Search NASA⌕ Search

SEARCH · Search NASA

Results for “numerical algorithms”

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 1,099 records · Page 61

CPOPT : optimization for fitting CANDECOMP/PARAFAC models.

Tensor decompositions (e.g., higher-order analogues of matrix decompositions) are powerful tools for data analysis. In particular, the CANDECOMP/PARAFAC (CP) model has proved useful in many applications such chemometrics, signal processing, and web analysis; see for details. The problem of computing the CP decomposition is typically solved using an alternating least squares (ALS) approach. We discuss the use of optimization-based algorithms for CP, including how to efficiently compute the derivatives necessary for the optimization methods. Numerical studies highlight the positive features of our CPOPT algorithms, as compared with ALS and Gauss-Newton approaches.

99 GENERAL AND MISCELLANEOUS↗

A Comparison of Monte Carlo Particle Transport Algorithms for an Interior Source Binary Stochastic Medium Benchmark Suite

Particle transport through binary stochastic mixtures has received considerable research attention in the last two decades. Zimmerman and Adams proposed a Monte Carlo algorithm (Algorithm A) that solves the Levermore-Pomraning equations and another Monte Carlo algorithm (Algorithm B) that should be more accurate as a result of improved local material realization modeling. Zimmerman and Adams numerically confirmed these aspects of the Monte Carlo algorithms by comparing the reflection and transmission values computed using these algorithms to a standard suite of planar geometry binary stochastic mixture benchmark transport solutions. The benchmark transport problems are driven by an isotropic angular flux incident on one boundary of a binary Markovian statistical planar geometry medium. In a recent paper, we extended the benchmark comparisons of these Monte Carlo algorithms to include the scalar flux distributions produced. This comparison is important, because as demonstrated, an approximate model that gives accurate reflection and transmission probabilities can produce unphysical scalar flux distributions. Brantley and Palmer recently investigated the accuracy of the Levermore-Pomraning model using a new interior source binary stochastic medium benchmark problem suite. In this paper, we further investigate the accuracy of the Monte Carlo algorithms proposed by Zimmerman and Adams by comparing to the benchmark results from the interior source binary stochastic medium benchmark suite, including scalar flux distributions. Because the interior source scalar flux distributions are of an inherently different character than the distributions obtained for the incident angular flux benchmark problems, the present benchmark comparison extends the domain of problems for which the accuracy of these Monte Carlo algorithms has been investigated.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Modeling wave propagation in elastic solids via high-order accurate implicit-mesh discontinuous Galerkin methods

Here, a high-order accurate implicit-mesh discontinuous Galerkin framework for wave propagation in single-phase and bi-phase solids is presented. The framework belongs to the embedded-boundary techniques and its novelty regards the spatial discretization, which enables boundary and interface conditions to be enforced with high-order accuracy on curved embedded geometries. High-order accuracy is achieved via high-order quadrature rules for implicitly-defined domains and boundaries, whilst a cell-merging strategy addresses the presence of small cut cells. The framework is used to discretize the governing equations of elastodynamics, written using a first-order hyperbolic momentum-strain formulation, and an exact Riemann solver is employed to compute the numerical flux at the interface between dissimilar materials with general anisotropic properties. The space-discretized equations are then advanced in time using explicit high-order Runge–Kutta algorithms. Several two- and three-dimensional numerical tests including dynamic adaptive mesh refinement are presented to demonstrate the high-order accuracy and the capability of the method in the elastodynamic analysis of single- and bi-phases solids containing complex geometries.

42 ENGINEERING↗

PyDA: A Hands-On Introduction to Dynamical Data Assimilation with Python

Dynamic data assimilation offers a suite of algorithms that merge measurement data with numerical simulations to predict accurate state trajectories. Meteorological centers rely heavily on data assimilation to achieve trustworthy weather forecast. With the advance in measurement systems, as well as the reduction in sensor prices, data assimilation (DA) techniques are applicable to various fields, other than meteorology. However, beginners usually face hardships digesting the core ideas from the available sophisticated resources requiring a steep learning curve. In this tutorial, we lay out the mathematical principles behind DA with easy-to-follow Python module implementations so that this group of newcomers can quickly feel the essence of DA algorithms. We explore a series of common variational, and sequential techniques, and highlight major differences and potential extensions. We demonstrate the presented approaches using an array of fluid flow applications with varying levels of complexity.

97 MATHEMATICS AND COMPUTING↗

Developing an integrated hub location and revenue management model considering multi-classes of customers in the airline industry

The hub location and revenue management problems are two interesting research fields in the transportation and network design studies. The hub location model designs the transportation network structure and the revenue management model allocates the network’s capacity for customer classes based on their sensitivity to prices. In this paper, we consider the integrated hub location and revenue management problem in the airline industry to maximize the revenue of transportation network and minimize hub installation costs. p hubs have been chosen from a set of n nodes and connected to a central node based on star–star network design. The capacity of the installed links between the hubs and the central node and the hubs and non-hub nodes is limited. This limited capacity has been allocated to the stochastic demands of customer classes using the revenue management approach. A two-stage stochastic model has been derived to determine the locations of hubs and protection levels in the ticket sale process, and an efficient modification of genetic algorithm has been proposed for the large-scale problems. Numerical experiments have been carried out to assess the effectiveness of the proposed algorithm.

97 MATHEMATICS AND COMPUTING↗

A Cartesian-diffusion Langevin method for hybrid kinetic-fluid Coulomb scattering in particle-in-cell plasma simulations

A novel, drag-diffusion Langevin method of hybrid, kinetic-fluid Coulomb scattering in plasmas is presented. Unlike previous methods, the frictional drag is always applied in the simulation frame of reference. The velocity-space diffusion is performed in the stationary-fluid frame of reference when anisotropic, and in the laboratory frame when isotropic. While the general method is mass-ratio independent, we focus on interactions of kinetic-ions and fluid-electrons to show first-order modifications to the electron velocity distribution function that are an important correction for the accurate calculation of electric resistivity. Inclusion of sub-cycling and a limit to the maximum collision frequency is shown to allow for arbitrarily large timesteps without numerical failure. Here the Langevin method is compared with a grid-based binary method and found to require a much less restrictive timestep in cases of ion–electron slowing and temperature equilibration; this finding differs from previous work and is dependent on the mass ratio.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Scalable Methods for Electronic Excitations and Optical Responses in Nanostructures: Mathematics to Algorithms to Observables

The work reported here took place at the University of Minnesota from September 15, 2003 to November 14, 2005. This funding resulted in 10 invited articles or book chapters, 37 articles in refereed journals and 13 invited talks. The funding helped train 5 PhD students. The research supported by this grant focused on developing theoretical methods for predicting and understanding the properties of matter at the nanoscale. Within this regime, new phenomena occur that are characteristic of neither the atomic limit, nor the crystalline limit. Moreover, this regime is crucial for understanding the emergence of macroscopic properties such as ferromagnetism. For example, elemental Fe clusters possess magnetic moments that reside between the atomic and crystalline limits, but the transition from the atomic to the crystalline limit is not a simple interpolation between the two size regimes. To capitalize properly on predicting such phenomena in this transition regime, a deeper understanding of the electronic, magnetic and structural properties of matter is required, e.g., electron correlation effects are enhanced within this size regime and the surface of a confined system must be explicitly included. A key element of our research involved the construction of new algorithms to address problems peculiar to the nanoscale. Typically, one would like to consider systems with thousands of atoms or more, e.g., a silicon nanocrystal that is 7 nm in diameter would contain over 10,000 atoms. Previous ab initio methods could address systems with hundreds of atoms whereas empirical methods can routinely handle hundreds of thousands of atoms (or more). However, these empirical methods often rely on ad hoc assumptions and lack incorporation of structural and electronic degrees of freedom. The key theoretical ingredients in our work involved the use of ab initio pseudopotentials and density functional approaches. The key numerical ingredients involved the implementation of algorithms for solving the Kohn-Sham equation without the use of an explicit basis, i.e., a real space grid. We invented algorithms for a solution of the Kohn-Sham equation based on Chebyshev 'subspace filtering'. Our filtering algorithms dramatically enhanced our ability to explore systems with thousands of atoms, i.e., we examined silicon quantum dots with approximately 11,000 atoms (or 40,000 electrons). We applied this algorithm to a number of nanoscale systems to examine the role of quantum confinement on electronic and magnetic properties: (1) Doping of nanocrystals and nanowires, including both magnetic and non-magnetic dopants and the role of self-purification; (2) Optical excitations and electronic properties of nanocrystals; (3) Intrinsic defects in nanostructures; and (4) The emergence of ferromagnetism from atoms to crystals.

36 MATERIALS SCIENCE↗

Data-Driven Mathematical Modeling of the Effect of Particle Size Distribution on the Transitory Reaction Kinetics of Hot Metal Desulfurization

The aim of this work was to develop a prediction model for hot metal desulfurization. More specifically, the study aimed at finding a set of explanatory variables that are mandatory in prediction of the kinetics of the lime-based transitory desulfurization reaction and evolution of the sulfur content in the hot metal. The prediction models were built through multivariable analysis of process data and phenomena-based simulations. The model parameters for the suggested model types are identified by solving multivariable least-squares cost functions with suitable solution strategies. One conclusion we arrived at was that in order to accurately predict the rate of desulfurization, it is necessary to know the particle size distribution of the desulfurization reagent. It was also observed that a genetic algorithm can be successfully applied in numerical parameter identification of the proposed model type. It was found that even a very simplistic parameterized expression for the first-order rate constant provides more accurate prediction for the end content of sulfur compared to more complex models, if the data set applied for the modeling contains the adequate information.

36 MATERIALS SCIENCE↗

Chaos control of Hastings–Powell model by combining chaotic motions

In this paper, we propose a Parameter Switching (PS) algorithm as a new chaos control method for the Hastings–Powell (HP) system. The PS algorithm is a convergent scheme that switches the control parameter within a set of values while the controlled system is numerically integrated. The attractor obtained with the PS algorithm matches the attractor obtained by integrating the system with the parameter replaced by the averaged value of the switched parameter values. The switching rule can be applied periodically or randomly over a set of given values. In this way, every stable cycle of the HP system can be approximated if its underlying parameter value equalizes the average value of the switching values. Moreover, the PS algorithm can be viewed as a generalization of Parrondo's game, which is applied for the first time to the HP system, by showing that losing strategy can win: “losing + losing = winning.” If “loosing” is replaced with “chaos” and, “winning” with “order” (as the opposite to “chaos”), then by switching the parameter value in the HP system within two values, which generate chaotic motions, the PS algorithm can approximate a stable cycle so that symbolically one can write “chaos + chaos = regular.” Also, by considering a different parameter control, new complex dynamics of the HP model are revealed.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Asynchronous and Distributed Tracking of Time-Varying Fixed Points

NREL researcher Andrey Bernstein discusses asynchronous and distributed tracking of time-varying fixed points as part of the Innovative Optimization and Control Methods for Highly Distributed Autonomous Systems workshop. NREL hosted the workshop to exchange ideas on state-of-the-art control and optimization strategies as well as get feedback on work being done at NREL.

97 MATHEMATICS AND COMPUTING↗

Data-Driven Learning of Nonautonomous Systems

In this work, we present a numerical framework for recovering unknown nonautonomous dynamical systems with time-dependent inputs. To circumvent the difficulty presented by the nonautonomous nature of the system, our method transforms the solution state into piecewise integration of the system over a discrete set of time instances. The time-dependent inputs are then locally parameterized by using a proper model, for example, polynomial regression, in the pieces determined by the time instances. This transforms the original system into a piecewise parametric system that is locally time invariant. We then design a deep neural network structure to learn the local models. Once the network model is constructed, it can be iteratively used over time to conduct global system prediction. We provide theoretical analysis of our algorithm and present a number of numerical examples to demonstrate the effectiveness of the method.

97 MATHEMATICS AND COMPUTING↗

Data-Driven Learning of Non-Autonomous Systems

We present a numerical framework for recovering unknown non-autonomous dynamical systems with time-dependent inputs. To circumvent the difficulty presented by the non-autonomous nature of the system, our method transforms the solution state into piecewise integration of the system over a discrete set of time instances. The time-dependent inputs are then locally parameterized by using a proper model, for example, polynomial regression, in the pieces determined by the time instances. This transforms the original system into a piecewise parametric system that is locally time invariant. We then design a deep neural network structure to learn the local models. Once the network model is constructed, it can be iteratively used over time to conduct global system prediction. We provide theoretical analysis of our algorithm and present a number of numerical examples to demonstrate the effectiveness of the method.

97 MATHEMATICS AND COMPUTING↗

Sparse Cholesky factorization for solving nonlinear PDEs via Gaussian processes

In recent years, there has been widespread adoption of machine learning-based approaches to automate the solving of partial differential equations (PDEs). Among these approaches, Gaussian processes (GPs) and kernel methods have garnered considerable interest due to their flexibility, robust theoretical guarantees, and close ties to traditional methods. They can transform the solving of general nonlinear PDEs into solving quadratic optimization problems with nonlinear, PDE-induced constraints. However, the complexity bottleneck lies in computing with dense kernel matrices obtained from pointwise evaluations of the covariance kernel, and its partial derivatives, a result of the PDE constraint and for which fast algorithms are scarce. The primary goal of this paper is to provide a near-linear complexity algorithm for working with such kernel matrices. We present a sparse Cholesky factorization algorithm for these matrices based on the near-sparsity of the Cholesky factor under a novel ordering of pointwise and derivative measurements. The near-sparsity is rigorously justified by directly connecting the factor to GP regression and exponential decay of basis functions in numerical homogenization. We then employ the Vecchia approximation of GPs, which is optimal in the Kullback-Leibler divergence, to compute the approximate factor. This enables us to compute ϵ-approximate inverse Cholesky factors of the kernel matrices with complexity O(N log d (N/ϵ)) in space and O(N log 2d (N/ϵ)) in time. We integrate sparse Cholesky factorizations into optimization algorithms to obtain fast solvers of the nonlinear PDE. We numerically illustrate our algorithm’s near-linear space/time complexity for a broad class of nonlinear PDEs such as the nonlinear elliptic, Burgers, and Monge-Ampère equations. In summary, we provide a fast, scalable, and accurate method for solving general PDEs with GPs and kernel methods.

97 MATHEMATICS AND COMPUTING↗

A Global, Multi-Resolution Approach to Regional Ocean Modeling

In this collaborative research project between Pennsylvania State University, Colorado State University and Florida State University, we mainly focused on developing multi-resolution algorithms which are suitable to regional ocean modeling. We developed hybrid implicit and explicit adaptive multirate time integration method to solve systems of time-dependent equations that present two ignorantly different scales. We studied the e ects of spatial simplicial meshes on the stability and the conditioning of fully discrete approximations. We also studies adaptive nite element method (AFEM) based upon the Centroidal Voronoi Tessellation (CVT) and superconvergent gradient recovery. Some of these techniques are now being used by geoscientists(such as those at LANL).

58 GEOSCIENCES↗

Using a derivative-free optimization method for multiple solutions of inverse transport problems

Identifying unknown components of an object that emits radiation is an important problem for national and global security. Radiation signatures measured from an object of interest can be used to infer object parameter values that are not known. This problem is called an inverse transport problem. An inverse transport problem may have multiple solutions and the most widely used approach for its solution is an iterative optimization method. This paper proposes a stochastic derivative-free global optimization algorithm to find multiple solutions of inverse transport problems. The algorithm is an extension of a multilevel single linkage (MLSL) method where a mesh adaptive direct search (MADS) algorithm is incorporated into the local phase. Furthermore, numerical test cases using uncollided fluxes of discrete gamma-ray lines are presented to show the performance of this new algorithm.

42 ENGINEERING↗

A scalable algorithm for the optimization of neural network architectures

In this work, we propose a new scalable method to optimize the architecture of an artificial neural network. The proposed algorithm, called Greedy Search for Neural Network Architecture, aims to determine a neural network with minimal number of layers that is at least as performant as neural networks of the same structure identified by other hyperparameter search algorithms in terms of accuracy and computational cost. Numerical results performed on benchmark datasets show that, for these datasets, our method outperforms state-of-the-art hyperparameter optimization algorithms in terms of attainable predictive performance by the selected neural network architecture, and time-to-solution for the hyperparameter optimization to complete.

97 MATHEMATICS AND COMPUTING↗

iVPIC: A low-­dispersion, energy-­conserving relativistic PIC solver for LPI simulations

We have developed a novel low-­dispersion, exactly energy-­conserving PIC algorithm for the relativistic Vlasov-­Maxwell system. The approach features an exact energy conservation theorem while preserving the favorable performance and numerical dispersion properties of explicit PIC. The new algorithm has the potential to enable much longer laser-­plasma-­interaction (LPI) simulations than are currently possible.

97 MATHEMATICS AND COMPUTING↗