Search NASA⌕ Search

SEARCH · Search NASA

Results for “Error estimation”

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 235 records · Page 13

Active learning emulators for nuclear two-body scattering in momentum space

In this work we extend the active learning emulators for two-body scattering in coordinate space with error estimation, recently developed by Maldonado et al. [Phys. Rev. C 112, 024002], to coupled-channel scattering in momentum space. Our full-order model (FOM) solver is based on the Lippmann-Schwinger integral equation for the scattering t-matrix as opposed to the radial Schrödinger equation. We use (Petrov-)Galerkin projections and high-fidelity calculations at a few snapshots across the parameter space of the interaction to construct efficient reduced-order models (ROMs), trained by a greedy algorithm for locally optimal snapshot selection. Both the FOM solver and the corresponding ROMs are implemented efficiently in Python using Google's JAX library. We present results for emulating scattering phase shifts in coupled and uncoupled channels and cross sections, and assess the accuracy of the developed ROMs and their computational speedup factors. We also develop emulator error estimation for both the t-matrix and the total cross section. The software framework for reproducing and extending our results is publicly available. Together with our recent advances in developing active-learning emulators for three-body scattering, these emulator frameworks set the stage for full Bayesian calibrations of chiral nuclear interactions and optical models against scattering data with quantified emulator errors.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Greedy emulators for nuclear two-body scattering

Applications of reduced basis method emulators are increasing in low-energy nuclear physics because they enable fast and accurate sampling of high-fidelity calculations, enabling robust uncertainty quantification. Here, in this paper, we develop, implement, and test two model-driven emulators based on the (Petrov-)Galerkin projection using the prototypical test case of two-body scattering with the Minnesota potential and a more realistic local chiral potential. The high-fidelity scattering equations are solved with the matrix Numerov method, a reformulation of the popular Numerov recurrence relation for solving special second-order differential equations as a linear system of coupled equations. A novel error estimator based on reduced-space residuals is applied to an active learning approach (a greedy algorithm) to choosing training samples (“snapshots”) for the emulator and contrasted with a proper orthogonal decomposition (POD) approach. Both approaches allow for computationally efficient offline-online decompositions, but the greedy approach requires many fewer snapshot calculations. These developments set the groundwork for emulating scattering observables based on chiral nucleon-nucleon and three-nucleon interactions and optical models, where computational speed-ups are necessary for Bayesian uncertainty quantification. Our emulators and error estimators are widely applicable to linear systems.

Bayesian methods↗

An efficient algorithm for simulating ensembles of parameterized flow problems

Abstract Many applications of computational fluid dynamics require multiple simulations of a flow under different input conditions. In this paper, a numerical algorithm is developed to efficiently determine a set of such simulations in which the individually independent members of the set are subject to different viscosity coefficients, initial conditions and/or body forces. The proposed scheme, when applied to the flow ensemble, needs to solve a single linear system with multiple right-hand sides, and thus is computationally more efficient than solving for all the simulations separately. We show that the scheme is nonlinearly and long-term stable under certain conditions on the time-step size and a parameter deviation ratio. A rigorous numerical error estimate shows the scheme is of first-order accuracy in time and optimally accurate in space. Several numerical experiments are presented to illustrate the theoretical results.

Mathematics↗

Ab initio computation of neutron-rich oxygen isotopes

We compute the binding energy of neutron-rich oxygen isotopes and employ the coupled-cluster method and chiral nucleon-nucleon interactions at next-to-next-to-next-to-leading order with two different cutoffs. We obtain rather well-converged results in model spaces consisting of up to 21 oscillator shells. For interactions with a momentum cutoff of 500 MeV, we find that {sup 28}O is stable with respect to {sup 24}O, while calculations with a momentum cutoff of 600 MeV result in a slightly unbound {sup 28}O. The theoretical error estimates due to the omission of the three-nucleon forces and the truncation of excitations beyond three-particle-three-hole clusters indicate that the stability of {sup 28}O cannot be ruled out from ab initio calculations, and that three-nucleon forces and continuum effects play the dominant role in deciding this question.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Automated Symbolic Upscaling: 2. Model Generation for Extended Applicability Regimes

Abstract In this second part of the two paper series, we detail an algorithmic procedure for systematically implementing the generalized closure form strategy presented in Part 1. This strategy extends the applicability of homogenized models with respect to classical homogenization theory, as demonstrated in Part 1 where upscaled models are rigorously derived in moderately reactive physical regimes. After encoding the algorithm into Symbolica, an automated upscaling framework, we upscale two reactive mass transport problems and numerically validate the resulting nonlinear homogenized models by showing the absolute error estimates predicted by homogenization theory are satisfied. In both problems, nontrivial closure forms and closure problems are automatically formulated using the encoded strategy with no human interaction, nor prior knowledge regarding the closure required for the systems. We hope these demonstrations spark further interest in automated analytical frameworks for multiscale modeling, as such capabilities are invaluable for generating rigorous multiscale models of complex phenomena in porous media.

Pietrzyk, Kyle↗

Automated Symbolic Upscaling: 1. Model Generation for Extended Applicability Regimes

Abstract In porous media theory, upscaling techniques are fundamental to deriving rigorous Darcy‐scale models for flow and reactive transport in subsurface systems. Due to limitations in classical techniques, a number of ad hoc approaches have been proposed to address physical regimes in which reactive time scales are similar to, or faster than, diffusive time scales. In Part 1 of this two part series, we present a strategy for expanding the applicability of classical homogenization theory by generalizing the assumed closure form. We detail the implementation of this strategy on two reactive mass transport problems with moderately reactive physics. The strategy produces nontrivial homogenized models with emergent terms and effective parameters that couple reactive, diffusive, and advective transport. The differences in equation forms between the macroscopic and pore‐scale descriptions advise caution to further studies where the forms of macroscopic equations are assumed, as opposed to rigorously derived. Numerical validation is provided for each problem to show that the error estimates of homogenization theory are satisfied, and to justify the implemented strategy. In Part 2, the presented strategy is automated using symbolic computing to expedite its implementation.

Pietrzyk, Kyle↗

Advanced numerical methods in mesh generation and mesh adaptation

Numerical solution of partial differential equations requires appropriate meshes, efficient solvers and robust and reliable error estimates. Generation of high-quality meshes for complex engineering models is a non-trivial task. This task is made more difficult when the mesh has to be adapted to a problem solution. This article is focused on a synergistic approach to the mesh generation and mesh adaptation, where best properties of various mesh generation methods are combined to build efficiently simplicial meshes. First, the advancing front technique (AFT) is combined with the incremental Delaunay triangulation (DT) to build an initial mesh. Second, the metric-based mesh adaptation (MBA) method is employed to improve quality of the generated mesh and/or to adapt it to a problem solution. We demonstrate with numerical experiments that combination of all three methods is required for robust meshing of complex engineering models. The key to successful mesh generation is the high-quality of the triangles in the initial front. We use a black-box technique to improve surface meshes exported from an unattainable CAD system. The initial surface mesh is refined into a shape-regular triangulation which approximates the boundary with the same accuracy as the CAD mesh. The DT method adds robustness to the AFT. The resulting mesh is topologically correct but may contain a few slivers. The MBA uses seven local operations to modify the mesh topology. It improves significantly the mesh quality. The MBA method is also used to adapt the mesh to a problem solution to minimize computational resources required for solving the problem. The MBA has a solid theoretical background. In the first two experiments, we consider the convection-diffusion and elasticity problems. We demonstrate the optimal reduction rate of the discretization error on a sequence of adaptive strongly anisotropic meshes. The key element of the MBA method is construction of a tensor metric from hierarchical edge-based error estimates. We conclude that the quasi-optimal mesh must be quasi-uniform in this metric. All numerical experiments are based on the publicly available Ani3D package, the collection of advanced numerical instruments.

99 GENERAL AND MISCELLANEOUS↗

A Leray regularized ensemble-proper orthogonal decomposition method for parameterized convection-dominated flows

Abstract Partial differential equations (PDEs) are often dependent on input quantities that are uncertain. To quantify this uncertainty PDEs must be solved over a large ensemble of parameters. Even for a single realization this can be a computationally intensive process. In the case of flows governed by the Navier–Stokes equations, an efficient method has been devised for computing an ensemble of solutions. To further reduce the computational cost of this method, an ensemble-proper orthogonal decomposition (POD) method was recently proposed. The main contribution of this work is the introduction of POD spatial filtering for ensemble-POD methods. The POD spatial filter makes possible the construction of the Leray ensemble-POD model, which is a regularized-reduced order model for the numerical simulation of convection-dominated flows of moderate Reynolds number. The Leray ensemble-POD model employs the POD spatial filter to smooth (regularize) the convection term in the Navier–Stokes equations, and diminishes the numerical inaccuracies produced by the ensemble-POD method in the numerical simulation of convection-dominated flows. Specifically, for the numerical simulation of a convection-dominated two-dimensional flow between two offset cylinders, we show that the Leray ensemble-POD method better reflects the dynamics of the benchmark results than the ensemble-POD scheme. The second contribution of this work is a new numerical discretization of the variable viscosity ensemble algorithm in which the average viscosity is replaced with the maximum viscosity. It is shown that this new numerical discretization is significantly more stable than those in current use. Furthermore, error estimates for the novel Leray ensemble-POD algorithm with this new numerical discretization are also proven.

Mathematics↗

Modeling and control of fuel distribution in a dual-fuel internal combustion engine leveraging late intake valve closings

In internal combustion engines, cycle-to-cycle and cylinder-to-cylinder variations of the combustion process have been shown to negatively impact the fuel efficiency of the engine and lead to higher exhaust emissions. The combustion variations are generally tied to differences in the composition and condition of the trapped mass throughout each cycle and across individual cylinders. Thus, advanced engines featuring exhaust gas recirculation, flexible valve actuation systems, advanced fueling strategies, and turbocharging systems are prone to exhibit higher variations in the combustion process. In this study, the cylinder-to-cylinder variations of the combustion process in a dual-fuel internal combustion engine leveraging late intake valve closing are investigated and a model to predict and address one of the root causes for these variations across cylinders is developed. The study is conducted on an inline six-cylinder heavy-duty dual-fuel engine equipped with exhaust gas recirculation, a variable geometry turbocharger, and a fully flexible variable intake valve actuation system. The engine is operated with late intake valve closure timings in a dual-fuel combustion mode in which a high reactivity fuel is directly injected into the cylinders and a low reactivity fuel is port injected into the cylinders. The cylinder-to-cylinder variations observed in the study have been associated with the maldistribution of the port-injected fuel, which is exacerbated at late intake valve timings. The resulting difference in indicated mean effective pressure between the cylinders ranges from 9% at an intake valve closing of 570° after top dead center to 38% at an intake valve closing of 620° after top dead center and indicates an increasingly uneven fuel distribution. The study leverages both experimental and simulation studies to investigate the distribution of the port-injected fuel and its impact on cylinder-to-cylinder variation. The effects of intake valve closing as well as the impact of intake runner length on fuel distribution were quantitatively analyzed, and a model was developed that can be used to accurately predict the fuel distribution of the port-injected fuel at different operating conditions with an average estimation error of 1.5% in cylinder-specific fuel flow. A model-based control strategy is implemented to adjust the fueling at each port and shown to significantly reduce the cylinder-to-cylinder variations in fuel distribution.

Kassa, Mateos↗

Estimation of Errors Associated With Use of Linear-Quadratic Formalism for Evaluation of Biologic Equivalence Between Single and Hypofractionated Radiation Doses: An In Vitro Study

Purpose: To investigate the reliability of the linear-quadratic (LQ) formalism and the magnitude of errors associated with its use in assessing biologic equivalence between single, high radiation doses and hypofractionated radiation doses. Methods and Materials: V79 and EMT6 single cells received single doses of 2-12 Gy or two or three fractions of 4 or 5 Gy, each at 4-h intervals. Single and fractionated doses to actually reduce the cell survival to the same level were determined by a colony assay. The {alpha}/{beta} ratio was obtained from the cell survival curves. Using the {alpha}/{beta} ratio and the LQ formalism, equivalent single doses for the hypofractionated doses were calculated. They were then compared with the actually determined equivalent single doses for the hypofractionated doses. The V79 spheroids received single doses of 5-26 Gy or two to five fractions of 5-12 Gy at 2 or 4-h interval, and then were assayed for cell survival. Next, equivalent single doses for the hypofractionated doses were determined, as were done for the single cells. Results: The {alpha}/{beta} ratio was 5.1 Gy for the V79 single cells and 0.36 Gy for EMT6. In V79, the equivalent single doses for the hypofractionated doses calculated using the LQ formalism were 12-19% lower than the actually measured biologically equivalent single doses. In the EMT6 cells, this trend was also seen, but the differences were not significant. In the V79 spheroids, the calculated doses were 18-30% lower than the measured doses. Conclusion: Conversion of hypofractionated radiation doses to single doses using the LQ formalism could underestimate the effect of hypofractionated radiation by {<=}30%.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Computational and analytical comparison of flux discretizations for the semiconductor device equations beyond Boltzmann statistics

We compare three thermodynamically consistent numerical fluxes known in the literature, appearing in a Voronoï finite volume discretization of the van Roosbroeck system with general charge carrier statistics. Our discussion includes an extension of the Scharfetter–Gummel scheme to non-Boltzmann (e.g. Fermi–Dirac) statistics. It is based on the analytical solution of a two-point boundary value problem obtained by projecting the continuous differential equation onto the interval between neighboring collocation points. Hence, it serves as a reference flux. The exact solution of the boundary value problem can be approximated by computationally cheaper fluxes which modify certain physical quantities. One alternative scheme averages the nonlinear diffusion (caused by the non-Boltzmann nature of the problem), another one modifies the effective density of states. To study the differences between these three schemes, we analyze the Taylor expansions, derive an error estimate, visualize the flux error and show how the schemes perform for a carefully designed p-i-n benchmark simulation. We present strong evidence that the flux discretization based on averaging the nonlinear diffusion has an edge over the scheme based on modifying the effective density of states.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

RECONCILING THE OBSERVED STAR-FORMING SEQUENCE WITH THE OBSERVED STELLAR MASS FUNCTION

We examine the connection between the observed star-forming sequence (SFR ∝ M {sup α}) and the observed evolution of the stellar mass function in the range 0.2 < z < 2.5. We find that the star-forming sequence cannot have a slope α ≲ 0.9 at all masses and redshifts because this would result in a much higher number density at 10 < log (M/M {sub ☉}) < 11 by z = 1 than is observed. We show that a transition in the slope of the star-forming sequence, such that α = 1 at log (M/M {sub ☉}) < 10.5 and α = 0.7-0.13z (Whitaker et al.) at log (M/M {sub ☉}) > 10.5, greatly improves agreement with the evolution of the stellar mass function. We then derive a star-forming sequence that reproduces the evolution of the mass function by design. This star-forming sequence is also well described by a broken power law, with a shallow slope at high masses and a steep slope at low masses. At z = 2, it is offset by ∼0.3 dex from the observed star-forming sequence, consistent with the mild disagreement between the cosmic star formation rate (SFR) and recent observations of the growth of the stellar mass density. It is unclear whether this problem stems from errors in stellar mass estimates, errors in SFRs, or other effects. We show that a mass-dependent slope is also seen in other self-consistent models of galaxy evolution, including semianalytical, hydrodynamical, and abundance-matching models. As part of the analysis, we demonstrate that neither mergers nor hidden low-mass quiescent galaxies are likely to reconcile the evolution of the mass function and the star-forming sequence. These results are supported by observations from Whitaker et al.

79 ASTRONOMY AND ASTROPHYSICS↗

Errors in estimating Ω{sub Λ} due to the fluid approximation

The matter content of the Universe is strongly inhomogeneous on small scales. Motivated by this fact, we consider a model of the Universe that has regularly spaced discrete masses, rather than a continuous fluid. The optical properties of such space-times can differ considerably from the continuous fluid case, even if the 'average' dynamics are the same. We show that these differences have consequences for cosmological parameter estimation, and that fitting to recent supernovae observations gives a correction to the inferred value of Ω{sub Λ} of ∼ 10%.

79 ASTRONOMY AND ASTROPHYSICS↗

Optimal sensing on an asymmetric exceptional surface

We study the connection between exceptional points (EPs) and optimal parameter estimation, in a simple system consisting of two counterpropagating traveling wave modes in a microring resonator. The unknown parameter to be estimated is the strength of a perturbing cross-coupling between the two modes. Partially reflecting the output of one mode into the other creates a non-Hermitian Hamiltonian that exhibits a family of EPs, creating an exceptional surface (ES). We use a fully quantum treatment of field inputs and noise sources to obtain a quantitative bound on the estimation error by calculating the quantum Fisher information (QFI) in the output fields, whose inverse gives the Cramér-Rao lower bound on the mean-squared error of any unbiased estimator. We determine the bounds for two input states, namely, a semiclassical coherent state and a highly nonclassical NOON state. We find that the QFI is enhanced in the presence of an EP for both of these input states and that both states can saturate the Cramér-Rao bound. We then identify idealized yet experimentally feasible measurements that achieve the minimum bound for these two input states. We also investigate how the QFI changes for parameter values that do not lie on the ES, finding that these can have a larger QFI, suggesting alternative routes to optimize the parameter estimation for this problem.

Exceptional points↗

Calibr8 v.1.0

Calibr8 provides an application to rapidly prototype and perform material model calibration for complex plasticity models using advanced adjoint or forward sensitivity analyses for execution on massively parallel machines. These techniques can be orders of magnitude faster than traditional finite difference approaches for material model calibration. The underlying technology used in Calibr8 is automatic differentiation, which allows for the rapid implementation and testing of new plasticity models within its framework. Additionally, Calibr8 can perform adjoint-based error estimation to approximate discretization errors for user-implemented plasticity models. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525. SAND2021-10630 O

Granzow, BrianN.↗