Search NASA⌕ Search

SEARCH · Search NASA

Results for “stochastic methods”

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 595 records · Page 33

Multi-scenario modelling of uncertainty in stochastic chemical systems

Uncertainty analysis has not been well studied at the molecular scale, despite extensive knowledge of uncertainty in macroscale systems. The ability to predict the effect of uncertainty allows for robust control of small scale systems such as nanoreactors, surface reactions, and gene toggle switches. However, it is difficult to model uncertainty in such chemical systems as they are stochastic in nature, and require a large computational cost. To address this issue, a new model of uncertainty propagation in stochastic chemical systems, based on the Chemical Master Equation, is proposed in the present study. The uncertain solution is approximated by a composite state comprised of the averaged effect of samples from the uncertain parameter distributions. This model is then used to study the effect of uncertainty on an isomerization system and a two gene regulation network called a repressilator. The results of this model show that uncertainty in stochastic systems is dependent on both the uncertain distribution, and the system under investigation. -- Highlights: •A method to model uncertainty on stochastic systems was developed. •The method is based on the Chemical Master Equation. •Uncertainty in an isomerization reaction and a gene regulation network was modelled. •Effects were significant and dependent on the uncertain input and reaction system. •The model was computationally more efficient than Kinetic Monte Carlo.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Improved Bayesian Infrasonic Source Localization for regional infrasound

The Bayesian Infrasonic Source Localization (BISL) methodology is examined and simplified providing a generalized method of estimating the source location and time for an infrasonic event and the mathematical framework is used therein. The likelihood function describing an infrasonic detection used in BISL has been redefined to include the von Mises distribution developed in directional statistics and propagation-based, physically derived celerity-range and azimuth deviation models. Frameworks for constructing propagation-based celerity-range and azimuth deviation statistics are presented to demonstrate how stochastic propagation modelling methods can be used to improve the precision and accuracy of the posterior probability density function describing the source localization. Infrasonic signals recorded at a number of arrays in the western United States produced by rocket motor detonations at the Utah Test and Training Range are used to demonstrate the application of the new mathematical framework and to quantify the improvement obtained by using the stochastic propagation modelling methods. Moreover, using propagation-based priors, the spatial and temporal confidence bounds of the source decreased by more than 40 per cent in all cases and by as much as 80 per cent in one case. Further, the accuracy of the estimates remained high, keeping the ground truth within the 99 per cent confidence bounds for all cases.

58 GEOSCIENCES↗

RADIATION TRANSPORT FOR EXPLOSIVE OUTFLOWS: OPACITY REGROUPING

Implicit Monte Carlo (IMC) and Discrete Diffusion Monte Carlo (DDMC) are methods used to stochastically solve the radiative transport and diffusion equations, respectively. These methods combine into a hybrid transport-diffusion method we refer to as IMC-DDMC. We explore a multigroup IMC-DDMC scheme that in DDMC, combines frequency groups with sufficient optical thickness. We term this procedure ''opacity regrouping''. Opacity regrouping has previously been applied to IMC-DDMC calculations for problems in which the dependence of the opacity on frequency is monotonic. We generalize opacity regrouping to non-contiguous groups and implement this in SuperNu, a code designed to do radiation transport in high-velocity outflows with non-monotonic opacities. We find that regrouping of non-contiguous opacity groups generally improves the speed of IMC-DDMC radiation transport. We present an asymptotic analysis that informs the nature of the Doppler shift in DDMC groups and summarize the derivation of the Gentile-Fleck factor for modified IMC-DDMC. We test SuperNu using numerical experiments including a quasi-manufactured analytic solution, a simple 10 group problem, and the W7 problem for Type Ia supernovae. We find that opacity regrouping is necessary to make our IMC-DDMC implementation feasible for the W7 problem and possibly Type Ia supernova simulations in general. We compare the bolometric light curves and spectra produced by the SuperNu and PHOENIX radiation transport codes for the W7 problem. The overall shape of the bolometric light curves are in good agreement, as are the spectra and their evolution with time. However, for the numerical specifications we considered, we find that the peak luminosity of the light curve calculated using SuperNu is ∼10% less than that calculated using PHOENIX.

79 ASTRONOMY AND ASTROPHYSICS↗

Space-Time Reduced-Order Modeling for Uncertainty Quantification

This work focuses on the space-time reduced-order modeling (ROM) method for solving large-scale uncertainty quantification (UQ) problems with multiple random coefficients. In contrast with the traditional space ROM approach, which performs dimension reduction in the spatial dimension, the space-time ROM approach performs dimension reduction on both the spatial and temporal domains, and thus enables accurate approximate solutions at a low cost. We incorporate the space-time ROM strategy with various classical stochastic UQ propagation methods such as stochastic Galerkin and Monte Carlo. Numerical results demonstrate that our methodology has significant computational advantages compared to state-of-the-art ROM approaches. By testing the approximation errors, we show that there is no obvious loss of simulation accuracy for space-time ROM given its high computational efficiency.

97 MATHEMATICS AND COMPUTING↗

Markov stochasticity coordinates

Markov dynamics constitute one of the most fundamental models of random motion between the states of a system of interest. Markov dynamics have diverse applications in many fields of science and engineering, and are particularly applicable in the context of random motion in networks. In this paper we present a two-dimensional gauging method of the randomness of Markov dynamics. The method–termed Markov Stochasticity Coordinates–is established, discussed, and exemplified. Also, the method is tweaked to quantify the stochasticity of the first-passage-times of Markov dynamics, and the socioeconomic equality and mobility in human societies.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Measuring alpha eigenvalue of a subcritical system by a intense pulsed neutron source

Intense-Pulsed-Neutron-Source (IPNS) technique is a new approach to measure the subcriticality of a reactor system, which has been theoretically analyzed by us while not been verified in experiment in other reference. In INPS technique the reactivity is derived via the system response to a prompt pulsed neutron source. The method utilizes very intense neutron pulses (about 1010 neutrons/pulse) with the duration ranging from 10 ns to 100 ns, which are generated by the dense plasma focus (DPF) device filled with pure DT mixture as the working gas. The neutron pulse in high intensity provides the opportunity to measure the signal using a scintillator and a photo-multiplier tube (PMT) (the intense pulse radiation measure method) with the flight distance of only twenty centimeters. The attenuation of the dense plasma focus device must be faster than the subcritical system so that the attenuation of the system can be researched. In this paper, after a neutron pulse is injected into the subcritical reactor, the subcriticality is determined by measuring the instantaneous characteristics of the leaking reactor neutron. The Monte Carlo analysis shows that, with the linear arrangement of the dense plasma focus, the subcritical system and the detector, the signal-to-noise ratio at the measure point meets the investigate requirement, in which the 14 MeV neutrons are attenuated by the subcritical system. The detector is close with the subcritical system to reduce the influence of the outer background and the spread of time-of-flight of the neutrons. The measure system utilizing gated detection technology is characterized with fast attenuation time and large linear current, the a value obtained with this method is 2.05 μs{sup -1} while the measurements using the {sup 252}Cf stochastic pulsed source method and the Rossi-a method were 2.18μs{sup -1} and 2.16μs{sup -1}.It showed that the measured result obtained with dense plasma focus instantaneous pulsed source is consistent with those obtained with the other two traditional methods. This new method utilized only a single-shot pulse to determine the a value of the reactor system. Therefore, this quick measure method is recommendable. The influence caused by energy response of the detector should been taken into account since the measured result reflects the time behavior of group particle. The influence is discussed preliminarily in this paper. (authors)

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗

A High-Performance Embedded Hybrid Methodology for Uncertainty Quantification With Applications

Multiphysics processes modeled by a system of unsteady differential equations are naturally suited for partitioned (modular) solution strategies. We consider such a model where probabilistic uncertainties are present in each module of the system and represented as a set of random input parameters. A straightforward approach in quantifying uncertainties in the predicted solution would be to sample all the input parameters into a single set, and treat the full system as a black-box. Although this method is easily parallelizable and requires minimal modifications to deterministic solver, it is blind to the modular structure of the underlying multiphysical model. On the other hand, using spectral representations polynomial chaos expansions (PCE) can provide richer structural information regarding the dynamics of these uncertainties as they propagate from the inputs to the predicted output, but can be prohibitively expensive to implement in the high-dimensional global space of uncertain parameters. Therefore, we investigated hybrid methodologies wherein each module has the flexibility of using sampling or PCE based methods of capturing local uncertainties while maintaining accuracy in the global uncertainty analysis. For the latter case, we use a conditional PCE model which mitigates the curse of dimension associated with intrusive Galerkin or semi-intrusive Pseudospectral methods. After formalizing the theoretical framework, we demonstrate our proposed method using a numerical viscous flow simulation and benchmark the performance against a solely Monte-Carlo method and solely spectral method.

97 MATHEMATICS AND COMPUTING↗

Improved semileptonic form factor calculations in lattice QCD

We investigate the computational efficiency of two stochastic based alternatives to the sequential propagator method used in lattice QCD calculations of heavy-light semileptonic form factors. In the first method, we replace the sequential propagator, which couples the calculation of two of the three propagators required for the calculation, with a stochastic propagator so that the calculations of all three propagators are independent. This method is more flexible than the sequential propagator method but introduces stochastic noise. We study the noise to determine when this method becomes competitive with the sequential propagator method, and find that for any practical calculation it is competitive with or superior to the sequential propagator method. We also examine a second stochastic method, the so-called 'one-end trick', concluding it is relatively inefficient in this context. The investigation is carried out on two gauge field ensembles, using the nonperturbatively improved Wilson-Sheikholeslami-Wohlert action with N{sub f}=2 mass-degenerate sea quarks. The two ensembles have similar lattice spacings but different sea-quark masses. We use the first stochastic method to extract O(a)-improved, matched lattice results for the semileptonic form factors on the ensemble with lighter sea quarks, extracting f{sub +}(0).

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Linear regulator design for stochastic systems by a multiple time-scales method

This short paper develops a hierarchically structured, suboptimal controller for a linear stochastic system composed of fast and slow subsystems. It is proved that the controller is optimal in the limit as the separation of time scales of the subsystems becomes infinite. The methodology is illustrated by design of a controller to suppress the phugoid and short-period modes of the longitudinal dynamics of the F-8 aircraft.

Teneketzis, D.↗

Exploring the performance of spatial stochastic simulation algorithms

Since the publication of Gillespie's direct method, diverse methods have been developed to improve the performance of stochastic simulation methods and to enter the spatial realm. In this paper we discuss a spatial {tau}-leaping variant (S{tau}) that extends the basic leap method. S{tau} takes reaction and both outgoing and incoming diffusion events into account when calculating a leap candidate. A performance analysis shall reveal details on the achieved success in balancing speed and accuracy in comparison to other methods. However, performance analysis of spatial stochastic algorithms requires significant effort - it is crucial to choose suitable (benchmark) models and to carefully define model and simulation setups that take problem and simulation design spaces into account.

97 MATHEMATICS AND COMPUTING↗

Mixed Stochastic-Deterministic Approach for Many-Body Perturbation Theory Calculations

Here, we present an approach for GW calculations of quasiparticle energies with quasi-quadratic scaling by approximating high-energy contributions to the Green's function in its Lehmann representation with effective stochastic vectors. The method is easy to implement without altering the GW code, converges rapidly with stochastic parameters, and treats systems of various dimensionality and screening response. Our calculations on a 5.75° twisted MoS 2 bilayer show how large-scale GW methods include geometry relaxations and electronic correlations on an equal basis in structurally nontrivial materials.

36 MATERIALS SCIENCE↗

Learning generative neural networks with physics knowledge

Deep generative neural networks have enabled modeling complex distributions, but incorporating physics knowledge into the neural networks is still challenging and is at the core of current physics-based machine learning research. To this end, we propose a physics generative neural network (PhysGNN), a new class of generative neural networks for learning unknown distributions in a physical system described by partial differential equations (PDE). PhysGNN couples PDE systems with generative neural networks. It is a fully differentiable model that allows back-propagation of gradients through both numerical PDE solvers and generative neural networks, and is trained by minimizing the discrete Wasserstein distance between generated and observed probability distributions of the PDE outputs using the stochastic gradient descent method. Moreover, PhysGNN does not require adversarial training like standard generative neural networks, which offers better stability than adversarial training. We show that PhysGNN can learn complex distributions in stochastic inverse problems, where conventional methods such as maximum likelihood estimation and momentum matching methods may be inapplicable when little knowledge is known about the form of unknown distributions or the physical model is too complex. Furthermore, our method allows physics-based generative neural network training for learning complex distributions in the context of differential equations.

97 MATHEMATICS AND COMPUTING↗

Robustness analysis of an air heating plant and control law by using polynomial chaos

This paper presents a robustness analysis of an air heating plant with a multivariable closed-loop control law by using the polynomial chaos methodology (MPC). The plant consists of a PVC tube with a fan in the air input (that forces the air through the tube) and a mass flux sensor in the output. A heating resistance warms the air as it flows inside the tube, and a thermo-couple sensor measures the air temperature. The plant has thus two inputs (the fan's rotation intensity and heat generated by the resistance, both measured in percent of the maximum value) and two outputs (air temperature and air mass flux, also in percent of the maximal value). The mathematical model is obtained by System Identification techniques. The mass flux sensor, which is nonlinear, is linearized and the delays in the transfer functions are properly approximated by non-minimum phase transfer functions. The resulting model is transformed to a state-space model, which is used for control design purposes. The multivariable robust control design techniques used is the LQG/LTR, and the controllers are validated in simulation software and in the real plant. Finally, the MPC is applied by considering some of the system's parameters as random variables (one at a time, and the system's stochastic differential equations are solved by expanding the solution (a stochastic process) in an orthogonal basis of polynomial functions of the basic random variables. This method transforms the stochastic equations in a set of deterministic differential equations, which can be solved by traditional numerical methods (That is the MPC). Statistical data for the system (like expected values and variances) are then calculated. The effects of randomness in the parameters are evaluated in the open-loop and closed-loop pole's positions.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

An investigation of Newton-Sketch and subsampled Newton methods

Sketching, a dimensionality reduction technique, has received much attention in the statistics community. In this paper, we study sketching in the context of Newton's method for solving finite-sum optimization problems in which the number of variables and data points are both large. In this work, we study two forms of sketching that perform dimensionality reduction in data space: Hessian subsampling and randomized Hadamard transformations. Each has its own advantages, and their relative tradeoffs have not been investigated in the optimization literature. Additionally, our study focuses on practical versions of the two methods in which the resulting linear systems of equations are solved approximately, at every iteration, using an iterative solver. The advantages of using the conjugate gradient method vs. a stochastic gradient iteration are revealed through a set of numerical experiments, and a complexity analysis of the Hessian subsampling method is presented.

97 MATHEMATICS AND COMPUTING↗

Impact of parametric uncertainty on estimation of the energy deposition into an irradiated brain tumor

Predictions of the total energy deposited into a brain tumor through X-ray irradiation are notoriously error-prone. We investigate how this predictive uncertainty is affected by uncertainty in both the location of the region occupied by a dose-enhancing iodinated contrast agent and the agent's concentration. This is done within the probabilistic framework in which these uncertain parameters are modeled as random variables. We employ the stochastic collocation (SC) method to estimate statistical moments of the deposited energy in terms of statistical moments of the random inputs, and the global sensitivity analysis (GSA) to quantify the relative importance of uncertainty in these parameters on the overall predictive uncertainty. A nonlinear radiation–diffusion equation dramatically magnifies the coefficient of variation of the uncertain parameters, yielding a large coefficient of variation for the predicted energy deposition. This demonstrates that accurate prediction of the energy deposition requires a proper treatment of even small parametric uncertainty. Our analysis also reveals that SC outperforms standard Monte Carlo, but its relative efficiency decreases as the number of uncertain parameters increases from one to three. A robust GSA ameliorates this problem by reducing this number.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Stochastically Realized Observables for Excitonic Molecular Aggregates

In this work, we show that a stochastic approach enables calculations of the optical properties of large 2-dimensional and nanotubular excitonic molecular aggregates. Previous studies of such systems relied on numerically diagonalizing the dense and disordered Frenkel Hamiltonian, which scales approximately as $\mathcal{O}$($N$ 3 ) for $N$ dye molecules. Our approach scales much more efficiently as $\mathcal{O}$($N$ log($N$)), enabling quick study of systems with a million of coupled molecules on the micrometer size scale. We calculate several important experimental observables, including the optical absorption spectrum and density of states, and develop a stochastic formalism for the participation ratio. Quantitative agreement with traditional matrix diagonalization methods is demonstrated for both small- and intermediate-size systems. The stochastic methodology enables the study of the effects of spatial-correlation in site energies on the optical signatures of large 2D aggregates. Our results demonstrate that stochastic methods present a path forward for screening structural parameters and validating experiments and theoretical predictions in large excitonic aggregates.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Deep Generative Models that Solve PDEs: Distributed Computing for Training Large Data-Free Models

Recent progress in scientific machine learning (SciML) has opened up the possibility of training novel neural network architectures that solve complex partial differential equations (PDEs). Several (nearly data free) approaches have been recently reported that successfully solve PDEs, with examples including deep feed forward networks, generative networks, and deep encoder-decoder networks. However, practical adoption of these approaches is limited by the difficulty in training these models, especially to make predictions at large output resolutions (≥1024×1024). Here we report on a software framework for data parallel distributed deep learning that resolves the twin challenges of training these large SciML models - training in reasonable time as well as distributing the storage requirements. Our framework provides several out of the box functionality including (a) loss integrity independent of number of processes, (b) synchronized batch normalization, and (c) distributed higher-order optimization methods. We show excellent scalability of this framework on both cloud as well as HPC clusters, and report on the interplay between bandwidth, network topology and bare metal vs cloud. We deploy this approach to train generative models of sizes hitherto not possible, showing that neural PDE solvers can be viably trained for practical applications. We also demonstrate that distributed higher-order optimization methods are 2-3× faster than stochastic gradient-based methods and provide minimal convergence drift with higher batch-size.

PDEs↗