Search NASA⌕ Search

SEARCH · Search NASA

Results for “Ensemble computing.”

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 1,801 records · Page 100

New Efficient Sparse Space Time Algorithms for Superparameterization on Mesoscales

Superparameterization (SP) is a large-scale modeling system with explicit representation of small-scale and mesoscale processes provided by a cloud-resolving model (CRM) embedded in each column of a large-scale model. New efficient sparse space-time algorithms based on the original idea of SP are presented. The large-scale dynamics are unchanged, but the small-scale model is solved in a reduced spatially periodic domain to save the computation cost following a similar idea applied by one of the authors for aquaplanet simulations. In addition, the time interval of integration of the small-scale model is reduced systematically for the same purpose, which results in a different coupling mechanism between the small- and large-scale models. The new algorithms have been applied to a stringent two-dimensional test suite involving moist convection interacting with shear with regimes ranging from strong free and forced squall lines to dying scattered convection as the shear strength varies. The numerical results are compared with the CRM and original SP. It is shown here that for all of the regimes of propagation and dying scattered convection, the large-scale variables such as horizontal velocity and specific humidity are captured in a statistically accurate way (pattern correlations above 0.75) based on space-time reduction of the small-scale models by a factor of 1/3; thus, the new efficient algorithms for SP result in a gain of roughly a factor of 10 in efficiency while retaining a statistical accuracy on the large-scale variables. Even the models with 1/6 reduction in space-time with a gain of 36 in efficiency are able to distinguish between propagating squall lines and dying scattered convection with a pattern correlation above 0.6 for horizontal velocity and specific humidity. These encouraging results suggest the possibility of using these efficient new algorithms for limited-area mesoscale ensemble forecasting.

99 GENERAL AND MISCELLANEOUS↗

Collision-free poisson motion planning in ultra high-dimensional molecular conformation spaces

The function of protein, RNA, and DNA is modulated by fast, dynamic exchanges between three-dimensional conformations. Conformational sampling of biomolecules with exact and nullspace inverse kinematics, using rotatable bonds as revolute joints and noncovalent interactions as holonomic constraints, can accurately characterize these native ensembles. However, sampling biomolecules remains challenging owing to their ultra-high dimensional configuration spaces, and the requirement to avoid (self-) collisions, which results in low acceptance rates. In this paper, we present two novel mechanisms to overcome these limitations. First, we introduce temporary constraints between near-colliding links. The resulting constraint varieties instantaneously redirect the search for collision-free conformations, and couple motions between distant parts of the linkage. Second, we adapt a randomized Poisson-disk motion planner, which prevents local oversampling and widens the search, to ultra-high dimensions. Tests on several model systems show that the sampling acceptance rate can increase from 16% to 70%, and that the conformational coverage in loop modeling measured as average closeness to existing loop conformations doubled. Finally, correlated protein motions identified with our algorithm agree with those from MD simulations.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

RDX Compression, α→ γ Phase Transition, and Shock Hugoniot Calculations from Density-Functional-Theory-Based Molecular Dynamics Simulations

Prediction of the density and lattice compression properties of the α and γ phases of the hexahydro-1,3,5-trinitro-1,3,5-s-triazine (RDX) crystal and of the low-pressure α → γ phase transition upon pressure increase are general tests used to assess the accuracy of density-functional-theory- (DFT-) based computational methods and to identify the essential parameters that govern the behavior of this high-energy-density material under extreme conditions. The majority of previous DFT studies have analyzed such issues under static optimization conditions by neglecting the corresponding temperature effects. In this paper, we extend previous investigations and analyze the performance of dispersion-corrected density functional theory to predict the compression of RDX in the pressure range of 0–9 GPa and the corresponding α → γ phase transition under realistic temperature and pressure conditions. We demonstrate that, by using static dispersion-corrected density functional theory calculations, direct interconversion between the α and γ phases upon compression is not observed. This limitation can be addressed by using isobaric–isothermal molecular dynamic simulations in conjunction with DFT-D2-calculated potentials, an approach that is shown to provide an accurate description of both the crystallographic RDX lattice parameters and the dynamical effects associated with the α→ γ phase transformation. An even more comprehensive and demanding analysis was done by predicting the corresponding shock Hugoniot curve of RDX in the pressure range of 0–9 GPa. It was found that the theoretical results reproduce reasonably well the available experimental Hugoniot shock data for both the α and γ phases. Finally, the results obtained demonstrate that a satisfactory prediction of the shock properties in high-energy-density materials undergoing crystallographic and configurational transformations is possible through the combined use of molecular dynamics simulations in the isobaric–isothermal ensemble with dispersion-corrected density functional theory methods.

74 ATOMIC AND MOLECULAR PHYSICS↗

SSAGES: Software Suite for Advanced General Ensemble Simulations

Molecular simulation has emerged as an essential tool for modern-day research, but obtaining proper results and making reliable conclusions from simulations requires adequate sampling of the system under consideration. To this end, a variety of methods exist in the literature that can enhance sampling considerably, and increasingly sophisticated, effective algorithms continue to be developed at a rapid pace. Implementation of these techniques, however, can be challenging for experts and non-experts alike. There is a clear need for software that provides rapid, reliable, and easy access to a wide range of advanced sampling methods, and that facilitates implementation of new techniques as they emerge. Here we present SSAGES, a publicly available Software Suite for Advanced General Ensemble Simulations designed to interface with multiple widely used molecular dynamics simulations packages. SSAGES allows facile application of a variety of enhanced sampling techniques—including adaptive biasing force, string methods, and forward flux sampling—that extract meaningful free energy and transition path data from all-atom and coarse grained simulations. A noteworthy feature of SSAGES is a user-friendly framework that facilitates further development and implementation of new methods and collective variables. In this study, the use of SSAGES is illustrated in the context of simple representative applications involving distinct methods and different collective variables that are available in the current release of the suite.

97 MATHEMATICS AND COMPUTING↗

Nationwide Buildings Energy Research enabled through an integrated Data Intensive Scientific Workflow and Advanced Analysis Environment

Modern workflow systems enable scientists to run ensemble simulations at unprecedented scales and levels of complexity, allowing them to study system sizes previously impossible to achieve, due to the inherent resource requirements needed for the modeling work. However as a result of these new capabilities the science teams suddenly also face unprecedented data volumes that they are unable to analyze with their existing tools and methodologies in a timely fashion. In this paper we will describe the ongoing development work to create an integrated data intensive scientific workflow and analysis environment that offers researchers the ability to easily create and execute complex simulation studies and provides them with different scalable methods to analyze the resulting data volumes. The integration of simulation and analysis environments is hereby not only a question of ease of use, but supports fundamental functions in the correlated analysis of simulation input, execution details and derived results for multi-variant, complex studies. To this end the team extended and integrated the existing capabilities of the Velo data management and analysis infrastructure, the MeDICi data intensive workflow system and RHIPE the R for Hadoop version of the well-known statistics package, as well as developing a new visual analytics interface for the result exploitation by multi-domain users. The capabilities of the new environment are demonstrated on a use case that focusses on the Pacific Northwest National Laboratory (PNNL) building energy team, showing how they were able to take their previously local scale simulations to a nationwide level by utilizing data intensive computing techniques not only for their modeling work, but also for the subsequent analysis of their modeling results. As part of the PNNL research initiative PRIMA (Platform for Regional Integrated Modeling and Analysis) the team performed an initial 3 year study of building energy demands for the US Eastern Interconnect domain, which they are now planning to extend to predict the demand for the complete century. The initial study raised their data demands from a few GBs to 400GB for the 3year study and expected tens of TBs for the full century.

99 GENERAL AND MISCELLANEOUS↗

Cycle-to-Cycle Variations in Multi-Cycle Engine RANS Simulations

Reynolds-averaged Navier-Stokes (RANS) modeling is expected to deliver an ensemble-averaged result for the majority of turbulent flows. This could lead to the conclusion that multi-cycle internal combustion engine (ICE) simulations performed using RANS must exhibit a converging numerical solution after a certain number of consecutive cycles. However, for some engine configurations unsteady RANS simulations are not guaranteed to deliver an ensemble-averaged result. In this paper it is shown that, when using RANS modeling to simulate multiple engine cycles, the cycle-to-cycle variations (CCV) generated from different initial conditions at each cycle are not damped out even after a large number of cycles. A single-cylinder GDI research engine is simulated using RANS modeling and the numerical results for 20 consecutive engine cycles are evaluated for two specific operating conditions. One condition is characterized by stoichiometric operation and stable combustion (COVIMEP < 2%) and the other features dilute combustion resulting in misfire events and much higher COVIMEP. In both cases, multi-cycle RANS simulation results show cyclic variability. An in-depth analysis of the most significant physical and chemical quantities highlights that CCV are caused primarily by the variability of the in-cylinder flow. For stoichiometric combustion this does not greatly affect flame propagation, and the resulting fluctuations of the pressure traces are narrow. However, in the event of dilute combustion, the variability of the flow gains more importance, therefore delivering large cyclic fluctuations of in-cylinder pressure. Even though unsteady RANS simulations are not expected to predict as much variability in engine flows as LES, in particular for stable cases, they can still be used to capture typical combustion stability features in an internal combustion engine. This paper shows that the occurrence of CCV when using multi-cycle RANS modeling is not a numerical artifact or the effect of changes in the computational grid from cycle to cycle. Cold-flow analysis reveals that the variability of the simulated flow is intrinsic for the unsteady RANS approach and can only be damped out by increasing numerical viscosity, either by increasing the cell size or by adding up-winding. This is proved in a simplified non-ICE case such as the simulation of the vortex shredding from a cylinder in cross flow. Also, the effect of numerical viscosity on CCV is shown for multi-cycle engine simulations by using a simplified combustion model. Finally, it is shown that multi-cycle RANS can successfully predict the effect of changing specific engine parameters, such as the properties of the ignition system, on cyclic variability and combustion stability.

Scarcelli, Riccardo↗

Inferring subsurface heterogeneity from push-drift tracer tests

In this study, we consider the late-time tailing in a tracer test performed with a push-drift methodology (i.e., quasi-radial injection followed by drift under natural gradient). Numerical simulations of such tests are performed on 1000 multi-Gaussian 2-D log-hydraulic conductivity field realizations of varying heterogeneity, each under eight distinct mean flow directions. The ensemble pdfs of solute return times are found to exhibit power law tails for each considered variance of the log-hydraulic conductivity field, σ$2\atop{ln K}$. The tail exponent is found to relate straightforwardly to σ$2\atop{ln K}$ and, within the parameter space we explored, to be independent of push-phase pumping rate, pumping duration, and local-scale dispersivity. We conjecture that individual push-drift tracer tests in wells with screened intervals much greater than the vertical correlation length of the aquifer will exhibit quasi-ergodicity and that their tail exponent may be used to infer σ$2\atop{ln K}$. Lastly, we calibrate a predictive relationship of this sort from our Monte Carlo study, and apply it to data from a push-drift test performed at a site of approximately known heterogeneity—closely matching the existing best estimate of heterogeneity.

58 GEOSCIENCES↗

A new class of ensemble conserving algorithms for approximate quantum dynamics: Theoretical formulation and model problems

We develop two classes of quasi-classical dynamics that are shown to conserve the initial quantum ensemble when used in combination with the Feynman-Kleinert approximation of the density operator. These dynamics are used to improve the Feynman-Kleinert implementation of the classical Wigner approximation for the evaluation of quantum time correlation functions known as Feynman-Kleinert linearized path-integral. As shown, both classes of dynamics are able to recover the exact classical and high temperature limits of the quantum time correlation function, while a subset is able to recover the exact harmonic limit. A comparison of the approximate quantum time correlation functions obtained from both classes of dynamics is made with the exact results for the challenging model problems of the quartic and double-well potentials. It is found that these dynamics provide a great improvement over the classical Wigner approximation, in which purely classical dynamics are used. In a special case, our first method becomes identical to centroid molecular dynamics.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An Ensemble Score Filter for Tracking High-Dimensional Nonlinear Dynamical Systems

We propose an ensemble score filter (EnSF) for solving high-dimensional nonlinear filtering problems with superior accuracy. A major drawback of existing filtering methods, e.g., particle filters or ensemble Kalman filters, is the low accuracy in handling high-dimensional and highly nonlinear problems. EnSF attacks this challenge by exploiting the score-based diffusion model, defined in a pseudo-temporal domain, to characterizing the evolution of the filtering density. EnSF stores the information of the recursively updated filtering density function in the score function, in stead of storing the information in a set of finite Monte Carlo samples (used in particle filters and ensemble Kalman filters). Unlike existing diffusion models that train neural networks to approximate the score function, we develop a training-free score estimation that uses mini-batch-based Monte Carlo estimator to directly approximate the score function at any pseudo-spatial-temporal location, which provides sufficient accuracy in solving high-dimensional nonlinear problems as well as saves tremendous amount of time spent on training neural networks. Another essential aspect of EnSF is its analytical update step, gradually incorporating data information into the score function, which is crucial in mitigating the degeneracy issue faced when dealing with very high-dimensional nonlinear filtering problems. High-dimensional Lorenz systems are used to demonstrate the performance of our method. EnSF provides surprisingly impressive performance in reliably tracking extremely high-dimensional Lorenz systems (up to 1,000,000 dimension) with highly nonlinear observation processes, which is a well-known challenging problem for existing filtering methods.

97 MATHEMATICS AND COMPUTING↗

An iterative finite difference method for solving the quantum hydrodynamic equations of motion

The quantum hydrodynamic equations of motion associated with the de Broglie-Bohm description of quantum mechanics describe a time evolving probability density whose 'fluid' elements evolve as a correlated ensemble of particle trajectories. These equations are intuitively appealing due to their similarities with classical fluid dynamics and the appearance of a generalized Newton's equation of motion in which the total force contains both a classical and quantum contribution. However, the direct numerical solution of the quantum hydrodynamic equations (QHE) is fraught with challenges: the probability 'fluid' is highly-compressible, it has zero viscosity, the quantum potential ('pressure') is non-linear, and if that weren't enough the quantum potential can also become singular during the course of the calculations. Collectively these properties are responsible for the notorious numerical instabilities associated with a direct numerical solution of the QHE. The most successful and stable numerical approach that has been used to date is based on the Moving Least Squares (MLS) algorithm. The improved stability of this approach is due to the repeated local least squares fitting which effectively filters or reduces the numerical noise which tends to accumulate with time. However, this method is also subject to instabilities if it is pushed too hard. In addition, the stability of the MLS approach often comes at the expense of reduced resolution or fidelity of the calculation (i.e., the MLS filtering eliminates the higher-frequency components of the solution which may be of interest). Recently, a promising new solution method has been developed which is based on an iterative solution of the QHE using finite differences. This method (referred to as the Iterative Finite Difference Method or IFDM) is straightforward to implement, computationally efficient, stable, and its accuracy and convergence properties are well understood. A brief overview of the IFDM will be presented followed by a couple of benchmark applications on one- and two-dimensional Eckart barrier scattering problems.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Interannual Variability of OLR as Observed by AIRS and CERES

This paper compares spatial anomaly time series of OLR (Outgoing Longwave Radiation) and OLR(sub CLR) (Clear Sky OLR) as determined using observations from CERES Terra and AIRS over the time period September 2002 through June 2011. Both AIRS and CERES show a significant decrease in global mean and tropical mean OLR over this time period. We find excellent agreement of the anomaly time-series of the two OLR data sets in almost every detail, down to 1 deg X 1 deg spatial grid point level. The extremely close agreement of OLR anomaly time series derived from observations by two different instruments implies that both sets of results must be highly stable. This agreement also validates to some extent the anomaly time series of the AIRS derived products used in the computation of the AIRS OLR product. The paper also examines the correlations of anomaly time series of AIRS and CERES OLR, on different spatial scales, as well as those of other AIRS derived products, with that of the NOAA Sea Surface Temperature (SST) product averaged over the NOAA Nino-4 spatial region. We refer to these SST anomalies as the El Nino Index. Large spatially coherent positive and negative correlations of OLR anomaly time series with that of the El Nino Index are found in different spatial regions. Anomalies of global mean, and especially tropical mean, OLR are highly positively correlated with the El Nino Index. These correlations explain that the recent global and tropical mean decreases in OLR over the period September 2002 through June 2011, as observed by both AIRS and CERES, are primarily the result of a transition from an El Nino condition at the beginning of the data record to La Nina conditions toward the end of the data period. We show that the close correlation of global mean, and especially tropical mean, OLR anomalies with the El Nino Index can be well accounted for by temporal changes of OLR within two spatial regions which lie outside the NOAA Nino-4 region, in which anomalies of cloud cover and mid-tropospheric water vapor are both highly negatively correlated with the El Nino Index. Agreement of the AIRS and CERES OLR(sub CLR) anomaly time series is less good, which may be a result of the large sampling differences in the ensemble of cases included in each OLR(sub CLR) data set.

Susskind, Joel↗

Polarized Radiative Transfer of a Cirrus Cloud Consisting of Randomly Oriented Hexagonal Ice Crystals: The 3 x 3 Approximation for Non-Spherical Particles

The reflection and transmission of polarized light for a cirrus cloud consisting of randomly oriented hexagonal columns were calculated by two very different vector radiative transfer models. The forward peak of the phase function for the ensemble-averaged ice crystals has a value of order 6 x 10(exp 3) so a truncation procedure was used to help produce numerically efficient yet accurate results. One of these models, the Vectorized Line-by-Line Equivalent model (VLBLE), is based on the doubling- adding principle, while the other is based on a vector discrete ordinates method (VDISORT). A comparison shows that the two models provide very close although not entirely identical results, which can be explained by differences in treatment of single scattering and the representation of the scattering phase matrix. The relative differences in the reflected I and Q Stokes parameters are within 0.5 for I and within 1.5 for Q for all viewing angles. In 1971 Hansen showed that for scattering by spherical particles the 3 x 3 approximation is sufficient to produce accurate results for the reflected radiance I and the degree of polarization (DOP), and he conjectured that these results would hold also for non-spherical particles. Simulations were conducted to test Hansen's conjecture for the cirrus cloud particles considered in this study. It was found that the 3 x 3 approximation also gives accurate results for the transmitted light, and for Q and U in addition to I and DOP. For these non-spherical ice particles the 3 x 3 approximation leads to an absolute error 2 x 10(exp -6) for the reflected and transmitted I, Q and U Stokes parameters. Hence, it appears to be an excellent approximation, which significantly reduces the computational complexity and burden required for multiple scattering calculations.

Polarization↗

Autonomous Radiological Survey Systems for Locating Fired Depleted Uranium Munitions - 19212

An ensemble of autonomous radiological survey platforms that utilize a suite of scintillation detectors that can be used to conduct radiological site characterization, assist with remediation efforts, or conduct site clearance in accordance with the Multi-Agency Radiologic Survey and Site Investigation Manual are described. The overall systems are composed of inorganic scintillators, light detection and ranging, global positioning, inertial guidance, wireless telemetry, and signal processing systems. Two primary designs are discussed: a three-meter wide platform and multiple half-meter-wide systems. The survey systems are used to conduct radiological surveys over large areas autonomously, without human involvement. The three-meter system has been designed to cover large, open areas free of obstacles. Smaller robotic platforms are designed to handle a large range of terrain types and to conduct surveys in areas that are plagued with obstacles, such as bushes and trees. The primary design criteria for these systems were accurate locating of fired depleted uranium (DU) munitions. Several U.S. Army facilities have remnant DU material on active firing ranges where DU rounds have been used in performance evaluations. Many ranges also have unexploded ordinance (UXO) challenges which are problematic for conducting walkover surveys. The combination of having active DU impacted firing ranges with UXO make conducting site characterization surveys very difficult and hazardous. The autonomous surveying systems allow radiological scanning data collection to be scheduled around range activities and reduces human exposure to ionizing radiation and UXO. All of the robotic platforms act independently, and are controlled by a centralized computer system via a wireless telemetry system. The central computer system is responsible for setting up survey units, survey paths, and scheduling which platform is to complete the scheduled survey. Unit selection and scheduling are completed by technicians, and the work is conducted autonomously by these survey platforms. The survey areas and survey paths are transmitted to each platform through the telemetry system. Radiological and positional data are collected by the systems and transmitted back to the central computer system in real-time. Collected data is saved to a database for post-survey analysis or can be analyzed in real-time. Analysis of the radiological data is used to construct radiation intensity and estimated DU concentration maps. Automatically generated reports for each survey unit will provide range managers with: survey coverage data maps, estimated uranium concentration maps, and a detailed list of areas that are above a user defined concentration. This paper describes the components of each of the surveying systems, their capabilities, and limitations. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Predicting permeation of compounds across the outer membrane of P. aeruginosa using molecular descriptors

The ability Gram-negative pathogens have at adapting and protecting themselves against antibiotics has increasingly become a public health threat. Data-driven models identifying molecular properties that correlate with outer membrane (OM) permeation and growth inhibition while avoiding efflux could guide the discovery of novel classes of antibiotics. Here we evaluate 174 molecular descriptors in 1260 antimicrobial compounds and study their correlations with antibacterial activity in Gram-negative Pseudomonas aeruginosa. The descriptors are derived from traditional approaches quantifying the compounds’ intrinsic physicochemical properties, together with, bacterium-specific from ensemble docking of compounds targeting specific MexB binding pockets, and all-atom molecular dynamics simulations in different subregions of the OM model. Using these descriptors and the measured inhibitory concentrations, we design a statistical protocol to identify predictors of OM permeation/inhibition. We find consistent rules across most of our data highlighting the role of the interaction between the compounds and the OM. An implementation of the rules uncovered in our study is shown, and it demonstrates the accuracy of our approach in a set of previously unseen compounds. Our analysis sheds new light on the key properties drug candidates need to effectively permeate/inhibit P. aeruginosa, and opens the gate to similar data-driven studies in other Gram-negative pathogens.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Monte Carlo chord length sampling for d-dimensional Markov binary mixtures

The Chord Length Sampling (CLS) algorithm is a powerful Monte Carlo method that models the effects of stochastic media on particle transport by generating on-the-fly the material interfaces seen by the random walkers during their trajectories. This annealed disorder approach, which formally consists of solving the approximate Levermore–Pomraning equations for linear particle transport, enables a considerable speed-up with respect to transport in quenched disorder, where ensemble-averaging of the Boltzmann equation with respect to all possible realizations is needed. However, CLS intrinsically neglects the correlations induced by the spatial disorder, so that the accuracy of the solutions obtained by using this algorithm must be carefully verified with respect to reference solutions based on quenched disorder realizations. When the disorder is described by Markov mixing statistics, such comparisons have been attempted so far only for one-dimensional geometries, of the rod or slab type. In this work we extend these results to Markov media in two-dimensional (extruded) and three-dimensional geometries, by revisiting the classical set of benchmark configurations originally proposed by Adams, Larsen and Pomraning and extended by Brantley. In particular, we examine the discrepancies between CLS and reference solutions for scalar particle flux and transmission/reflection coefficients as a function of the material properties of the benchmark specifications and of the system dimensionality.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Variational deep learning of equilibrium transition path ensembles

Here, we present a time-dependent variational method to learn the mechanisms of equilibrium reactive processes and efficiently evaluate their rates within a transition path ensemble. This approach builds off of the variational path sampling methodology by approximating the time-dependent commitment probability within a neural network ansatz. The reaction mechanisms inferred through this approach are elucidated by a novel decomposition of the rate in terms of the components of a stochastic path action conditioned on a transition. This decomposition affords an ability to resolve the typical contribution of each reactive mode and their couplings to the rare event. The associated rate evaluation is variational and systematically improvable through the development of a cumulant expansion. We demonstrate this method in both over- and under-damped stochastic equations of motion, in low-dimensional model systems, and in the isomerization of a solvated alanine dipeptide. In all examples, we find that we can obtain quantitatively accurate estimates of the rates of the reactive events with minimal trajectory statistics and gain unique insights into transitions through the analysis of their commitment probability.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Resolving turbulent magnetohydrodynamics: a hybrid operator-diffusion framework

We present a hybrid machine learning framework that combines physics-informed neural operators (PINOs) with score-based generative diffusion models to simulate the full spatio-temporal evolution of two-dimensional, incompressible, resistive magnetohydrodynamic turbulence across a broad range of Reynolds numbers (Re). The framework leverages the equation-constrained generalization capabilities of PINOs to predict coherent, low-frequency dynamics, while a conditional diffusion model stochastically corrects high-frequency residuals, enabling accurate modeling of fully developed turbulence. Trained on a comprehensive ensemble of high-fidelity simulations with Re ϵ {100, 250, 500, 750, 1000, 3000, 10000}, the approach achieves state-of-the-art accuracy in regimes previously inaccessible to deterministic surrogates. At Re = 1000 and 3000, the model faithfully reconstructs the full spectral energy distributions of both velocity and magnetic fields late into the simulation, capturing non-Gaussian statistics, intermittent structures, and cross-field correlations with high fidelity. At extreme turbulence levels (Re = 10 000), it remains the first surrogate capable of recovering the high-wavenumber evolution of the magnetic field, preserving large-scale morphology and enabling statistically meaningful predictions.

Diffusion-Integrated Neural Operators↗

Inverse design of photonic surfaces via multi fidelity ensemble framework and femtosecond laser processing

We demonstrate a multi-fidelity (MF) machine learning ensemble framework for the inverse design of photonic surfaces, trained on a dataset of 11,759 samples that we fabricate using high throughput femtosecond laser processing. The MF ensemble combines an initial low fidelity model for generating design solutions, with a high fidelity model that refines these solutions through local optimization. The combined MF ensemble can generate multiple disparate sets of laser-processing parameters that can each produce the same target input spectral emissivity with high accuracy (root mean squared errors < 2%). SHapley Additive exPlanations analysis shows transparent model interpretability of the complex relationship between laser parameters and spectral emissivity. Finally, the MF ensemble is experimentally validated by fabricating and evaluating photonic surface designs that it generates for improved efficiency energy harvesting devices. Our approach provides a powerful tool for advancing the inverse design of photonic surfaces in energy harvesting applications.

97 MATHEMATICS AND COMPUTING↗

Refine your search to explore more results.