Search NASA⌕ Search

SEARCH · Search NASA

Results for “Smoothing 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 1,261 records · Page 70

One shot methods for optimal control of distributed parameter systems 1: Finite dimensional control

The efficient numerical treatment of optimal control problems governed by elliptic partial differential equations (PDEs) and systems of elliptic PDEs, where the control is finite dimensional is discussed. Distributed control as well as boundary control cases are discussed. The main characteristic of the new methods is that they are designed to solve the full optimization problem directly, rather than accelerating a descent method by an efficient multigrid solver for the equations involved. The methods use the adjoint state in order to achieve efficient smoother and a robust coarsening strategy. The main idea is the treatment of the control variables on appropriate scales, i.e., control variables that correspond to smooth functions are solved for on coarse grids depending on the smoothness of these functions. Solution of the control problems is achieved with the cost of solving the constraint equations about two to three times (by a multigrid solver). Numerical examples demonstrate the effectiveness of the method proposed in distributed control case, pointwise control and boundary control problems.

Taasan, Shlomo↗

SU-F-R-38: Impact of Smoothing and Noise On Robustness of CBCT Textural Features for Prediction of Response to Radiotherapy Treatment of Head and Neck Cancers

Purpose: To examine the impact of image smoothing and noise on the robustness of textural information extracted from CBCT images for prediction of radiotherapy response for patients with head/neck (H/N) cancers. Methods: CBCT image datasets for 14 patients with H/N cancer treated with radiation (70 Gy in 35 fractions) were investigated. A deformable registration algorithm was used to fuse planning CT’s to CBCT’s. Tumor volume was automatically segmented on each CBCT image dataset. Local control at 1-year was used to classify 8 patients as responders (R), and 6 as non-responders (NR). A smoothing filter [2D Adaptive Weiner (2DAW) with 3 different windows (ψ=3, 5, and 7)], and two noise models (Poisson and Gaussian, SNR=25) were implemented, and independently applied to CBCT images. Twenty-two textural features, describing the spatial arrangement of voxel intensities calculated from gray-level co-occurrence matrices, were extracted for all tumor volumes. Results: Relative to CBCT images without smoothing, none of 22 textural features extracted showed any significant differences when smoothing was applied (using the 2DAW with filtering parameters of ψ=3 and 5), in the responder and non-responder groups. When smoothing, 2DAW with ψ=7 was applied, one textural feature, Information Measure of Correlation, was significantly different relative to no smoothing. Only 4 features (Energy, Entropy, Homogeneity, and Maximum-Probability) were found to be statistically different between the R and NR groups (Table 1). These features remained statistically significant discriminators for R and NR groups in presence of noise and smoothing. Conclusion: This preliminary work suggests that textural classifiers for response prediction, extracted from H&N CBCT images, are robust to low-power noise and low-pass filtering. While other types of filters will alter the spatial frequencies differently, these results are promising. The current study is subject to Type II errors. A much larger cohort of patients is needed to confirm these results. This work was supported in part by a grant from Varian Medical Systems (Palo Alto, CA)

60 APPLIED LIFE SCIENCES↗

Inverse methods for assessing ship-of-opportunity networks and estimating circulation and winds from tropical expendable bathythermograph data

Inverse methods for estimating the surface ciculation of the equatorial Pacific by combining a linear reduced-gravity shallow-water model with the Tropical Ocean-Global Atmosphere ship-of-opportunity expendable bathythermograph (TOGA SOP XBT) observing program are examined. It is demonstrated that a simple linear model of the upper circulation of the equatorial Pacific can be successfully used as a weak constraint when smoothing the TOGA SOP XBT data. A circulation is sought as the weighted least squares fit to the dynamics and the data. The solution method is an expansion in representer functions, and the generalized inverse problem is thereby reduced from a functional problem to an algebraic problem for the coefficients of the representer. A specific inverse calculation using synthetic forcing and data is presented.

Bennett, Andrew F.↗

Systematic reduction of sign errors in many-body problems: generalization of self-healing diffusion Monte Carlo to excited states

A recently developed Self-Healing Diffusion Monte Carlo Algorithm [PRB {\bf 79}, 195117 ] is extended to the calculation of excited states. The formalism is based on a excited-state fixed-node approximation and the mixed estimator of the excited-state probability density. The fixed-node ground state wave-functions of inequivalent nodal pockets are found simultaneously using a recursive approach. The decay of the wave-function into lower energy states is prevented using two methods: i) The projection of the improved trial-wave function into previously calculated eigenstates is removed. ii) The reference energy for each nodal pocket is adjusted in order to create a kink in the global fixed-node wave-function which, when locally smoothed out, increases the volume of the higher energy pockets at the expense of the lower energy ones until the energies of every pocket become equal. This reference energy method is designed to find nodal structures that are local minima for arbitrary fluctuations of the nodes within a given nodal topology. We demonstrate in a model system that the algorithm converges to many-body eigenstates in bosonic-like and fermionic cases.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Modular Flooring System

The modular flooring system (MFS) was developed to provide a portable, modular, durable carpeting solution for NASA fs Robotics Alliance Project fs (RAP) outreach efforts. It was also designed to improve and replace a modular flooring system that was too heavy for safe use and transportation. The MFS was developed for use as the flooring for various robotics competitions that RAP utilizes to meet its mission goals. One of these competitions, the FIRST Robotics Competition (FRC), currently uses two massive rolls of broadloom carpet for the foundation of the arena in which the robots are contained during the competition. The area of the arena is approximately 30 by 72 ft (approximately 9 by 22 m). This carpet is very cumbersome and requires large-capacity vehicles, and handling equipment and personnel to transport and deploy. The broadloom carpet sustains severe abuse from the robots during a regular three-day competition, and as a result, the carpet is not used again for competition. Similarly, broadloom carpets used for trade shows at convention centers around the world are typically discarded after only one use. This innovation provides a green solution to this wasteful practice. Each of the flooring modules in the previous system weighed 44 lb (.20 kg). The improvements in the overall design of the system reduce the weight of each module by approximately 22 lb (.10 kg) (50 %), and utilize an improved "module-to-module" connection method that is superior to the previous system. The MFS comprises 4-by-4-ft (.1.2-by- 1.2-m) carpet module assemblies that utilize commercially available carpet tiles that are bonded to a lightweight substrate. The substrate surface opposite from the carpeted surface has a module-to-module connecting interface that allows for the modules to be connected, one to the other, as the modules are constructed. This connection is hidden underneath the modules, creating a smooth, co-planar flooring surface. The modules are stacked and strapped onto durable, commercially available drywall carts for storage and/or transportation. This method of storage and transportation makes it very convenient and safe when handling large quantities of modules.

Thate, Robert↗

Combined Dimensional and Topology Optimization of Interior Permanent Magnet Synchronous Machine Rotors Using a Permanent Magnet Function Interpolation Method

This paper presents a magneto-structural combined dimensional and topology optimization technique for interior permanent magnet synchronous machine (IPMSM) rotors. Dimensional changes to the permanent magnet (PM) location or size are accomplished by interpolating or projecting a smoothed Heaviside rectangular function representing the presence of PM material onto the IPMSM rotor design domain mesh. A density based Solid Isotropic with Material Penalization (SIMP) topology optimization approach is then used to vary the presence of electrical steel in mesh elements to form flux barriers around the PM. The proposed method enforces a defined shape for the PM without requiring the mesh in the design domain to be deformed. Three examples are presented to demonstrate the technique: two flat bar IPMSM and one V-shaped IPMSM.

33 ADVANCED PROPULSION SYSTEMS↗

Cold Seeded Epitaxy and Flexomagnetism in GdAuGe Membranes Exfoliated From Graphene/Ge(111)

Remote and van der Waals epitaxy are promising approaches for synthesizing single crystalline membranes for flexible electronics and discovery of new properties via extreme strain; however, a fundamental challenge is that most materials do not wet the graphene surface. We develop a cold seed approach for synthesizing smooth intermetallic films on graphene that can be exfoliated to form few nanometer thick single crystalline membranes. Our seeded GdAuGe films have narrow X-ray rocking curve widths of 9–24 arc seconds, which is 2 orders of magnitude lower than their counterparts grown by typical high temperature methods, and have atomically sharp interfaces observed by transmission electron microscopy. Upon exfoliation and rippling, strain gradients in GdAuGe membranes induce an antiferromagnetic to ferri/ferromagnetic transition. Finally, our smooth, ultrathin membranes provide a clean platform for discovering new flexomagnetic effects in quantum materials.

36 MATERIALS SCIENCE↗

Deep Neural Networks are Adaptive to Function Regularity and Data Distribution in Approximation and Estimation

Deep learning has exhibited remarkable results across diverse areas. To understand its success, substantial research has been directed towards its theoretical foundations. Nev- ertheless, the majority of these studies examine how well deep neural networks can model functions with uniform regularities. In this paper, we explore a different angle: how deep neural networks can adapt to varying degrees of smoothness in functions and nonuni- form data distributions across different locations and scales. More precisely, we focus on a broad class of functions defined by nonlinear tree-based approximation methods. This class encompasses a range of function types, such as functions with uniform regularities and discontinuous functions. We develop nonparametric approximation and estimation theories for this class using deep ReLU networks. Our results show that deep neural networks are adaptive to the nonuniform smoothness of functions and nonuniform data distributions at different locations and scales. We apply our results to several function classes, and derive the corresponding approximation and generalization errors. The validity of our results is demonstrated through numerical experiments.

97 MATHEMATICS AND COMPUTING↗

Advances in Automation of Overset Structured Volume Mesh Generation and Domain Connectivity

Automation of overset structured surface mesh generation has recently been accomplished by the creation of face, edge, and node meshes based on Boundary Representation solids as the geometry input. The current work continues the automation effort in overset volume mesh generation and domain connectivity based on the auto-generated surface meshes. All near-body curvilinear volume meshes are automatically generated using hyperbolic methods. Automation of this step is enabled by appropriate surface grid point distribution, and selection of boundary-splay and smoothing parameters based on concave and convex surface features. The off-body domain is covered by two automatically generated grid systems. The first contains a single Cartesian mesh with a uniform core enclosing all near-body volume meshes and stretched layers that extend to the far field, while the second consists of a set of small stretched Cartesian grids covering pockets of off-body orphan points. With high quality mesh overlap mostly guaranteed by the surface meshing scheme, orphan points that need to be covered by the second Cartesian mesh system are located away from the fine grid spacing region near the wall. Using line-segment and ray-pierce tests against the surface grids, hole-cutting is accomplished on both near and off-body volume grids resulting in appropriate clearances from the wall. The complete mesh generation automation process is demonstrated on five test cases where flow solutions are also computed and compared with solutions obtained using other methods.

TTT↗

Mesoscale Simulation of Proton Transport in Proton Exchange Membranes

Previous efforts to model proton transport through fuel cell membranes have largely focused on disparate length scales: molecular dynamics at the atomistic level and fuel cell stack engineering approaches at the macroscale. A new multiscale approach to bridge these extremes is proposed in this work which combines concepts from coarse-grained (CG) modeling with smoothed particle hydrodynamics (SPH) to capture the qualitative morphology and transport behavior of a proton exchange membrane at the length scale of tens of nanometers. This method allows for connection to atomistic simulations via the inclusion of transport coefficients from molecular dynamics and coarse-grained forces derived for the polymer backbone, side chain, proton, and water interactions. Information pertaining to macroscopic conductivity is obtained by volume averaging based on the flux and chemical potential fields within the membrane. Proton transport is effectively coarse-grained via introduction of a composition variable associated with each interacting site which carries the field information. By combining this technique with local electrostatics and coordinate dependent diffusion constants, the effects of double layer formation within the water pores and the influence of proximity to sulfonate groups on transport is recovered. The combined CG-SPH method is validated and subsequently applied to an equilibrated hydrated Nafion structure with a box length of 40 nm. The resulting conductivities calculated for the material agree very well with trends from experiment and provide insight into the complex interplay of morphology, proton distribution, and diffusion coefficients at a length scale that can be expanded beyond feasible atomistic molecular dynamics simulations to capture the effects of mesoscopic morphology on proton conduction.

Jorn, Ryan↗

Nuclear Metal Fuel Casting Modeling (Annual Report 2012)

In fiscal year 2012, FLOW3D was used to develop an understanding of the metallic nuclear fuel casting processes. There were two major accomplishments for FY12, 1) a rank-order list of the most important material properties to measure, 2) simulation of alternative mold-fill methods. We performed a design of experiments study to quantify several material properties and their impact on effective mold fill, for both circular and annular fuel rod geometries. The most important fuel material property to be measured is the wetting angle of the molten metal on the coated mold material (coated graphite). The second most important property to measure is the surface tension of the molten metal in argon gas at the design mold pressure. The magnitude of these material properties had the largest effect on whether the mold casting processes produced large voids in the fuel rods. The nominal mold fill method used in the above study was the gravity assisted top-fill method of the Idaho National Laboratory Bench Scale Casting facility. Alternative mold fill methods were investigated and using a bottom-fill method shows promise. In the bottom-fill method, the molten metal is injected into the bottom of the mold and the argon cover gas is allowed to escape through vents at the top of the mold. This allows for a smooth, uniform fill of the mold, where as the top-fill geometry has a counter-flow condition of the molten metal and argon cover gas. The mold is initially filled with argon, and using a top fill method, the molten metal must flow downward into the mold and the argon is displaced by the metal and must flow upward out of the mold. This counter-flow condition creates large voids in the cast because bubbles of argon get trapped as the metal flows into the mold. We also investigated various venting schemes for the top fill casting process and this also shows some promise to mitigate the void formation. In fiscal year 2013, we will continue to investigate mold fill approaches. We will use the results of the FY2012 work and begin investigating the segregation of binary alloys for these casting approaches and recommend separate effects testing requirements. If data becomes available from the BSC, we will calibrate the computational models and identify optimum casting parameters. Finally, we will study the feasibility of extending the calibrated model to ternary flow problem by developing appropriate scaled models.

1.02.02.02↗

Correlated Trajectory Uncertainty for Adaptive Sequential Decision Making

One of the great challenges with decision making tasks on real world systems is the fact that data is sparse and acquiring additional data is expensive. In these cases, it is often crucial to make a model of the environment to assist in making decisions. At the same time, limited data means that learned models are erroneous, making it just as important to equip the model with good predictive uncertainties. In the context of learning sequential decision making policies, these uncertainties can prove useful for informing which data to collect for the greatest improvement in policy performance \citep{mehta2021experimental, mehta2022exploration} or informing the policy about unsure regions of state and action space to avoid during test time \citep{yu2020mopo}. Additionally, assuming that realistic samples of the environment can be drawn, an adaptable policy can be trained that attempts to make optimal decisions for any given possible instance of the environment \citep{ghosh2022offline, chen2021offline}. In this work, we examine the so-called ``probabilistic neural network'' (PNN) model that is ubiquitous in model-based reinforcement learning (MBRL) works. We argue that while PNN models may have good marginal uncertainties, they form a distribution of non-smooth transition functions. Not only are these samples unrealistic and may hamper adaptability, but we also assert that this leads to poor uncertainty estimates when predicting multiple step trajectory estimates. To address this issue, we propose a simple sampling method that can be implemented on top of pre-existing models.We evaluate our sampling technique on a number of environments, including a realistic nuclear fusion task, and find that, not only do smooth transition function samples produce more calibrated uncertainties, but they also lead to better downstream performance for an adaptive policy.

Offline Reinforcement Learning↗

An improved multilevel Monte Carlo method for estimating probability distribution functions in stochastic oil reservoir simulations

In this paper, we develop an improved multilevel Monte Carlo (MLMC) method for estimating cumulative distribution functions (CDFs) of a quantity of interest, coming from numerical approximation of large-scale stochastic subsurface simulations. Compared with Monte Carlo (MC) methods, that require a significantly large number of high-fidelity model executions to achieve a prescribed accuracy when computing statistical expectations, MLMC methods were originally proposed to significantly reduce the computational cost with the use of multifidelity approximations. The improved performance of the MLMC methods depends strongly on the decay of the variance of the integrand as the level increases. However, the main challenge in estimating CDFs is that the integrand is a discontinuous indicator function whose variance decays slowly. To address this difficult task, we approximate the integrand using a smoothing function that accelerates the decay of the variance. In addition, we design a novel a posteriori optimization strategy to calibrate the smoothing function, so as to balance the computational gain and the approximation error. The combined proposed techniques are integrated into a very general and practical algorithm that can be applied to a wide range of subsurface problems for high-dimensional uncertainty quantification, such as a fine-grid oil reservoir model considered in this effort. The numerical results reveal that with the use of the calibrated smoothing function, the improved MLMC technique significantly reduces the computational complexity compared to the standard MC approach. Finally, we discuss several factors that affect the performance of the MLMC method and provide guidance for effective and efficient usage in practice.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Extended Physics-Informed Neural Networks (XPINNs): A Generalized Space-Time Domain Decomposition Based Deep Learning Framework for Nonlinear Partial Differential Equations

Here we propose a generalized space-time domain decomposition approach for the physics-informed neural networks (PINNs) to solve nonlinear partial differential equations (PDEs) on arbitrary complex-geometry domains. The proposed framework, named eXtended PINNs ( X P I N N s ), further pushes the boundaries of both PINNs as well as conservative PINNs (cPINNs), which is a recently proposed domain decomposition approach in the PINN framework tailored to conservation laws. Compared to PINN, the XPINN method has large representation and parallelization capacity due to the inherent property of deployment of multiple neural networks in the smaller subdomains. Unlike cPINN, XPINN can be extended to any type of PDEs. Moreover, the domain can be decomposed in any arbitrary way (in space and time), which is not possible in cPINN. Thus, XPINN offers both space and time parallelization, thereby reducing the training cost more effectively. In each subdomain, a separate neural network is employed with optimally selected hyperparameters, e.g., depth/width of the network, number and location of residual points, activation function, optimization method, etc. A deep network can be employed in a subdomain with complex solution, whereas a shallow neural network can be used in a subdomain with relatively simple and smooth solutions. We demonstrate the versatility of XPINN by solving both forward and inverse PDE problems, ranging from one-dimensional to three-dimensional problems, from time-dependent to time-independent problems, and from continuous to discontinuous problems, which clearly shows that the XPINN method is promising in many practical problems. The proposed XPINN method is the generalization of PINN and cPINN methods, both in terms of applicability as well as domain decomposition approach, which efficiently lends itself to parallelized computation. The XPINN code is available on h t t p s : / / g i t h u b . c o m / A m e y a J a g t a p / X P I N N s .

97 MATHEMATICS AND COMPUTING↗

Shock-ramp analysis test problem

Quasi-isentropic (ramp) compression is now a well-established experimental method and so are the analysis techniques to give Lagrangian sound speed, pressure, and density along the sample material's isentrope. A shock followed by ramp compression is a natural extension to investigate, for example, shock melt and refreeze on compression, or isentropes of states off the Hugoniot or principal isentrope. In practice, graded-density impactors produce initial shocks, compression by shaped laser pulses may be unable to produce a smooth pressure increase from zero, and incidental perturbations on the drive pulse may also give rise to shocks, so robust shock-ramp analysis methods will be needed. Appropriate analysis methods are needed for shock-ramp experiments, based on those for quasi-isentropic compression, and these require validation. This paper describes three different analyses of a shock-ramp test problem, including an assessment of their estimated errors. The methods tested were based on hydrodynamic characteristics or integration backward in space. All methods gave the known Lagrangian sound speed to within ~1%, and pressure and volume to within less than 2% and 1%, demonstrating that the analysis methods of isentropic compression experiments can be confidently extended to the analysis of shock and ramp compression.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Pseudospectral collocation methods for the computation of the self-force on a charged particle: Generic orbits around a Schwarzschild black hole

The inspiral of a stellar compact object into a massive black hole, an extreme-mass-ratio inspiral, is one of the main sources of gravitational waves for the future space-based Laser Interferometer Space Antenna. We expect to be able to detect and analyze many cycles of these slowly inspiraling systems, which makes them truly high-precision tools for gravitational-wave astronomy. To that end, the use of very precise theoretical waveform templates in the data analysis is required. To build them, we need to have a deep understanding of the gravitational backreaction mechanism responsible for the inspiral. The self-force approach describes the inspiral as the action of a local force that can be obtained from the regularization of the perturbations created by the stellar compact object on the massive black hole geometry. In this paper we extend a new time-domain technique for the computation of the self-force from the circular case to the case of eccentric orbits around a nonrotating black hole. The main idea behind our scheme is to use a multidomain framework in which the small compact object, described as a particle, is located at the interface between two subdomains. Then, the equations at each subdomain are homogeneous wave-type equations, without distributional sources. In this particle-without-particle formulation, the solution of the equations is smooth enough to provide good convergence properties for the numerical computations. This formulation is implemented by using a pseudospectral collocation method for the spatial discretization, combined with a Runge-Kutta algorithm for the time evolution. We present results from several simulations of eccentric orbits in the case of a scalar charged particle around a Schwarzschild black hole, an excellent test bed model for testing the techniques for self-force computations. In particular, we show the convergence of the method and its ability to resolve the field and its derivatives across the particle location. Finally, we provide numerical values of the self-force for different orbital parameters.

79 ASTRONOMY AND ASTROPHYSICS↗

Effects of Varying the Three-Body Molecular Hydrogen

The transformation of atomic hydrogen to molecular hydrogen through three-body reactions is a crucial stage in the collapse of primordial, metal-free halos, where the first generation of stars (Population III stars) in the Universe are formed. However, in the published literature, the rate coefficient for this reaction is uncertain by nearly an order of magnitude. We report on the results of both adaptive mesh refinement (AMR) and smoothed particle hydrodynamics (SPH) simulations of the collapse of metal-free halos as a function of the value of this rate coefficient. For each simulation method, we have simulated a single halo three times, using three different values of the rate coefficient. We find that while variation between halo realizations may be greater than that caused by the three-body rate coefficient being used, both the accretion physics onto Population III protostars as well as the long-term stability of the disk and any potential fragmentation may depend strongly on this rate coefficient.

08 HYDROGEN↗

Bootstrap AMG for spectral clustering

Graph Laplacian is a popular tool for analyzing graphs, particularly in graph partitioning and clustering. Given a notion of similarity (via an adjacency matrix), graph clustering refers to identifying different groups such that vertices in the same group are more similar compared to vertices across different groups. Data clustering can be reformulated in terms of a graph clustering problem when the given set of data is represented as a graph, also known as similarity graph. In this context, eigenvectors of the graph Laplacian are often used to obtain a new geometric representation of the original data set that generally enhances cluster properties and improves cluster detection. Here, we apply a bootstrap algebraic multigrid (AMG) method that constructs a set of vectors associated with the graph Laplacian. These vectors, referred to as algebraically smooth ones, span a low-dimensional Euclidean space, which we use to represent the data, enabling cluster detection both in synthetic and in realistic well-clustered graphs. We show that, in the case of a good quality bootstrap AMG, the computed smooth vectors employed in the construction of the final AMG operator, which by construction is spectrally equivalent to the originally given graph Laplacian, accurately approximate the space in the lower portion of the spectrum of the preconditioned operator. Thus, our approach can be viewed as a spectral clustering technique associated with the generalized spectral problem (Laplace operator versus the final AMG operator), and hence, it can be seen as an extension of the classical spectral clustering that employs a standard eigenvalue problem.

97 MATHEMATICS AND COMPUTING↗