Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Propagation of Variance”

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 199 records · Page 11

Scintillation in an earth-to-space propagation path.

Results of an experiment to measure the scintillation at a satellite from a ground-based laser transmitter. The GEOS-II satellite in a nearly circular orbit of 1250 km was illuminated by a ground-based laser tracking station using a 4880-A CW argon laser. A detector aboard the satellite measured the incident light and telemetered the data to recording equipment on the ground. Log-amplitude variance, probability distributions, and scintillation frequency distributions are derived from the data. The probability distribution is shown to be log-normal. Log-amplitude variance and normalized power spectral density are shown to be within the limits measured for stellar scintillation.-

Minott, P. O.↗

Uncertainty Quantification Framework for Predicting Material Response with Large Number of Parameters: Application to Creep Prediction in Ferritic-Martensitic Steels Using Combined Crystal Plasticity and Grain Boundary Models

This paper presents an uncertainty quantification (UQ) framework for the physics-based model prediction of material response with a large number of parameters. The application problem presented in this work is that of predicting creep in Grade 91 steel at 600°C. The material response is defined with a physically based microstructural model with constitutive equations emulating several observed phenomena in Grade 91 and embodied into an explicit geometry mesoscale finite element model for prior austenite grains and grain boundaries. Creep within the grains and in grain boundaries are represented by crystal plasticity for dislocation motion and a physics-based model for cavity growth and nucleation, respectively. The creep behavior of this material is influenced by several parameters, some of which have a wide range of variation based on experimental data. UQ combined with microstructural modeling can discover the core microstructural causes of experimental variability, leading to improved materials with lower variability in critical long-term material properties. In this study, we investigate the model's uncertainty to identify material properties that may be modified during production to increase creep life and analyze different components of the crystal plasticity model for improvements. For this purpose, a quantity of interest is defined as time to minimum creep rate, which correlates well to the creep failure of the material. A deep neural network model was trained and validated to be used as a surrogate for the finite element model. Then, a variance-based sensitivity analysis is performed on the surrogate model to find the Sobol indices of the input parameters in respect to the output quantity of interest. The Sobol indices are used to reduce the dimensionality of the model. Generalized polynomial chaos expansion is used on the reduced basis models to propagate the uncertainty from the input parameters to the quantity of interest using the deep neural network surrogate model. These results are benchmarked against uncertainty propagation using Monte Carlo simulations. In conclusion, the UQ performed through the reduced basis model captures almost all the uncertainty in the model with significantly fewer simulations, making it possible to perform the UQ directly via simulations with the finite element model rather than surrogate machine-learned models.

36 MATERIALS SCIENCE↗

The acceleration of electrons at collisionless shocks moving through a turbulent magnetic field

We perform a numerical-simulation study of the acceleration of electrons at shocks that propagate through a prespecified, kinematically defined turbulent magnetic field. The turbulence consists of broadband magnetic fluctuations that are embedded in the plasma and cover a range of wavelengths, the smallest of which is larger than the gyroradii of electrons that are initially injected into the system. We find that when the variance of the turbulent component of the upstream magnetic field is sufficiently large—σ{sup 2} ∼ 10 B{sub 0}{sup 2}, where B{sub 0} is the strength of the background magnetic field—electrons can be efficiently accelerated at a collisionless shock regardless of the orientation of the mean upstream magnetic field relative to the shock-normal direction. Since the local angle between the incident magnetic-field vector and the shock-normal vector can be quite large, electrons can be accelerated through shock-drift acceleration at the shock front. In the upstream region, electrons are mirrored back to the shock front leading to multiple shock encounters. Eventually the accelerated electrons are energetic enough that their gyroradii are of the same order as the wavelength of waves that are included in our description of the turbulent magnetic field. Our results are consistent with recent in situ observations at Saturn’s bow shock. This study may help us to understand the acceleration of electrons at shocks in space and astrophysical systems.

79 ASTRONOMY AND ASTROPHYSICS↗

A 3-D Navier-Stokes CFD study of turbojet/ramjet nozzle plume interactions at Mach 3.0 and comparison with data

Advanced airbreathing propulsion systems used in Mach 4-6 mission scenarios, usually consist of a single integrated turboramjet or as in this study, a turbojet housed in an upper bay with a separate ramjet housed in a lower bay. As the engines transition from turbojet to ramjet, there is an operational envelope where both engines operate simultaneously. One nozzle concept under consideration has a common nozzle, where the plumes from the turbojet and ramjet interact with one another as they expand to ambient conditions. In this paper, the two plumes interact at the end of a common 2-D cowl, when they both reach an approximate Mach 3.0 condition and then jointly expand to Mach 3.6 at the common nozzle exit plane. At this condition, the turbojet engine operated at a higher NPR than the ramjet, where the turbojet overpowers the ramjet plume, deflecting it approximately 12 degrees downward and in turn the turbojet plume is deflected 6 degrees upward. In the process, shocks were formed at the deflections and a shear layer formed at the confluence of the two jets. This particular case was experimentally tested and the data used to compare with the PARC3D code with k-kl two equation turbulence model. The 2-D and 3-D centerline CFD solutions are in good agreement, but as the CFD solutions approach the outer sidewall, a slight variance occurs. The outer wall boundary layers are thin and do not present much of an interaction, however, where the confluence interaction shocks interact with the thin boundary layer on the outer wall, strong vortices run down each shock causing substantial disturbances in the boundary layer. These disturbances amplify somewhat as they propagate downstream axially from the confluence point. The nozzle coefficient (CFG) is reduced 1/2 percent as a result of this sidewall interaction, from 0.9850 to 0.9807. This three-dimensional reduction is in better agreement with the experimental value of 0.9790.

Chang, Ing↗

Gamma-ray Burst Energetics an the Gamma-ray Burst Hubble Diagram: Promises and Limitations

We present a complete sample of 29 gamma-ray bursts (GRBs) for which it has been possible to determine temporal breaks (or limits) from their afterglow light curves. We interpret these breaks within the framework of the uniform conical jet model, incorporating realistic estimates of the ambient density and propagating error estimates on the measured quantities. In agreement with our previous analysis of a smaller sample, the derived jet opening angles of those 16 bursts with redshifts result in a narrow clustering of geometrically corrected gamma-ray energies about 1.33 x 10(exp 51) ergs; the burst-to-burst variance about this value is 0.35 dex, a factor of 2.2. Despite this rather small scatter, we demonstrate in a series of GRB Hubble diagrams that the current sample cannot place meaningful constraints upon the fundamental parameters of the universe. Indeed, for GRBs to ever be useful in cosmographic measurements, we argue the necessity of two directions. First, GRB Hubble diagrams should be based upon fundamental physical quantities such as energy, rather than empirically derived and physically ill-understood distance indicators (such as those based upon prompt burst time-profiles and spectra). Second, a more homogeneous set should be constructed by culling subclasses from the larger sample. These subclasses, although now first recognizable by deviant energies, ultimately must be identifiable by properties other than those directly related to energy. We identify a new subclass of GRBs (" f-GRBs ") that appear both underluminous by factors of at least 10 and exhibit a rapid fading (f(sub nu is proportional to t(sup -2) at early times (t < or = 0.5 day). About 10%-20% of observed long-duration bursts appear to be f-GRBs.

Bloom, J. S.↗

Uncertainty Quantification of Inducer Natural Frequency using Conditional Assessment of Modeling and Modal Testing of Simpler Systems

The low pressure fuel pump inducer of the new Space Launch System RS25 core stage engine operates in a highly complex environment that substantially affects its modal characteristics. Some of the more important effects are fluid-added mass resulting from operation within a light liquid (Hydrogen), and the magnification of this effect due to tight tip clearance. Since higher order cavitation has been identified as a significant harmonic driver, knowledge of the natural frequency of potentially excitable modes is critical for safe operation, but this frequency cannot be measured during the severe operational environment. A comprehensive testing and analysis program has therefore been performed over the last four years to identify the nominal value and uncertainty of the frequency by modeling and testing four simpler structures which share some of the characteristics of the operational inducer. This testing was used to assess and adjust modeling techniques and excellent correlation was achieved. Identification of the uncertainty in the inducer frequency itself was still problematic, however. This difficulty led to an investigation of Bayesian uncertainty quantification techniques, and to the application of the relatively simple technique of Multi-Variate Normal conditional distributions to calculate the inducer natural frequency uncertainty. Assumptions on prior distributions of uncertainty of the fluid-added mass and tip clearance effect are initially applied to models of each of the simple structures and the inducer itself, and these uncertainties are propagated to generate natural frequencies using design of experiments. Simple response surfaces are then created from this data in order to calculate a Covariance Matrix relating all of these natural frequencies. Finally, the results from modal test of the simple structures are considered to be observations and used to calculate the conditional variance of the desired inducer frequencies. As this method is less rigorous than more complicated Bayesian methods reported in the literature, a conservative factor is applied to the result, but the resulting uncertainty is still significantly less than originally estimated and will greatly assist certification of the inducer for use in the engine.

Uncertainty Quantification↗

Uncertainty Quantification of Inducer Eigenvalues using Conditional Assessment of Models and Modal Test of Simpler Systems

The low pressure fuel pump inducer of the new Space Launch System RS25 core stage engine operates in a highly complex environment that substantially affects its modal characteristics. Some of the more important effects are fluid-added mass resulting from operation within a light liquid (Hydrogen), and the magnification of this effect due to tight tip clearance. Since higher order cavitation has been identified as a significant harmonic driver, knowledge of the natural frequency of potentially excitable modes is critical for safe operation, but this frequency cannot be measured during the severe operational environment. A comprehensive testing and analysis program has therefore been performed over the last four years to identify the nominal value and uncertainty of the frequency by modeling and testing four simpler structures which share some of the characteristics of the operational inducer. This testing was used to assess and adjust modeling techniques and excellent correlation was achieved. Identification of the uncertainty in the inducer frequency itself was still problematic, however. This difficulty led to an investigation of Bayesian uncertainty quantification techniques, and to the application of the relatively simple technique of Multi-Variate Normal conditional distributions to calculate the inducer natural frequency uncertainty. Assumptions on prior distributions of uncertainty of the fluid-added mass and tip clearance effect are initially applied to models of each of the simple structures and the inducer itself, and these uncertainties are propagated to generate natural frequencies using design of experiments. Simple response surfaces are then created from this data in order to calculate a Covariance Matrix relating all of these natural frequencies. Finally, the results from modal test of the simple structures are considered to be observations and used to calculate the conditional variance of the desired inducer frequencies. As this method is less rigorous than more complicated Bayesian methods reported in the literature, a conservative factor is applied to the result, but the resulting uncertainty is still significantly less than originally estimated and will greatly assist certification of the inducer for use in the engine.

Uncertainty Quantification↗

Extragalactic foreground contamination in temperature-based CMB lens reconstruction

We discuss the effect of unresolved point source contamination on estimates of the CMB lensing potential, from components such as the thermal Sunyaev-Zel'dovich effect, radio point sources, and the Cosmic Infrared Background. We classify the possible trispectra associated with such source populations, and construct estimators for the amplitude and scale-dependence of several of the major trispectra. We show how to propagate analytical models for these source trispectra to biases for lensing. We also construct a ''source-hardened'' lensing estimator which experiences significantly smaller biases when exposed to unresolved point sources than the standard quadratic lensing estimator. We demonstrate these ideas in practice using the sky simulations of Sehgal et al., for cosmic-variance limited experiments designed to mimic ACT, SPT, and Planck. We find that for radio sources and SZ the bias is significantly reduced, but for CIB it is essentially unchanged. However, by using the high-frequency, all-sky CIB measurements from Planck and Herschel it may be possible to suppress this contribution.

79 ASTRONOMY AND ASTROPHYSICS↗

A multishock model for the density variance of anisotropic, highly magnetized, supersonic turbulence

ABSTRACT Shocks form the basis of our understanding for the density and velocity statistics of supersonic turbulent flows, such as those found in the cool interstellar medium (ISM). The variance of the density field, $\sigma ^2_{\rho /\rho _0}$, is of particular interest for molecular clouds (MCs), the birthplaces of stars in the Universe. The density variance may be used to infer underlying physical processes in an MC, and parametrizes the star formation (SF) rate of a cloud. However, models for $\sigma ^2_{\rho /\rho _0}$ all share a common feature – the variance is assumed to be isotropic. This assumption does not hold when a trans-/sub-Alfvénic mean magnetic field, ${B}_0$, is present in the cloud, which observations suggest is relevant for some MCs. We develop an anisotropic model for $\sigma _{\rho /\rho _0}^2$, using contributions from hydrodynamical and fast magnetosonic shocks that propagate orthogonal to each other. Our model predicts an upper bound for $\sigma _{\rho /\rho _0}^2$ in the high Mach number $(\mathcal {M})$ limit as small-scale density fluctuations become suppressed by the strong ${B}_0$. The model reduces to the isotropic $\sigma _{\rho /\rho _0}^2\!-\!\mathcal {M}$ relation in the hydrodynamical limit. To validate our model, we calculate $\sigma _{\rho /\rho _0}^2$ from 12 high-resolution, three-dimensional, supersonic, sub-Alfvénic magnetohydrodynamical (MHD) turbulence simulations and find good agreement with our theory. We discuss how the two MHD shocks may be the bimodally oriented overdensities observed in some MCs and the implications for SF theory in the presence of a sub-Alfvénic ${B}_0$. By creating an anisotropic, supersonic density fluctuation model, this study paves the way for SF theory in the highly anisotropic regime of interstellar turbulence.

Beattie, James R.↗

U 2 Ni 2 Sn and the origin of magnetic anisotropy in uranium compounds

Here, U 2 Ni 2 Sn is a member of a large family of intermetallic compounds with the tetragonal Mo 2 FeB 2 crystal structure. It orders antiferromagnetically at 25 K with propagation vector $q = (0, 0, \frac{1}{2})$. Magnetization, magnetoacoustic, and neutron-diffraction experiments on a single crystal provide evidence that the uranium moments align parallel to the $c$ axis with the anisotropy energy of ≈170 K, indicating that U 2 Ni 2 Sn can be classified as an Ising system. The results are at variance with previous studies on polycrystals, which indicated different magnetic structure, and which were incompatible with the $5f - 5f$ two-ion anisotropy model dominant in most U band systems. High-field magnetization studies exhibit a weak linear response for fields along the basal plane up to the highest field applied (60 T), while the $c$-axis magnetization curve exhibits three metamagnetic transitions at approximately 30, 39, and 50 T. The U magnetic moments of $0.87 μ_B$, the low magnetic entropy, and the enhanced Sommerfeld coefficient $γ$ = 187 mJ/mol f.u.K 2 suggest that U 2 Ni 2 Sn can be classified as an itinerant antiferromagnet with strong electron-electron correlations.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

An estimate of equatorial wave energy flux at 9- to 90-day periods in the Central Pacific

Deep fluctuations in current along the equator in the Central Pacific are dominated by coherent structures which correspond closely to narrow-band propagating equatorial waves. Currents were measured roughly at 1500 and 3000 m depths at five moorings between 144 and 148 deg W from January 1981 to March 1983, as part of the Pacific Equatorial Ocean Dynamics program. In each frequency band resolved, a single complex empirical orthogonal function accounts for half to three quarters of the observed variance in either zonal or meridional current. Dispersion for equatorial first meridional Rossby and Rossby gravity waves is consistent with the observed vertical-zonal coherence structure. The observations indicate that energy flux is westward and downward in long first meridional mode Rossby waves at periods 45 days and longer, and eastward and downward in short first meridional mode Rossby waves and Rossby-gravity waves at periods 30 days and shorter. A local minimum in energy flux occurs at periods corresponding to a maximum in upper-ocean meridional current energy contributed by tropical instability waves. Total vertical flux across the 9- to 90-day period range is 2.5 kW/m.

Eriksen, Charles C.↗

Improved statistical models for limited datasets in uncertainty quantification using stochastic collocation

This paper presents a data-driven framework for performing uncertainty quantification (UQ) by choosing a stochastic model that accurately describes the sources of uncertainty in a system. This model is propagated through an appropriate response surface function that approximates the behavior of this system using stochastic collocation. Given a sample of data describing the uncertainty in the inputs, our goal is to estimate a probability density function (PDF) using the kernel moment matching (KMM) method so that this PDF can be used to accurately reproduce statistics like mean and variance of the response surface function. Instead of constraining the PDF to be optimal for a particular response function, we show that we can use the properties of stochastic collocation to make the estimated PDF optimal for a wide variety of response functions. We contrast this method with other traditional procedures that rely on the Maximum Likelihood approach, like kernel density estimation (KDE) and its adaptive modification (AKDE). We argue that this modified KMM method tries to preserve what is known from the given data and is the better approach when the available data is limited in quantity. We test the performance of these methods for both univariate and multivariate density estimation by sampling random datasets from known PDFs and then measuring the accuracy of the estimated PDFs, using the known PDF as a reference. Comparing the output mean and variance estimated with the empirical moments using the raw data sample as well as the actual moments using the known PDF, we show that the KMM method performs better than KDE and AKDE in predicting these moments with greater accuracy. This improvement in accuracy is also demonstrated for the case of UQ in electrostatic and electrothermomechanical microactuators. We show how our framework results in the accurate computation of statistics in micromechanical systems.

97 MATHEMATICS AND COMPUTING↗

Removing the Effects of Tropical Dynamics from North Pacific Climate Variability

Teleconnections from the Tropics energize variations of the North Pacific climate, but detailed diagnosis of this relationship has proven difficult. Simple univariate methods, such as regression on El Niño-Southern Oscillation (ENSO) indices, may be inadequate since the key dynamical processes involved -- including ENSO diversity in the Tropics, re-emergence of mixed layer thermal anomalies, and oceanic Rossby wave propagation in the North Pacific -- have a variety of overlapping spatial and temporal scales. Here we use a multivariate Linear Inverse Model to quantify tropical and extra-tropical multi-scale dynamical contributions to North Pacific variability, in both observations and CMIP6 models. In observations, we find that the Tropics are responsible for almost half of the seasonal variance, and almost three quarters of the decadal variance, along the North American coast and within the subtropical front region northwest of Hawaii. SST anomalies that are generated by local dynamics within the Northeast Pacific have much shorter time scales, consistent with transient weather forcing by Aleutian low anomalies. Variability within the Kuroshio-Oyashio Extension (KOE) region is considerably less impacted by the Tropics, on all time scales. Consequently, without tropical forcing the dominant pattern of North Pacific variability would be a KOE pattern, rather than the Pacific Decadal Oscillation (PDO). In contrast to observations, most CMIP6 historical simulations produce North Pacific variability that maximizes in the KOE region, with amplitude significantly higher than observed. Correspondingly, the simulated North Pacific in all CMIP6 models is shown to be relatively insensitive to the Tropics, with a dominant spatial pattern generally resembling the KOE pattern, not the PDO.

54 ENVIRONMENTAL SCIENCES↗

Inventory Uncertainty Quantification using TENDL Covariance Data in Fispact-II

The new inventory code Fispact-II provides predictions of inventory, radiological quantities and their uncertainties using nuclear data covariance information. Central to the method is a novel fast pathways search algorithm using directed graphs. The pathways output provides (1) an aid to identifying important reactions, (2) fast estimates of uncertainties, (3) reduced models that retain important nuclides and reactions for use in the code's Monte Carlo sensitivity analysis module. Described are the methods that are being implemented for improving uncertainty predictions, quantification and propagation using the covariance data that the recent nuclear data libraries contain. In the TENDL library, above the upper energy of the resolved resonance range, a Monte Carlo method in which the covariance data come from uncertainties of the nuclear model calculations is used. The nuclear data files are read directly by FISPACT-II without any further intermediate processing. Variance and covariance data are processed and used by FISPACT-II to compute uncertainties in collapsed cross sections, and these are in turn used to predict uncertainties in inventories and all derived radiological data.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Nonlinear evolution of baryon acoustic oscillations from improved perturbation theory in real and redshift spaces

We study the nonlinear evolution of baryon acoustic oscillations in the matter power spectrum and correlation function from the improved perturbation theory (PT). Based on the framework of renormalized PT, which provides a nonperturbative way to treat the gravitational clustering of large-scale structure, we apply the closure approximation that truncates the infinite series of loop contributions at one-loop order, and obtain a closed set of integral equations for power spectrum and nonlinear propagator. The resultant integral expressions are basically equivalent to those previously derived in the form of evolution equations, and they keep important nonperturbative properties which can dramatically improve the prediction of nonlinear power spectrum. Employing the Born approximation, we then derive the analytic expressions for nonlinear power spectrum and the predictions are made for nonlinear evolution of baryon acoustic oscillations in power spectrum and correlation function. We find that the improved PT possesses a better convergence property compared with standard PT calculation. A detailed comparison between improved PT results and N-body simulations shows that a percent-level agreement is achieved in a certain range in power spectrum and in a rather wider range in correlation function. Combining a model of nonlinear redshift-space distortion, we also evaluate the power spectrum and correlation function in redshift space. In contrast to the results in real space, the agreement between N-body simulations and improved PT predictions tends to be worse, and a more elaborate modeling for redshift-space distortion needs to be developed. Nevertheless, with the currently existing model, we find that the prediction of correlation function has a sufficient accuracy compared with the cosmic-variance errors for future galaxy surveys with volume of a few h{sup -3} Gpc{sup 3} at z > or approx. 0.5.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

The influence of physical and algorithmic factors on simulated far-field waveforms and source–time functions of underground explosions using unsupervised machine learning

SUMMARY Characterizing explosion sources and differentiating between earthquake and underground explosions using distributed seismic networks becomes non-trivial when explosions are detonated in cavities or heterogeneous ground material. Moreover, there is little understanding of how changes in subsurface physical properties affect the far-field waveforms we record and use to infer information about the source. Simulations of underground explosions and the resultant ground motions can be a powerful tool to systematically explore how different subsurface properties affect far-field waveform features, but there are added variables that arise from how we choose to model the explosions that can confound interpretation. To assess how both subsurface properties and algorithmic choices affect the seismic wavefield and the estimated source functions, we ran a series of 2-D axisymmetric non-linear numerical explosion experiments and wave propagation simulations that explore a wide array of parameters. We then inverted the synthetic far-field waveform data using a linear inversion scheme to estimate source–time functions (STFs) for each simulation case. We applied principal component analysis (PCA), an unsupervised machine learning method, to both the far-field waveforms and STFs to identify the most important factors that control variance in the waveform data and differences between cases. For the far-field waveforms, the largest variance occurs in the shallower radial receiver channels in the 0–50 Hz frequency band. For the STFs, both peak amplitude and rise times across different frequencies contribute to the variance. We find that the ground equation of state (i.e. lithology and rheology) and the explosion emplacement conditions (i.e. tamped versus cavity) have the greatest effect on the variance of the far-field waveforms and STFs, with the ground yield strength and fracture pressure being secondary factors. Differences in the PCA results between the far-field waveforms and STFs could possibly be due to near-field non-linearities of the source that are not accounted for in the estimation of STFs and could be associated with yield strength, fracture pressure, cavity radius and cavity shape parameters. Other algorithmic parameters are found to be less important and cause less variance in both the far-field waveforms and STFs, meaning algorithmic choices in how we model explosions are less important, which is encouraging for the further use of explosion simulations to study how physical Earth properties affect seismic waveform features and estimated STFs.

58 GEOSCIENCES↗

Derotation of the cosmic microwave background polarization: Full-sky formalism

Mechanisms have been proposed that might rotate the linear polarization of the cosmic microwave background (CMB) as it propagates from the surface of last scatter. In the simplest scenario, the rotation will be uniform across the sky, but the rotation angle may also vary across the sky. We develop in detail the complete set of full-sky quadratic estimators for the rotation of the CMB polarization that can be constructed from the CMB temperature and polarization. We derive the variance with which these estimators can be measured and show that these variances reduce to the simpler flat-sky expressions in the appropriate limit. We evaluate the variances numerically. While the flat-sky formalism may be suitable if the rotation angle arises as a realization of a random field, the full-sky formalism will be required to search for rotations that vary slowly across the sky as well as for models in which the angular power spectrum for the rotation angle peaks at large angles.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

An adaptive ANOVA-based PCKF for high-dimensional nonlinear inverse modeling

The probabilistic collocation-based Kalman filter (PCKF) is a recently developed approach for solving inverse problems. It resembles the ensemble Kalman filter (EnKF) in every aspect—except that it represents and propagates model uncertainty by polynomial chaos expansion (PCE) instead of an ensemble of model realizations. Previous studies have shown PCKF is a more efficient alternative to EnKF for many data assimilation problems. However, the accuracy and efficiency of PCKF depends on an appropriate truncation of the PCE series. Having more polynomial chaos basis functions in the expansion helps to capture uncertainty more accurately but increases computational cost. Selection of basis functions is particularly important for high-dimensional stochastic problems because the number of polynomial chaos basis functions required to represent model uncertainty grows dramatically as the number of input parameters (random dimensions) increases. In classic PCKF algorithms, the PCE basis functions are pre-set based on users' experience. Also, for sequential data assimilation problems, the basis functions kept in PCE expression remain unchanged in different Kalman filter loops, which could limit the accuracy and computational efficiency of classic PCKF algorithms. To address this issue, we present a new algorithm that adaptively selects PCE basis functions for different problems and automatically adjusts the number of basis functions in different Kalman filter loops. The algorithm is based on adaptive functional ANOVA (analysis of variance) decomposition, which approximates a high-dimensional function with the summation of a set of low-dimensional functions. Thus, instead of expanding the original model into PCE, we implement the PCE expansion on these low-dimensional functions, which is much less costly. We also propose a new adaptive criterion for ANOVA that is more suited for solving inverse problems. The new algorithm was tested with different examples and demonstrated great effectiveness in comparison with non-adaptive PCKF and EnKF algorithms.

97 MATHEMATICS AND COMPUTING↗