guiding in experimental design. Here, we have developed a scaling relationship for Fatty Acid Methyl Ester (FAME) ignition time data taken at high temperatures in 4%O 2 /Ar mixtures behind reflected shocks using an aerosol shock tube: τ ign [ms] = 2.24 x 10 -6 [ms] (P [atm]) -.41 ($\phi$) 0.30 (C n ) -.61 x exp $$ \left(\frac{37.1 [kcal/mol]}{\hat{R}_u [kcal / mol K] T [K]}\right) $$ In addition, we have combined our ignition delay time data for methyl decanoate, methyl palmitate, methyl oleate, and methyl linoleate with other experimental results in the literature in order to derive fuel-specific oxygen-mole-fraction scaling parameters for these surrogates. In conclusion, in this article, we discuss the significance of the parameter values, compare our correlation to others found in the literature for different classes of fuels, and contrast the above expression’s performance with correlations obtained using leading FAME kinetic models in 4%O 2 /Ar mixtures.
Combustion of a seven-component surrogate for a research grade 87 octane gasoline mixed with 10% ethanol is investigated experimentally and numerically from the perspective of an isolated droplet burning under conditions that promote one-dimensional gas transport. The numerical analysis included a kinetic mechanism comprised of 398 species and 24,814 reactions and a soot model that accounted for nucleation, surface growth, coalescence/aggregation of soot particles, and luminous flame radiation. Measurements of droplet and flame diameters were made for an initial droplet diameter (D o ) of approximately 0.63 mm. The simulations agreed well with the measurements including the location of the soot shell. Preferential vaporization was revealed by simulations of the liquid concentrations in the droplet. Predicted peak soot volume fractions coincided with temperatures between 1300 K and 1400 K as a soot inception temperature. Simulations were also carried out for D o between 0.25 mm and 5 mm to explore the effect of radiation and D o on burning. Below 0.25 mm radiation was negligible and burning rates and flame temperatures converged to a single value. Increasing D o up to 1.8 mm lowered the burning rate with luminous radiation having a strong effect. When radiation was entirely removed from the model the burning rate was nearly constant. Above D o = 2 mm droplets extinguished almost immediately after ignition. The flame temperature decreased with increasing D o while it increased when radiation was omitted. As a result, the simulations show that soot precursors including polyaromatic hydrocarbons were concentrated around the soot shell.
As modern gasoline direct injection (GDI) engines utilize sophisticated injection strategies, a detailed understanding of the air-fuel mixing process is crucial to further improvements in engine emission and fuel economy. In this study, a comprehensive evaluation of the spray process of single-component iso-octane (IC8) and multi-component gasoline surrogate E00 (36 % n-pentane, 46 % iso-octane, and 18 % n-undecane, by volume) fuels was conducted using an Engine Combustion Network (ECN) Spray G injector. High-speed extinction, schlieren, and microscopy imaging campaigns were carried out under engine-like ambient conditions in a spray vessel. Experimental results including liquid/vapor penetration, local liquid volume fraction, droplet size, and projected liquid film on the nozzle tip were compared under ECN G1 (573 K, 3.5 kg/m 3 ), G2 (333 K, 0.5 kg/m 3 ), and G3 (333 K, 1.01 kg/m 3 ) conditions. In addition to the experiments, preferential evaporation process of the E00 fuel was elucidated by Large–Eddy Simulations (LES). The three-dimensional liquid volume fraction measurement enabled by the computed tomographic reconstruction showed substantial plume collapse for E00 under the G2 and G3 conditions having wider plume growth and plume-to-plume interaction due to the fuel high vapor pressure. Here, the CFD simulation of E00 showed an inhomogeneity in the way fuel components vaporized, with more volatile components carried downstream in the spray after the end of injection. The high vapor pressure of E00 also results in ~4 μm smaller average droplet diameter than IC8, reflecting a higher rate of initial vaporization even though the final boiling point temperature is higher. Consistent with high vapor pressure, E00 had a wider plume cone angle and enhanced interaction with the wall to cover the entire surface of the nozzle tip in a film. However, the liquid fuel underwent faster evaporation, so the final projected tip wetting area was smaller than the IC8 under the flash-boiling condition.
In this work we introduce a manifold learning-based surrogate modeling framework for uncertainty quantification in high-dimensional stochastic systems. Our first goal is to perform data mining on the available simulation data to identify a set of low-dimensional (latent) descriptors that efficiently parameterize the response of the high-dimensional computational model. To this end, we employ Principal Geodesic Analysis on the Grassmann manifold of the response to identify a set of disjoint principal geodesic submanifolds, of possibly different dimension, that captures the variation in the data. Since operations on the Grassmann require the data to be concentrated, we propose an adaptive algorithm based on Riemannian K-means and the minimization of the sample Fréchet variance on the Grassmann manifold to identify “local” principal geodesic submanifolds that represent different system behavior across the parameter space. Polynomial chaos expansion is then used to construct a mapping between the random input parameters and the projection of the response on these local principal geodesic submanifolds. Here, the method is demonstrated on four test cases, a toy-example that involves points on a hypersphere, a Lotka-Volterra dynamical system, a continuous-flow stirred-tank chemical reactor system, and a two-dimensional Rayleigh-Bénard convection problem.
Here, we present a machine learning (ML)-based surrogate model using convolutional neural networks (CNN) designed to emulate the attenuation of neutron fields as they pass through various shielding materials. This model can compute the outgoing neutron flux almost instantaneously and achieves reasonable accuracy compared to traditional Monte Carlo (MC)-based codes, which are computationally intensive. This emulator alleviates the complexity of neutron radiation transport through shielding materials by reducing the dimensionality and enables shielding optimization for a known radiation environment. This optimization process, which would have taken an unrealistic timeline due to several complex radiation transport simulations, can now be achieved in minutes, thus increasing computational capabilities in radiation shielding assessment. We demonstrate the applications of this emulator in computing effective dose rates and optimizing shielding solutions for a heavy-ion accelerator facility, such as the Facility for Rare Isotope Beams, where secondary neutrons produced via beam interactions dominate the radiation environment.
The issue of fuel fragmentation, relocation, and dispersal is critical in the licensing and use of high-burnup (>62 GWd/MTU) nuclear fuel in light water reactors (LWRs). In this work, two test series are reported that examine the fragment dispersal during a burst event and an additional dispersal following the burst due to vibrations in the rod, such as those induced by accident recovery systems. To examine dispersal during the balloon and burst portion of a Loss-of-Coolant Accident event, HfO 2 fragments and yttria-stabilized zirconia pellets were filled into an as-fabricated cladding tube, which was then pressurized and subjected to loss-of-coolant accident testing in steam. Results from this test were found to be highly non-prototypic, and dispersal was both significantly more violent and significantly greater in magnitude than identified for actual fuel tests. These findings were attributed to the conservative (more dispersive) nature of the particles chosen for dispersal and to the details of the test conditions used that led to particularly wide bursts. Further, the second set of tests examined dispersal following the burst when vibrations were induced in the rod, primarily via recovery activities such as Emergency Core Cooling System actuation leading to rapid water addition. Post-burst dispersal testing was performed by inducing sinusoidal oscillations with 2–25 nm peak-to-peak amplitude and 2–5 Hz frequencies in pre-burst rods that had been refilled with HfO 2 fragments or high-burnup fragment surrogate mixture of HfO 2 fragments and yttria-stabilized zirconia sands. Testing revealed that rods with large burst openings (7 mm wide in this work) led to unmitigated dispersal from above the burst zone but that smaller bursts (5 mm wide), although still much larger than the mean fragment size of 3 mm, led to effectively no dispersal because of interparticle locking. Additionally, mixture and moisture were found to impact the amount of dispersal: mixtures increased dispersal, and moisture drastically reduced it. The implications of these findings on likely dispersal from actual fuel are discussed.
Here this work reports an experimental and modeling study on the chemical kinetic interactions of NO with a multi-component gasoline surrogate, namely PACE-20, using a twin-piston rapid compression machine at a stochiometric fuel loading with 20% EGR (exhaust gas recirculation) by mass, pressures of 20 and 40 bar, and temperatures from 700 to 930 K. Five NO concentrations are investigated, namely 0, 20, 50, 70 and 150 ppm, where NO addition effects are characterized through changes in PACE-20 ignition reactivity and heat release characteristics. Experiments indicate that within the low-temperature regime, NO promotes low-temperature heat release rate and main ignition reactivity at low addition levels, with saturation or even inhibiting effects observed at >50 ppm NO addition, while within the NTC/intermediate-temperature regime, adding NO only promotes reactivity. A recently updated, detailed chemical kinetic model with chemistry specific to NOx/hydrocarbons interaction incorporated is used to simulate the experiments, and reasonable agreement is obtained. In-depth sensitivity and rate of production analyses are further performed. The results indicate that NO interacts with PACE-20 via two types of interaction: (a) direct interactions between NO and PACE-20 derivatives, primarily through NO+HO 2 ↔NO 2 +OH and RO 2 +NO↔RO+NO 2 , and (b) indirect interactions between PACE-20 derivatives and NO 2 produced from the direct interactions, primarily through R+NO 2 ↔RO+NO. The observed NO inhibiting effect at low temperatures and 150 ppm NO addition is attributed to the lack of HO 2 radicals to sustain NO consumption via NO+HO 2 ↔NO 2 +OH, and the take-up of inhibiting pathways via RO 2 +NO↔RO+NO 2 . The results also indicate that even with the presence of multiple fuel components, NOx/hydrocarbons interactions are highly selective, and are mainly initiated by the interactions between NO and RO 2 radicals from cyclopentane and ethanol, as well as between NO 2 and R radicals from toluene, 1,2,4-trimethylbenzene and 1-hexene. Further studies on these interactive reactions are therefore highly recommended.