Search NASA⌕ Search

SEARCH · Search NASA

Results for “adaptive cross approximation”

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 73 records · Page 4

Bridging the Gap between Cosmological Simulations with Graph Neural Networks and Domain Adaptation

Deep learning models have been shown to outperform methods that rely on summary statistics, like the power spectrum, in extracting information from complex cosmological data sets. However, due to differences in the subgrid physics implementation and numerical approximations across different simulation suites, models trained on data from one cosmological simulation show a drop in performance when tested on another. Similarly, models trained on any of the simulations would also likely experience a drop in performance when applied to observational data. Training on data from two different suites of the CAMELS hydrodynamic cosmological simulations, we examine the generalization capabilities of Domain Adaptive Graph Neural Networks (DA-GNNs). By utilizing GNNs, we capitalize on their capacity to capture structured scale-free cosmological information from galaxy distributions. Moreover, by including unsupervised domain adaptation via Maximum Mean Discrepancy (MMD), we enable our models to extract domain-invariant features. We demonstrate that DA-GNN achieves higher accuracy and robustness on cross dataset tasks (up to 28% better relative error and up to almost an order of magnitude better χ 2 ). Using data visualizations, we show the effects of domain adaptation on proper latent space data alignment. This shows that DA-GNNs are a promising method for extracting domain-independent cosmological information, a vital step toward robust deep learning for real cosmic survey data.

97 MATHEMATICS AND COMPUTING↗

On approaches to analyze the sensitivity of simulated hydrologic fluxes to model parameters in the community land model

Here, effective sensitivity analysis approaches are needed to identify important parameters or factors and their uncertainties in complex Earth system models composed of multi-phase multi-component phenomena and multiple biogeophysical-biogeochemical processes. In this study, the impacts of 10 hydrologic parameters in the Community Land Model on simulations of runoff and latent heat flux are evaluated using data from a watershed. Different metrics, including residual statistics, the Nash-Sutcliffe coefficient, and log mean square error, are used as alternative measures of the deviations between the simulated and field observed values. Four sensitivity analysis (SA) approaches, including analysis of variance based on the generalized linear model, generalized cross validation based on the multivariate adaptive regression splines model, standardized regression coefficients based on a linear regression model, and analysis of variance based on support vector machine, are investigated. Results suggest that these approaches show consistent measurement of the impacts of major hydrologic parameters on response variables, but with differences in the relative contributions, particularly for the secondary parameters. The convergence behaviors of the SA with respect to the number of sampling points are also examined with different combinations of input parameter sets and output response variables and their alternative metrics. This study helps identify the optimal SA approach, provides guidance for the calibration of the Community Land Model parameters to improve the model simulations of land surface fluxes, and approximates the magnitudes to be adjusted in the parameter values during parametric model optimization.

58 GEOSCIENCES↗

Domain Adaptive Graph Neural Networks for Constraining Cosmological Parameters Across Multiple Data Sets

Deep learning models have been shown to outperform methods that rely on summary statistics, like the power spectrum, in extracting information from complex cosmological data sets. However, due to differences in the subgrid physics implementation and numerical approximations across different simulation suites, models trained on data from one cosmological simulation show a drop in performance when tested on another. Similarly, models trained on any of the simulations would also likely experience a drop in performance when applied to observational data. Training on data from two different suites of the CAMELS hydrodynamic cosmological simulations, we examine the generalization capabilities of Domain Adaptive Graph Neural Networks (DA-GNNs). By utilizing GNNs, we capitalize on their capacity to capture structured scale-free cosmological information from galaxy distributions. Moreover, by including unsupervised domain adaptation via Maximum Mean Discrepancy (MMD), we enable our models to extract domain-invariant features. We demonstrate that DA-GNN achieves higher accuracy and robustness on cross-dataset tasks. Using data visualizations, we show the effects of domain adaptation on proper latent space data alignment. This shows that DA-GNNs are a promising method for extracting domain-independent cosmological information, a vital step toward robust deep learning for real cosmic survey data.

79 ASTRONOMY AND ASTROPHYSICS↗

Domain Adaptive Graph Neural Networks for Constraining Cosmological Parameters Across Multiple Data Sets

Deep learning models have been shown to outperform methods that rely on summary statistics, like the power spectrum, in extracting information from complex cosmological data sets. However, due to differences in the subgrid physics implementation and numerical approximations across different simulation suites, models trained on data from one cosmological simulation show a drop in performance when tested on another. Similarly, models trained on any of the simulations would also likely experience a drop in performance when applied to observational data. Training on data from two different suites of the CAMELS hydrodynamic cosmological simulations, we examine the generalization capabilities of Domain Adaptive Graph Neural Networks (DA-GNNs). By utilizing GNNs, we capitalize on their capacity to capture structured scale-free cosmological information from galaxy distributions. Moreover, by including unsupervised domain adaptation via Maximum Mean Discrepancy (MMD), we enable our models to extract domain-invariant features. We demonstrate that DA-GNN achieves higher accuracy and robustness on cross-dataset tasks (up to $28\%$ better relative error and up to almost an order of magnitude better $\chi^2$). Using data visualizations, we show the effects of domain adaptation on proper latent space data alignment. This shows that DA-GNNs are a promising method for extracting domain-independent cosmological information, a vital step toward robust deep learning for real cosmic survey data.

79 ASTRONOMY AND ASTROPHYSICS↗

Effect of parallel flow on resonant layer responses in high beta plasmas

Abstract Resonant layers in a tokamak respond to non-axisymmetric magnetic perturbations by amplifying the mode amplitude and balancing the plasma rotation through magnetic reconnection and force balance, respectively. This resonant response can be characterized by local layer parameters and especially by a single quantity in the linear regime, the so-called inner-layer Δ. The computation of Δ under two-fluid drift-MHD formalism has been progressed by reducing the order of the system in the phase space, where the shielding current is approximated as being only carried by electrons, a posteriori . In this study, we relax the approximation and compute Δ accounted for by the parallel flow associated with the ion shielding current. The posteriori is numerically verified in great agreement with the original SLAYER developed in a previous paper (J.-K. Park 2022 Phys. Plasmas 29 072506). Extending the resonant layer response theory to high β plasmas, our research findings answer two important questions: how the parallel flow influences the resonant layer response and why the parallel flow effect appears in high β plasmas. The complicated plasma compression in high β regime allows the parallel flow response to give rise to the ion shielding current, which not only shifts the zero-crossing condition of the ExB flow but also enhances the field penetration threshold. Technically, the Riccati matrix transformation method is adapted to handle the numerical stiffness due to the increased order of the system. The high fidelity of this numerical method makes use of further extension of the model to higher-order systems to take other physical phenomena into account. This work is envisaged to predict the resonant layer response under high β fusion reactor conditions.

Lee, Yeongsun (ORCID:000000034474416X)↗

Reduction of Raman Spectroscopy Data for H2-CO2-Air Tubular Flame Measurements

An approach of reducing spontaneous Raman spectroscopy data, based on the matrix inversion method which logically removes undesired sources of the Raman scattering signals including cross-talk due to the signals from other species, is discussed. The approach contains extensive calibration of temperature-dependent system response. The Raman spectra of several cold and hot calibration flames and their resulting calibration matrix are elaborated. The structure of H2-CO2-air tubular flames, in terms of major species (CO2, O2, N2, H2O, and H2) concentrations and temperature, was successfully measured with the optical technique. The Raman spectra data used in this study are taken from a uniquely designed experimental apparatus – a tubular burner and Raman spectroscopy system at Vanderbilt University. The characteristic uncertainties for chemical species are approximately ±2% by mole fraction for hot products and ±0.5% for room temperature reactants. The approach described here for hydrogen/air tubular flames may be adapted to hydrocarbon/air tubular flames when other necessary major species, for example, CH4 and CO for methane/air tubular flame, are considered in the whole process.

Raman Spectroscopy↗

Changes in intervertebral disc cross-sectional area with bed rest and space flight

STUDY DESIGN. We measured the cross-sectional area of the intervertebral discs of normal volunteers after an overnight rest; before, during, and after 5 or 17 weeks of bed rest; and before and after 8 days of weightlessness. OBJECTIVES. This study sought to determine the degree of expansion of the lumbar discs resulting from bed rest and space flight. SUMMARY OF BACKGROUND DATA. Weightlessness and bed rest, an analog for weightlessness, reduce the mechanical loading on the musculoskeletal system. When unloaded, intervertebral discs will expand, increasing the nutritional diffusion distance and altering the mechanical properties of the spine. METHODS. Magnetic resonance imaging was used to measure the cross-sectional area and transverse relaxation time (T2) of the intervertebral discs. RESULTS. Overnight or longer bed rest causes expansion of the disc area, which reaches an equilibrium value of about 22% (range 10-40%) above baseline within 4 days. Increases in disc area were associated with modest increases in disc T2. During bed rest, disc height increased approximately 1 mm, about one-half of previous estimates based on body height measurements. After 5 weeks of bed rest, disc area returned to baseline within a few days of ambulation, whereas after 17 weeks, disc area remained above baseline 6 weeks after reambulation. After 8 days of weightlessness, T2, disc area, and lumbar length were not significantly different from baseline values 24 hours after landing. CONCLUSIONS. Significant adaptive changes in the intervertebral discs can be expected during weightlessness. These changes, which are rapidly reversible after short-duration flights, may be an important factor during and after long-duration missions.

Non-NASA Center↗

Analysis of nystagmus response to a pseudorandom velocity input

Space motion sickness was not reported during the first Apollo missions; however, since Apollo 8 through the current Shuttle and Skylab missions, approximately 50% of the crewmembers have experienced instances of space motion sickness. Space motion sickness, renamed space adaptation syndrome, occurs primarily during the initial period of a mission until habilation takes place. One of NASA's efforts to resolve the space adaptation syndrome is to model the individual's vestibular response for basis knowledge and as a possible predictor of an individual's susceptibility to the disorder. This report describes a method to analyse the vestibular system when subjected to a pseudorandom angular velocity input. A sum of sinusoids (pseudorandom) input lends itself to analysis by linear frequency methods. Resultant horizontal ocular movements were digitized, filtered and transformed into the frequency domain. Programs were developed and evaluated to obtain the (1) auto spectra of input stimulus and resultant ocular resonse, (2) cross spectra, (3) the estimated vestibular-ocular system transfer function gain and phase, and (4) coherence function between stimulus and response functions.

Lessard, C. S.↗

The human as a detector of changes in variance and bandwidth

The detection of changes in random process variance and bandwidth was studied. Psychophysical thresholds for these two parameters were determined using an adaptive staircase technique for second order random processes at two nominal periods (1 and 3 seconds) and damping ratios (0.2 and 0.707). Thresholds for bandwidth changes were approximately 9% of nominal except for the (3sec,0.2) process which yielded thresholds of 12%. Variance thresholds averaged 17% of nominal except for the (3sec,0.2) process in which they were 32%. Detection times for suprathreshold changes in the parameters may be roughly described by the changes in RMS velocity of the process. A more complex model is presented which consists of a Kalman filter designed for the nominal process using velocity as the input, and a modified Wald sequential test for changes in the variance of the residual. The model predictions agree moderately well with the experimental data. Models using heuristics, e.g. level crossing counters, were also examined and are found to be descriptive but do not afford the unification of the Kalman filter/sequential test model used for changes in mean.

Curry, R. E.↗

Evaluation of the Emission, Transport, and Deposition of Mercury and Fine Particulate Matter from Coal-Based Power Plants in the Ohio River Valley Region

As stated in the proposal: Ohio University, in collaboration with CONSOL Energy, Advanced Technology Systems, Inc (ATS) and Atmospheric and Environmental Research, Inc. (AER) as subcontractors, evaluated the impact of emissions from coal-fired power plants in the Ohio River Valley region as they relate to the transport and deposition of mercury and associated fine particulate matter. This evaluation involved two interrelated areas of effort: ambient air monitoring and regional-scale modeling analysis. The scope of work for the ambient air monitoring included the deployment of a surface air monitoring (SAM) station in southeastern Ohio. The SAM station contains sampling equipment to collect and measure mercury (including speciated forms of mercury and wet and dry deposited mercury), particulate matter (PM) mass, PM composition, and gaseous criteria pollutants (CO, NOx, SO2, O3, etc.). Laboratory analyses of time-integrated samples were used to obtain chemical speciation of ambient PM composition and mercury in precipitation. Nearreal- time measurements were used to measure the ambient concentrations of PM mass and all gaseous species including Hg0 and RGM. Approximately 30 months of field data were collected at the SAM site to validate the proposed regional model simulations for episodic and seasonal model runs. The ambient air quality data provides mercury, and fine particulate matter data that can be used by Ohio Valley industries to assess performance on multi-pollutant control systems. The scope of work for the modeling analysis includes (1) development of updated inventories of mercury emissions from coal plants and other important sources in the modeled domain; (2) adapting an existing 3-D atmospheric chemical transport model to incorporate recent advancements in the understanding of mercury transformations in the atmosphere; (3) analyses of the flux of Hg0, RGM, and fine particulate matter in the different sectors of the study region to identify key transport mechanisms; (4) comparison of cross correlations between species from the model results to observations in order to evaluate characteristics of specific air masses associated with long-range transport from a specified source region; and (5) evaluation of the sensitivity of these correlations to emissions from regions along the transport path. This is accomplished by multiple model runs with emissions simulations switched on and off from the various source regions. To the greatest extent possible, model results were compared to field data collected at other air monitoring sites in the Ohio Valley region, operated independently of this project. These sites may include (1) the DOE National Energy Technologies Laboratory’s monitoring site at its suburban Pittsburgh, PA facility; (2) sites in Pittsburgh (Lawrenceville) PA and Holbrook, PA operated by ATS; (3) sites in Steubenville, OH and Pittsburgh, PA operated by the USEPA and/or its contractors; and (4) sites operated by State or local air regulatory agencies. Field verification of model results and predictions provides critical information for the development of cost effective air pollution control strategies by the coal-fired power plants in the Ohio River Valley region.

01 COAL, LIGNITE, AND PEAT↗

Metabolic and morphologic properties of single muscle fibers in the rat after spaceflight, Cosmos 1887

The adaptation of a slow (soleus, Sol) and a fast (medial gastrocnemius, MG) skeletal muscle to spaceflight was studied in five young male rats. The flight period was 12.5 days and the rats were killed approximately 48 h after returning to 1 g. Five other rats that were housed in cages similar to those used by the flight rats were maintained at 1 g for the same period of time to serve as ground-based controls. Fibers were classified as dark or light staining for myosin adenosine triphosphatase (ATPase). On the average, the fibers in the Sol of the flight rats atrophied twice as much as those in the MG. Further, the fibers located in the deep (close to the bone and having the highest percentage of light ATPase and high oxidative fibers in the muscle cross section) region of the MG atrophied more than the fibers located in the superficial (away from the bone and having the lowest percentage of light ATPase and high oxidative fibers in the muscle cross-section) region of the muscle. Based on quantitative histochemical assays of single muscle fibers, succinate dehydrogenase (SDH) activity per unit volume was unchanged in fibers of the Sol and MG. However, in the Sol, but not the MG, the total amount of SDH activity in a 10-microns-thick section of a fiber decreased significantly in response to spaceflight. Based on population distributions, it appears that the alpha-glycerophosphate dehydrogenase (GPD) activities were elevated in the dark ATPase fibers in the Sol, whereas the light fibers in the Sol and both fiber types in the MG did not appear to change. The ratio of GPD to SDH activities increased in the dark (but not light) fibers of the Sol and was unaffected in the MG. Immunohistochemical analyses indicate that approximately 40% of the fibers in the Sol of flight rats expressed a fast myosin heavy chain compared with 22% in control rats. Further, 31% of the fibers in the Sol of flight rats expressed both fast and slow myosin heavy chains compared with 8% in control rats. Immunohistochemical changes in the MG were minimal. These data suggest that the magnitude and direction of enzymatic activity and cell volume changes are dependent on the muscle, the region of the muscle, and the type of myosin expressed in the fibers. Further, the ability of fibers to maintain normal or even elevated activities per unit volume of some metabolic enzymes is remarkable considering the marked and rapid decrease in fiber volume.

NASA Experiment Number COS 1887-26↗

Correlation-based network analysis combined with machine learning techniques highlight the role of the GABA shunt in Brachypodium sylvaticum freezing tolerance

Perennial grasses will account for approximately 16 billion gallons of renewable fuels by the year 2022, contributing significantly to carbon and nitrogen sequestration. However, perennial grasses productivity can be limited by severe freezing conditions in some geographical areas, although these risks could decrease with the advance of climate warming, the possibility of unpredictable early cold events cannot be discarded. We conducted a study on the model perennial grass Brachypodium sylvaticum to investigate the molecular mechanisms that contribute to cold and freezing adaption. The study was performed on two different B. sylvaticum accessions, Ain1 and Osl1, typical to warm and cold climates, respectively. Both accessions were grown under controlled conditions with subsequent cold acclimation followed by freezing stress. For each treatment a set of morphological parameters, transcription, metabolite, and lipid profiles were measured. State-of-the-art algorithms were employed to analyze cross-component relationships. Phenotypic analysis revealed higher adaption of Osl1 to freezing stress. Our analysis highlighted the differential regulation of the TCA cycle and the GABA shunt between Ain1 and Osl1. Osl1 adapted to freezing stress by repressing the GABA shunt activity, avoiding the detrimental reduction in fatty acid biosynthesis and the concomitant detrimental effects on membrane integrity.

54 ENVIRONMENTAL SCIENCES↗

Modeling error PDF optimization based wavelet neural network modeling of dynamic system and its application in blast furnace ironmaking

In general, the modeling errors of dynamic system model are a set of random variables. The traditional performance index of modeling such as means square error (MSE) and root means square error (RMSE) can not fully express the connotation of modeling errors with stochastic characteristics both in the dimension of time domain and space domain. Therefore, the probability density function (PDF) is introduced to completely describe the modeling errors in both time scales and space scales. Based on it, a novel wavelet neural network (WNN) modeling method is proposed by minimizing the two-dimensional (2D) PDF shaping of modeling errors. First, the modeling error PDF by the tradional WNN is estimated using data-driven kernel density estimation (KDE) technique. Then, the quadratic sum of 2D deviation between the modeling error PDF and the target PDF is utilized as performance index to optimize the WNN model parameters by gradient descent method. Since the WNN has strong nonlinear approximation and adaptive capability, and all the parameters are well optimized by the proposed method, the developed WNN model can make the modeling error PDF track the target PDF, eventually. Simulation example and application in a blast furnace ironmaking process show that the proposed method has a higher modeling precision and better generalization ability compared with the conventional WNN modeling based on MSE criteria. Furthermore, the proposed method has more desirable estimation for modeling error PDF that approximates to a Gaussian distribution whose shape is high and narrow.

Modelling↗

Full wave modeling of radio-frequency beams in tokamaks in the electron cyclotron frequency range

Simulation of full wave, without paraxial approximation, high-resolution solution of wave equations in frequency domain in the electron cyclotron resonance (ECR) frequency range for realistic Tokamak plasma parameters became possible by using recently formulated hybrid iterative algorithm [Svidzinski et al., Phys. Plasmas 25, 082509 (2018)] for numerically solving discretized wave equations. This approach combines time evolution and iterative relaxation techniques into iteration cycles. This algorithm is implemented in 2D code FullWave, solving wave equations in Tokamaks in cold and hot plasma models, and it has been tested in 3D full wave iterative RF beams simulation tool, which is presently being developed to model 3D ECRH RF beams in fusion devices using dynamic grid adaptation. The results of 2D full wave modeling, assuming specified toroidal mode number, of ECRH RF beams in DIII-D plasma, performed in the cold and hot plasma models for outboard and top launch scenarios using FullWave are presented. Nonlocal hot plasma response model, based on accurate numerical solution of linearized Vlasov equation, is used to model beam propagation and absorption in the 2nd electron cyclotron harmonic region. Demonstration of capability of the hybrid iterative algorithm to model ECRH RF beams in 3D is made by simulating a substantial part of realistic beam in DIII-D, launched from outboard side of the machine. All relevant physics of RF beam propagation, most of which is not captured in paraxial approximation, such as beam's divergence, interference between the X and O modes in the beam, X-O mode conversion, beam splitting into the X and O mode beams, transformation of beam's cross section, and absorption at the 2nd electron cyclotron harmonic, is captured in the simulations. A numerical technique to find an optimal beam polarization at the launcher to launch a nearly pure X or O mode beam in plasma is developed and tested.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

A new methodology for free wake analysis using curved vortex elements

A method using curved vortex elements was developed for helicopter rotor free wake calculations. The Basic Curve Vortex Element (BCVE) is derived from the approximate Biot-Savart integration for a parabolic arc filament. When used in conjunction with a scheme to fit the elements along a vortex filament contour, this method has a significant advantage in overall accuracy and efficiency when compared to the traditional straight-line element approach. A theoretical and numerical analysis shows that free wake flows involving close interactions between filaments should utilize curved vortex elements in order to guarantee a consistent level of accuracy. The curved element method was implemented into a forward flight free wake analysis, featuring an adaptive far wake model that utilizes free wake information to extend the vortex filaments beyond the free wake regions. The curved vortex element free wake, coupled with this far wake model, exhibited rapid convergence, even in regions where the free wake and far wake turns are interlaced. Sample calculations are presented for tip vortex motion at various advance ratios for single and multiple blade rotors. Cross-flow plots reveal that the overall downstream wake flow resembles a trailing vortex pair. A preliminary assessment shows that the rotor downwash field is insensitive to element size, even for relatively large curved elements.

Bliss, Donald B.↗

Mitigating Headward Fluid Shifts with Venoconstrictive Thigh Cuffs during Spaceflight

Venoconstrictive thigh cuffs (VTC) are a mechanical countermeasure capable of attenuating the spaceflight induced headward fluid shift, and thus may be a viable spaceflight associated neuro-ocular syndrome (SANS) countermeasure. Crewmembers can use VTC to mitigate the headward fluid shift to aid in adapting to spaceflight. However, data are needed to determine if VTC affect ocular structures. PURPOSE The purpose of this study is to determine the efficacy of long duration use of VTC application to mitigate the spaceflight-induced headward fluid shift. We hypothesize that a VTC countermeasure will temporarily reverse the headward fluid shift and attenuate spaceflight-induced changes of internal jugular vein (IJV) cross-sectional area, IJV pressure, stroke volume, cardiac output, intraocular pressure (IOP), and optic nerve head and retinal morphology. METHODS This study will evaluate the effectiveness of VTC countermeasure application on the headward fluid shift, as well as cardiovascular and ocular variables. VTC during spaceflight will be worn for an extended duration (up to 6 hours) with data collected at three time points (30 minutes, 3 hours, and 6 hours) to characterize the temporal profile of key fluid shift outcome measures of the vascular fluid shift, IOP, and ocular structure changes. Ten astronauts will be recruited to participate and will be studied before and during approximately 180-day International Space Station (ISS) spaceflight missions. Baseline data collection will occur approximately 90-days before launch (Figure). Preflight, each leg of the crewmember will be measured to determine the VTC cuff size and a cuff fit check session will occur prior to the preflight baseline ground imaging to verify the appropriate fit measured via a surface contact pressure. Prior to donning the VTC, baseline measures without VTC will be collected seated, supine, and supine followed by data collection with the VTC. The baseline data collection will include ultrasound (IJV area and pressure, stroke volume and cardiac output), brachial blood pressure and heart rate, optical coherence tomography (OCT) imaging (total retinal thickness and choroid thickness), and IOP. An inflight cuff fit check session, same as the preflight fit check, will occur prior to VTC use on ISS. The inflight VTC experiment will be conducted early (FD45) and late (R-45) to determine if mission duration affects VTC fit and the efficacy of fluid redistribution. A system usability scale comfort questionnaire will be included in each VTC session to capture feedback from crewmembers regarding the comfort and usability of the VTC. SCIENTIFIC & MISSION IMPACT Results will narrow knowledge gaps described in the Human Research Roadmap to mitigate the headward fluid shift during spaceflight and help NASA to 1) determine the efficacy of extended use VTC application to mitigate the spaceflight-induced headward fluid shift and 2) further the understanding of the use of VTC on vascular fluid shifts, IOP, and ocular structure during spaceflight. Supported by NASA Human Research Program Directed Research. Figure. Detailed Testing Schedule. Baseline data collection at L-90 will be performed in a randomized order of position.

J V Jasien↗

Ground Tests on the BOLT Geometry at Mach 7: Cross-Facility Comparison and Stability Analysis

Transition measurements have been obtained through two experimental campaigns conducted independently by the German Aerospace Center and the French Aerospace Lab with the French Alternative Energies and Atomic Energy Commission on subscale models of the BOLT-1 flight experiment geometry. This paper details a cross-facility comparison of measurements obtained at Mach 7, as well as subsequent computational analysis. Infrared (IR) thermography measurements obtained by both campaigns have facilitated a global comparison of the transition front across facilities at analagous conditions, which are found to be in good agreement. High-frequency surface pressure fluctuation data demonstrate significant amplification of instabilities with Mack-mode characteristics in the outboard regions of the acreage. These measurements are compared to stability analyses of varying fidelity. The computational methods employed to characterize the boundary-layer transition phenomena include the traditional line-marching implementation of the Parabolized Stability Equations (PSE), 2D eigenvalue analysis coupled with PSE, and Adaptive Mesh Refinement Wavepacket Tracking. While the 2D eigenvalue analysis is found to predict instabilities, which correlate in terms of frequency and acreage location to the experimental measurements, the predicted amplification for these instabilities is lower than would typically be expected for transition. Line-marching results for traveling crossflow produce the best match to the experimental transition front, with a consistent transition N factor of approximately 3-3.5.

Boundary Layer Transition↗

Multi-fidelity Bayesian neural networks: Algorithms and applications

Here we propose a new class of Bayesian neural networks (BNNs) that can be trained using noisy data of variable fidelity, and we apply them to learn function approximations as well as to solve inverse problems based on partial differential equations (PDEs). These multi-fidelity BNNs consist of three neural networks: The first is a fully connected neural network, which is trained following the maximum a posteriori probability (MAP) method to fit the low-fidelity data; the second is a Bayesian neural network employed to capture the cross-correlation with uncertainty quantification between the low- and high-fidelity data; and the last one is the physics-informed neural network, which encodes the physical laws described by PDEs. For the training of the last two neural networks, we first employ the mean-field variational inference (VI) to maximize the evidence lower bound (ELBO) to obtain informative prior distributions for the hyperparameters in the BNNs, and subsequently we use the Hamiltonian Monte Carlo (HMC) method to estimate accurately the posterior distributions for the corresponding hyperparameters. We demonstrate the accuracy of the present method using synthetic data as well as real measurements. Specifically, we first approximate a one- and four-dimensional function, and then infer the reaction rates in one- and two-dimensional diffusion-reaction systems. Moreover, we infer the sea surface temperature (SST) in the Massachusetts and Cape Cod Bays using satellite images and in-situ measurements. Taken together, our results demonstrate that the present method can capture both linear and nonlinear correlation between the low- and high-fidelity data adaptively, identify unknown parameters in PDEs, and quantify uncertainties in predictions, given a few scattered noisy high-fidelity data. Finally, we demonstrate that we can effectively and efficiently reduce the uncertainties and hence enhance the prediction accuracy with an active learning approach, using as examples a specific one-dimensional function approximation and an inverse PDE problem.

97 MATHEMATICS AND COMPUTING↗