Maria Rightley - Academia.edu (original) (raw)

Papers by Maria Rightley

Combustion and Flame, May 1, 1995

Abstract Since hydrogen-containing species catalyze CO oxidation, at small H/C and H/O ratios ste... more Abstract Since hydrogen-containing species catalyze CO oxidation, at small H/C and H/O ratios steady-state approximations for reaction intermediaries should be accurate and should produce a good one-step reduced-chemistry approximation for describing the flame structure. Such an approximation is complicated, however, because of the many elementary kinetic steps involved. In the present work simplifications are introduced into a one-step approximation that enable analytical expressions to be obtained for the burning velocities of premixed laminar flames by asymptotic methods. The resulting burning velocities are shown to be in good agreement with experiment and with detailed numerical predictions, and qualitative aspects of the internal flame structure are shown to agree with those deduced from full-chemistry numerical calculations. Estimates are given of the ranges of conditions over which the approximations can reasonably be employed and of flammability limits controlled by chemical kinetics.

Combustion and Flame, Aug 1, 1997

Asymptotic and computational methods are employed to investigate premixed laminar burning velocit... more Asymptotic and computational methods are employed to investigate premixed laminar burning velocities of planar adiabatic flames in ideal gas mixtures containing carbon monoxide, oxygen, an inert, and trace amounts of hydrogen-containing species. The chemical-kinetic steps that control the burning velocity are identified, and the dependence of the burning velocity on composition, pressure, and temperature is determined. A lower limit to the burning velocity that is greater than 5 mm/s, independent of pressure, is established for totally dry stoichiometric mixtures at 300 K. The results contribute to understanding of CO flame structure.

Combustion Science and Technology, 1997

Abstract Computational results are reported for structures of laminar counterflow diffusion flame... more Abstract Computational results are reported for structures of laminar counterflow diffusion flames between carbon monoxide and air, initially at room temperature and pressure from 1 to 100 atm, with total hydrogen-atom mole fractions in the system ranging from zero to about 0.02. All strain rates considered are within a factor of ten of the critical extinction strain rate. This critical strain rate is calculated as a function of pressure and of hydrogen content and is shown to lie below measured values under most conditions. For hydrogen-free flames, activation-energy asymptotics is employed and supports the computational results. It is reasoned that trace hydrogen amounts in air and preferential hydrogen diffusion through nonplanar diffusive-thermal instability contribute to enhanced flame robustness in the experiments, while increasing buoyant convective heat loss with increasing pressure promotes extinction at the higher pressures.

Dispersed droplet or particulate two-phase flows are frequently used in industrial applications, ... more Dispersed droplet or particulate two-phase flows are frequently used in industrial applications, and their accurate prediction by means of computer flow simulations is of great practical importance. To capture the essential dynamics of dispersed two-phase flows, numerical models must often solve for a probability distribution of dispersed phase entity locations, sizes, velocities, and temperatures. Because in general the problem is eight dimensional, Monte Carlo or stochastic computational particle methods are usually used to solve the dispersed phase equations. Each computational particle carries with it a number of properties equal to the number of independent variables of the probability distribution function, and these properties are updated in time by numerically solving Lagrangian equations. While stochastic particle methods have many advantages for serial computer calculations, they are difficult to implement efficiently on parallel platforms. Therefore we are investigating continuum, or non-Lagrangian, formulations of the dispersed phase equations that are of lower dimensionality and that parallelize well. Our initial efforts have concentrated on deriving equations for a four-dimensional model in which the dispersed phase entities have arbitrary physical space and size distributions and on incorporating into this model the effects of turbulent dispersion of particles subject to Stokes drag.

Aps Division of Fluid Dynamics Meeting Abstracts, Nov 1, 2001

Equations of state (EOS) for gaseous products of detonation of high explosives (HE) have generall... more Equations of state (EOS) for gaseous products of detonation of high explosives (HE) have generally been empirically based. That is, an EOS is calibrated so that simulations of a particular experiment agree well with the experimental data. For high explosives, the test used to calibrate these EOS values has been the cylinder test. A cylinder test begins with a metal (usually copper) cylinder filled with high explosive; the explosive is then detonated at one end. The detonation velocity and cylinder wall expansion profiles and/or cylinder wall velocity profiles are measured. This work will look at the sensitivity of the match (of experiment and calculation) to perturbations of the EOS parameters, and draw conclusions about the ability to validate equations of state of high explosives using such methods from the results.

AIP Conference Proceedings, 1998

The technique called Differential Sensitivity has been applied to the system of Eulerian continuu... more The technique called Differential Sensitivity has been applied to the system of Eulerian continuum mechanics equations solved by a hydrocode. Differential Sensitivity uses forward and adjoint techniques to obtain output response sensitivity to input parameters. Previous papers have described application of the technique to two-dimensional, multi-component problems. Inaccuracies in the adjoint solutions have prompted us to examine our numerical techniques in more detail. Here we examine one-dimensional, one material shock problems. Solution accuracy is assessed by comparison to sensitivities obtained by automatic differentiation and a code-based adjoint differentiation technique.

Many problems in physics and modern computing are inverse problems -- problems where the desired ... more Many problems in physics and modern computing are inverse problems -- problems where the desired output is known, and the task is to find the set of input parameters that will best reproduce that output in a hydrodynamics code (hydrocode). Optimization methods tackle this type of problem, and a central task in applying optimization methods is to be able to determine the gradient of the output with respect to the input parameters that are being adjusted. Presented here is the authors' progress (through the use of adjoint differentiation) in obtaining those gradients in the case of some relatively simple hydrocodes. 1 Introduction When a program simulates a physical system, it does so through the use of a set of equations and mathematical relationships known as a physical model. This model will generally contain a number of parameters that influence the system. Depending on the problem of interest, there may be several situations that arise. One such situation is the inverse prob...

When calculating the sensitivities of hydrocode results, one generally chooses either equation-ba... more When calculating the sensitivities of hydrocode results, one generally chooses either equation-based or code-based methods. The automatic differentiation (AD) programs ADIFOR, GRESS and TAMC utilize a code-based approach, determining adjoint code on a line-by-line basis, producing an enhanced source that is compiled and run to give the derivatives of interest. The Adjoint Differentiation In Code Technique (ADICT) also utilizes a code-based approach. The adjoint code is written to reverse the path of the forward calculation, accumulating sensitivities as it does the reverse calculation, which decreases the storage requirements compared to the AD methods. This process is simplified by making extensive use of subroutines for both the forward and the adjoint code. Differential Sensitivity (DS) uses equation-based techniques to determine sensitivity derivatives; either the original pdes or the hydrocode's actual difference equations are differentiated. Those differentiated equations are then either solved numerically (forward mode) or adjointed (transposed, when using the difference equations) and then solved to obtain the derivatives. We present a comparison of these methods (DS,AD,ADICT) applied to a relatively simple one-dimensional hydrocode for analyzing shock problems. Progress with application to more complex codes will also be discussed.

AdjointDierentiationofHydro dynamicCo desMariaLJRightleyRudolphHenningerandKennethMHansonXNHMSFXH... more AdjointDierentiationofHydro dynamicCo desMariaLJRightleyRudolphHenningerandKennethMHansonXNHMSFXHMMSDDXMSP LosAlamosNationalLab oratoryNMAprilAbstractManyproblemsinphysicsandmo derncomputingareversewherethedesiredoutputisknownandtasktondsetofinputparametersthatwillb estrepro duceoutputinahydro dynamicsco deydro co deOptimizationmetho dstacklethistyp eofproblemandacentraltaskinapplyingoptimizationmetho dsistobeabledeterminethegradientofoutputwithresp ectinputparametersthatareb eingadjustedPresentedhereistheauthorsprogressthroughtheuseofadjointdi erentiationinobtainingthosegradientscasesomerelativelysimplehydro co desIntro ductionWhenaprogramsimulatesphysicalsystemitdo essothroughtheuseofsetequationsandmathematicalrelationshipsknownasaphysicalmo delThiswillgenerallycontainanumberofparametersthatinuencethesystemDep endingonproblemofinteresttheremaybeseveralsituationsthatariseOnesuchsituationistheinverseproblemwheretheoutputoratleastdesiredisknownanditinputparametersthatneedtobedetermin...

The setup for a cylinder test consists of a copper cylinder filled with high explosive; the explo... more The setup for a cylinder test consists of a copper cylinder filled with high explosive; the explosive is then detonated at one end. The detonation velocity and cylinder wall expansion profiles and/or cylinder wall velocity profiles are measured (if only the cylinder wall expansion profiles are measured, it is straightforward to calculate the wall velocities by differentiating the expansion profile values with respect to time). There have been theories occasionally put forward that the ratio of the wall velocity value at 19 mms expansion to that at 6 mms expansion will discriminate explosives with a fast reaction from those with a slow reaction. It seems more likely, however, that the velocity ratio really shows, for the most part, the difference in product sound speeds, and those speeds determine how rapidly energy is transferred from the expanding gases to the metal cylinder. By looking at the velocity ratio results for numerous cylinder tests that have been compiled locally in a database of experimental results, these conflicting ideas will be put to the test.

AIP Conference Proceedings

Aps Division of Fluid Dynamics Meeting Abstracts, Nov 1, 1997

There have been many equations of state (EOS) proposed for gaseous products of detonation, from s... more There have been many equations of state (EOS) proposed for gaseous products of detonation, from simple theoretically-based EOS to empirically-based EOS with many adjustable parameters. How well these EOS approximate the real behavior depends on the material which is detonated. Gases, for example, are much easier to represent simply than are condensed or solid materials. If we concentrate solely on the particular class of condensed materials known as high explosives (HE) and their products of detonation, the most common formulation is the Jones-Wilkins-Lee (JWL) EOS. It has six adjustable parameters, and is as popular as it is in some part due to the fact that it can represent well the experiments which are primary source for HE EOS data, the explosive cylinder test. This test uses a cylinder of copper filled with an HE, which is then initiated, and the expansion is recorded. The obtained expansion profile is then used to calibrate the EOS, generally the JWL form. The work presented here will describe the rereading of some old film data, as well as recalibration of that data for the JWL EOS, and progress with modified EOS forms.

Many problems in physics and modern computing are inverse problems -- problems wherethe desired o... more Many problems in physics and modern computing are inverse problems -- problems wherethe desired output is known, and the task is to find the set of input parameters that will bestreproduce that output in a hydrodynamics code (hydrocode). Optimization methods tackle thistype of problem, and a central task in applying optimization methods is to be able to determinethe gradient of

Journal of the American Statistical Association, 2008

This work focuses on combining observations from field experiments with detailed computer simulat... more This work focuses on combining observations from field experiments with detailed computer simulations of a physical process to carry out statistical inference. Of particular interest here is determining uncertainty in resulting predictions. This typically involves calibration of parameters in the computer simulator as well as accounting for inadequate physics in the simulator. The problem is complicated by the fact that simulation code is sufficiently demanding that only a limited number of simulations can be carried out. We consider applications in characterizing material properties for which the field data and the simulator output are highly multivariate. For example, the experimental data and simulation output may be an image or may describe the shape of a physical object. We make use of the basic framework of Kennedy and O'Hagan (2001). However, the size and multivariate nature of the data lead to computational challenges in implementing the framework. To overcome these challenges, we make use of basis representations (e.g. principal components) to reduce the dimensionality of the problem and speed up the computations required for exploring the posterior distribution. This methodology is applied to applications, both ongoing and historical, at Los Alamos National Laboratory.