SEARCH · Search NASA

Results for “stochastic approaches”

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.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 379 records · Page 21

All sources NASA NTRS DOE OSTI

DOE OSTI2023

Tailoring resonant magnetic perturbation to optimize fast-ion confinement during ELM control in KSTAR

Abstract 3D resonant magnetic perturbation (RMP) is one promising way to control edge localized modes that can cause excessive material erosion of tokamak first walls. However, RMP can lead to undesired degradation of plasma confinement, including fast-particle losses, which can impact the performance and safety of the reactor. This work investigates the optimization of the poloidal spectrum of the 3D field to optimize fast ion confinement during edge localized mode (ELM) suppression. In the initial step, the validity of the modeling framework is tested against experimental data. Simulations successfully replicate an increase in poloidal limiter temperature with different poloidal spectra. Then, the simulation shows improvement of fast ion confinement with a reduction of core resonant response, while edge resonant magnetic fields are maintained above the threshold to sustain the ELM suppression. Reduction of the core resonant fields keeps the Kolmogorov–Arnold–Moser surface and reduces the fast particle losses due to the stochastic magnetic field lines. The results highlight the potential of edge localization of the resonant fields to enhance the performance of fusion reactors, but further investigation is needed to improve the validation of this approach.

3D field↗

DOE OSTI2020

Real-time Ridesharing for Transportation Hubs with Demand and Supply Uncertainty

Transportation hubs in major cities generate a significant amount of trips by taxis and for-hail vehicles (FHV), with many of the trips sharing similar destinations. This suggests promising opportunities to leverage the collective travel needs with dedicated ridesharing solutions to reduce the externalities of excessive traffic at transportation hubs. In this study, we develop a novel dynamic ridesharing approach to serve trips from the transportation hub by considering (1) demand (new passengers) and supply (newly available vehicles) in the near future and (2) the uncertainty of future predictions. Our approach consists of two stages. In the first stage, we develop a data structure called hub mobility tree to generate potential combinations of shareable trips as candidate schedules efficiently. Then the generated schedules are used in the second stage to formulate the stochastic hub-based ridesharing problem (SHRP), which is a stochastic integer programming problem with the objective to maximize the total expected ridesharing profit over time. Due to the prohibitive number of shareable trips, we then approximately solve SHRP by the sample average approximate method (SAA), and a dual Lagrangian technique is implemented to further improve the scalability of the solution approach. We demonstrate the performances of the proposed method by simulating the ridesharing service at JFK airport using NYC taxi and FHV data. The results indicate that the proposed method outperforms the myopic ridesharing (maximize profit for a single time step) and the rolling horizon method with point estimation of future demand and supply.

dynamic ridesharing↗

DOE OSTI2021

Simulated Dust Transport in the Convective Boundary Layer

Soil dusts are an important source of aerosol in agricultural regions and can affect the Earth's radiation budget through the modification of cloud properties, and in particular, through their ability to act as ice nucleating particles. In order to impact cloud properties, agricultural soil dusts need to be transported from the point of emission to cloud–relevant altitudes. Vertical transport within the planetary boundary layer is strongly controlled by turbulence and is challenging to represent accurately in regional and global models. Large–eddy simulations (LES) are run at resolutions capable of resolving most of the turbulent energy directly and can thus better simulate vertical transport. In this study, we leverage the LES ARM Symbiotic Simulation and Observation (LASSO) large–eddy simulations to simulate vertical transport of agricultural dust within the turbulent boundary layer using a modified version of the stochastic dispersion model FLEXPART–WRF. We find that the modified model is better capable of simulating particle transport due to turbulence, and that particle size was the greatest factor in determining particle lifetime. Individual meteorology and particle density had intermediate effects upon particle transport, while release height had little effect upon simulation results. Finally, we utilize a quasi–single column model (QSCM) approach to determine how our results compare to a parameterized treatment of turbulence. The QSCM simulations led to greater tracer transport out of the boundary layer, with ramifications for any studies utilizing a Lagrangian stochastic model to understand tracer dispersion. Furthermore, these results highlight the importance of accurately simulating turbulence for understanding particle transport.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2020

Modern Monte Carlo methods for efficient uncertainty quantification and propagation: A survey

Uncertainty quantification (UQ) includes the characterization, integration, and propagation of uncertainties that result from stochastic variations and a lack of knowledge or data in the natural world. Monte Carlo (MC) method is a sampling-based approach that has widely used for quantification and propagation of uncertainties. However, the standard MC method is often time-consuming if the simulation-based model is computationally intensive. This article gives an overview of modern MC methods to address the existing challenges of the standard MC in the context of UQ. Specifically, multilevel Monte Carlo (MLMC) extending the concept of control variates achieves a significant reduction of the computational cost by performing most evaluations with low accuracy and corresponding low cost, and relatively few evaluations at high accuracy and corresponding high cost. Multifidelity Monte Carlo (MFMC) accelerates the convergence of standard Monte Carlo by generalizing the control variates with different models having varying fidelities and varying computational costs. Multimodel Monte Carlo method (MMMC), having a different setting of MLMC and MFMC, aims to address the issue of UQ and propagation when data for characterizing probability distributions are limited. Multimodel inference combined with importance sampling is proposed for quantifying and efficiently propagating the uncertainties resulting from small data sets. All of these three modern MC methods achieve a significant improvement of computational efficiency for probabilistic UQ, particularly uncertainty propagation. Finally, an algorithm summary and the corresponding code implementation are provided for each of the modern MC methods. The extension and application of these methods are discussed in detail.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2025

On-lattice kinetic Monte Carlo approaches for modeling molecular anisotropy in resveratrol crystallization

Stilbenes are a class of organic compounds with broad-ranging pharmaceutical and agricultural applications, which are typically isolated and purified through recrystallization. We are motivated by reducing experimental waste and optimizing yield via developing predictive simulations for processing-dependent crystal morphologies. Using resveratrol as a model stilbene system, we have developed an approach for simulating crystallization with molecular resolution using on-lattice kinetic Monte Carlo. In this work, we highlight modifications to the Stochastic Parallel PARticle Kinetic Simulator (SPPARKS) software package, which were essential to this application. Key enhancements include the incorporation of non-orthogonal cell shapes and monomer anisotropy approximations using bound hard spheres. This new SPPARKS application has been applied to resveratrol with attachment energy libraries obtained from density functional theory, resulting in excellent agreement with experimental morphology prediction.

crystallization↗

DOE OSTI2022

Machine learning and atomic layer deposition: Predicting saturation times from reactor growth profiles using artificial neural networks

In this work, we explore the application of deep neural networks to the optimization of atomic layer deposition (ALD) processes. In particular, we focus on a one-shot optimization problem, where we try to predict the optimal dose time that leads to saturation everywhere in the reactor based on thickness values measured at different points of an ALD reactor after a single trial growth. In order to tackle this problem, we introduce a dataset designed to train neural networks to predict saturation times based on these inputs for a cross-flow ALD reactor. Here, we then explore the predictive ability of artificial neural networks of different depths and sizes using a separate testing dataset to evaluate their accuracies. The results obtained show that networks trained using stochastic gradient descent methods can accurately predict saturation times without requiring any additional information on the surface kinetics. This provides a viable approach to minimize the number of experiments required to optimize new ALD processes in a known reactor, and it highlights the way machine learning can be leveraged for thin film growth and manufacturing. While the datasets and training procedure depend on the reactor geometry, the trained neural networks provide a general surrogate model connecting thickness values and trial dose times with optimal saturation times that can be reused for different ALD processes within the same reactor.

36 MATERIALS SCIENCE↗

DOE OSTI2020

Learning in Modal Space: Solving Time-Dependent Stochastic PDEs Using Physics-Informed Neural Networks

One of the open problems in scientific computing is the long-time integration of nonlinear stochastic partial differential equations (SPDEs), especially with arbitrary initial data. We address this problem by taking advantage of recent advances in scientific machine learning and the spectral dynamically orthogonal (DO) and borthogonal (BO) methods for representing stochastic processes. The recently introduced DO/BO methods reduce the SPDE to solving a system of deterministic PDEs and a system of stochastic ordinary differential equations. Specifically, we propose two new physics-informed neural networks (PINNs) for solving time-dependent SPDEs, namely the neural network (NN)-DO/BO methods. The proposed methods incorporate the DO/BO constraints into the loss function (along with the modal decomposition of the SPDE) with an implicit form instead of generating explicit expressions for the temporal derivatives of the DO/BO modes. Hence, the NN-DO/BO methods can overcome some of the drawbacks of the original DO/BO methods. For example, we do not need the assumption that the covariance matrix of the random coefficients is invertible as in the original DO method, and we can remove the assumption of no eigenvalue crossing as in the original BO method. Moreover, the NN-DO/BO methods can be used to solve time-dependent stochastic inverse problems with the same formulation and same computational complexity as for forward problems. Furthermore, we demonstrate the capability of the proposed methods via several numerical examples, namely: (1) A linear stochastic advection equation with deterministic initial condition: we obtain good results with the proposed methods, while the original DO/BO methods cannot be applied directly in this case. (2) Long-time integration of the stochastic Burgers' equation: we show the good performance of NN-DO/BO methods, especially the effectiveness of the NN-BO approach for such problems with many eigenvalue crossings during the whole time evolution, while the original BO method fails. (3) Nonlinear reaction diffusion equation: we consider both the forward problem and the inverse problems, including very noisy initial point values, to investigate the flexibility of the NN-DO/BO methods in handling inverse and mixed type problems. Taken together, these simulation results demonstrate that the NN-DO/BO methods can be employed to effectively quantify uncertainty propagation in a wide range of physical problems, but future work should address the efficiency issue of PINNs for forward problems.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2023

Ground Motion Models (GMMs) Improvements Using Earthquake Simulations on High Performance Computers

A computationally efficient simulation platform was developed that can provide representative synthetic ground motions from crustal earthquakes in the Stable Continental Regions of Central and Eastern US (CEUS), using 3D modeling and high-performance computing. The main objective was to use synthetic ground motion to provide constrains to refinements of exiting ergodic Ground Motion Models (GMMs), for large magnitude earthquakes and near-fault distances, for which these models are less reliable. Physics-based broadband (0-5Hz) ground motion simulations were used to estimate the near-fault ground motion amplitudes and within event and between-event variabilities associated with fault rupture characteristics. As part of a strategy for selecting a reginal velocity model and validation of developed rupture modeling technique, ground motions from the moment magnitude Mw5.0 November 7, 2016, Cushing Oklahoma, and Mw5.8 September 3, 2016, Pawnee Oklahoma earthquakes were simulated. In our simulations we used a 3D regional velocity model that was based on Saikia’s 1D velocity model. Saikia’s model demonstrated better performance in modelling high frequency regional wave propagation for CEUS region. The proposed 3D model includes lateral variations added to the 1D background model using the stochastic scheme of Pitarka and Mellors. Comparisons of the simulations with recordings of both earthquakes demonstrated the reliability of our deterministic simulation approach while emphasizing the importance of including small-scale variability in the regional velocity model needed to reproduce the observed high-frequency wave scattering effects. As part of validation analysis, comparisons with different GMMs for a Mw6.5 earthquake in the CESUS region resulted in a very good match between the simulated and empirical ground motion models. Initial investigations of within-event and between-event ground motion variabilities for Mw6.5 scenario earthquakes on a strike-slip fault, suggest that they are strongly related to spatial slip and slip rate variations, average rupture velocity, rupture area and rupture initiation location. For certain scenarios we found that the ground motion variability observed at near-fault distances (< 5 km) also persists at longer distances. Regardless of the rupture scenario, the simulated ground motion tends to fully saturate at short distances and for all periods. The near-fault saturation has to do with the attenuation of waves propagating along the fault and local rupture radiation pattern that also contribute to stronger ground motion variation at such distances. Analysis of effects of rupture initiation location suggest that the peak ground motion (PGV) and spectral acceleration (SA) can be quite variable due to rupture directivity effects. Such effects are stronger at periods longer than 1s. The effect of the 1D velocity models and surface topography on simulated ground motion were investigated by comparing three component synthetic seismograms computed at selected sites. Effect of surface topography was considered using the ratio between spectral accelerations simulated for two 1D models with flat surface topography and realistic model with surface topography. Overall, the topography slightly amplifies (by ~30%) the ground motion amplitude in the frequency range 1-3Hz. The effect of topography is more visible in the surface and coda waves portion of the seismograms.

58 GEOSCIENCES↗

DOE OSTI2020

Dynamical Mass Estimates of the β Pictoris Planetary System through Gaussian Process Stellar Activity Modeling

Nearly 15 yr of radial velocity (RV) monitoring and direct imaging enabled the detection of two giant planets orbiting the young, nearby star β Pictoris. The δ Scuti pulsations of the star, which overwhelm planetary signals, need to be carefully suppressed. In this work, we independently revisit the analysis of the RV data following a different approach than available in the literature to model the activity of the star. We show that a Gaussian process (GP) with a stochastically driven damped harmonic oscillator kernel can model the δ Scuti pulsations. It provides similar results to parametric models but with a simpler framework, using only three hyperparameters. It also enables us to model poorly sampled RV data that were excluded from previous analyses, hence extending the RV baseline by nearly five years. Altogether, the orbit and mass of both planets can be constrained from RV only, which was not possible with the parametric modeling. To characterize the system more accurately, we also perform a joint fit of all available relative astrometry and RV data. Our orbital solutions for β Pic b favor a low eccentricity of 0.029$_{−0.024}^{+0.061}$ and a relatively short period of 21.1$_{−0.8}^{+2.0}$ yr. The orbit of β Pic c is eccentric with 0.206$_{−0.063}^{+0.074}$ with a period of 3.36 ± 0.03 yr. We find model-independent masses of 11.7 ± 1.4 and 8.5 ± 0.5 M Jup for β Pic b and c, respectively, assuming coplanarity. The mass of β Pic b is consistent with the hottest start evolutionary models, at an age of 25 ± 3 Myr. A direct detection of β Pic c would provide a second calibration measurement in a coeval system.

79 ASTRONOMY AND ASTROPHYSICS↗

DOE OSTI2020

Wake steering optimization under uncertainty

Turbines in wind power plants experience significant power losses when wakes from upstream turbines affect the energy production of downstream turbines. A promising plant-level control strategy to reduce these losses is wake steering, where upstream turbines are yawed to direct wakes away from downstream turbines. However, there are significant uncertainties in many aspects of the wake steering problem. For example, infield sensors do not give perfect information, and inflow to the plant is complex and difficult to forecast with available information, even over short time periods. Here, we formulate and solve an optimization under uncertainty (OUU) problem for determining optimal plant-level wake steering strategies in the presence of independent uncertainties in the direction, speed, turbulence intensity, and shear of the incoming wind, as well as in turbine yaw positions. The OUU wake steering strategy is first examined for a two-turbine test case to explore the impacts of different types of inflow uncertainties, and it is then demonstrated for a more realistic 11-turbine wind power plant. Of the sources of uncertainty considered, we find that wake steering strategies are most sensitive to uncertainties in the wind speed and direction. When maximizing expected power production, the OUU strategy also tends to favor smaller yaw angles, which have been shown in previous work to reduce turbine loading. Ultimately, the plant-level wake steering strategy formulated using an OUU approach yields 0.48?% more expected annual energy production for the 11-turbine wind plant than a strategy that neglects uncertainty when considering stochastic inputs. Thus, not only does the present OUU strategy produce more power in realistic conditions, but it also reduces risk by prescribing strategies that call for less extreme yaw angles.

17 WIND ENERGY↗

DOE OSTI2020

Dataset for "FLEXPART-WRF simulations to assess dust transport in the convective boundary layer at the Southern Great Plains site"

Soil dusts are an important source of aerosol in agricultural regions and can affect the Earth’s radiation budget through the modification of cloud properties, and in particular, through their ability to act as ice nucleating particles. In order to impact cloud properties, agricultural soil dusts need to be transported from the point of emission to cloud-relevant altitudes. Vertical transport within the planetary boundary layer is strongly controlled by turbulence and is challenging to represent accurately in regional and global models. Large-eddy simulations (LES) are run at resolutions capable of resolving most of the turbulent energy directly and can thus better simulate vertical transport. In this study, we leverage the LES ARM Symbiotic Simulation and Observation (LASSO) large-eddy simulations to simulate vertical transport of agricultural dust within the turbulent boundary layer using a modified version of the stochastic dispersion model FLEXPART-WRF. We find that the modified model is better capable of simulating particle transport due to turbulence, and that particle size was the greatest factor in determining particle lifetime. Individual meteorology and particle density had intermediate effects upon particle transport, while release height had little effect upon simulation results. Finally, we utilize a quasi-single column model (QSCM) approach to determine how our results compare to a parameterized treatment of turbulence. The QSCM simulations led to greater tracer transport out of the boundary layer, with ramifications for any studies utilizing a Lagrangian stochastic model to understand tracer dispersion. These results highlight the importance of accurately simulating turbulence for understanding particle transport.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2022

Moving the needle: Employing deep reinforcement learning to push the boundaries of coarse-grained vaccine models

Highly mutable infectious disease pathogens (hm-IDPs) such as HIV and influenza evolve faster than the human immune system can contain them, allowing them to circumvent traditional vaccination approaches and causing over one million deaths annually. Agent-based models can be used to simulate the complex interactions that occur between immune cells and hm-IDP-like proteins (antigens) during affinity maturation—the process by which antibodies evolve. Compared to existing experimental approaches, agent-based models offer a safe, low-cost, and rapid route to study the immune response to vaccines spanning a wide range of design variables. However, the highly stochastic nature of affinity maturation and vast sequence space of hm-IDPs render brute force searches intractable for exploring all pertinent vaccine design variables and the subset of immunization protocols encompassed therein. To address this challenge, we employed deep reinforcement learning to drive a recently developed agent-based model of affinity maturation to focus sampling on immunization protocols with greater potential to improve the chosen metrics of protection, namely the broadly neutralizing antibody (bnAb) titers or fraction of bnAbs produced. Using this approach, we were able to coarse-grain a wide range of vaccine design variables and explore the relevant design space. Our work offers new testable insights into how vaccines should be formulated to maximize protective immune responses to hm-IDPs and how they can be minimally tailored to account for major sources of heterogeneity in human immune responses and various socioeconomic factors. Our results indicate that the first 3 to 5 immunizations, depending on the metric of protection, should be specially tailored to achieve a robust protective immune response, but that beyond this point further immunizations require only subtle changes in formulation to sustain a durable bnAb response.

60 APPLIED LIFE SCIENCES↗

DOE OSTI2022

Non-intrusive reduced order modeling of natural convection in porous media using convolutional autoencoders: Comparison with linear subspace techniques

Natural convection in porous media is a highly nonlinear multiphysical problem relevant to many engineering applications (e.g., the process of CO 2 sequestration). Here, we extend and present a non-intrusive reduced order model of natural convection in porous media employing deep convolutional autoencoders for the compression and reconstruction and either radial basis function (RBF) interpolation or artificial neural networks (ANNs) for mapping parameters of partial differential equations (PDEs) on the corresponding nonlinear manifolds. To benchmark our approach, we also describe linear compression and reconstruction processes relying on proper orthogonal decomposition (POD) and ANNs. Further, we present comprehensive comparisons among different models through three benchmark problems. The reduced order models, linear and nonlinear approaches, are much faster than the finite element model, obtaining a maximum speed-up of 7 × 10 6 because our framework is not bound by the Courant–Friedrichs–Lewy condition; hence, it could deliver quantities of interest at any given time contrary to the finite element model. Our model’s accuracy still lies within a relative error of 7% in the worst-case scenario. We illustrate that, in specific settings, the nonlinear approach outperforms its linear counterpart and vice versa. We hypothesize that a visual comparison between principal component analysis (PCA) and t-Distributed Stochastic Neighbor Embedding (t-SNE) could indicate which method will perform better prior to employing any specific compression strategy.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2021

High-level coupled-cluster energetics by merging moment expansions with selected configuration interaction

Inspired by our earlier semi-stochastic work aimed at converging high-level coupled-cluster (CC) energetics, we propose a novel form of the CC(P; Q) theory in which the stochastic Quantum Monte Carlo propagations, used to identify dominant higher-than-doubly excited determinants, are replaced by the selected configuration interaction (CI) approach using the perturbative selection made iteratively (CIPSI) algorithm. The advantages of the resulting CIPSI-driven CC(P; Q) methodology are illustrated by a few molecular examples, including the dissociation of F 2 and the automerization of cyclobutadiene, where we recover the electronic energies corresponding to the CC calculations with a full treatment of singles, doubles, and triples based on the information extracted from compact CI wave functions originating from relatively inexpensive Hamiltonian diagonalizations.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2020

Fat Crayon Toolkit

The Fat Crayon Toolkit (FCT) provides python modules for assisting in the design and visualization of boreholes and tunnels and their relationship with natural features such as fractures. Distances of closest approach among the boreholes can be calculated, along with other key design metrics. Additionally, the shear propensity of natural fractures (either deterministic or stochastic) can be visualized in either 3D (as VTK files) or in 2D (stereonets). The example provided demonstrates a test bed developed for the Enhanced Geothermal Systems (EGS) Collab project. This material is based upon work supported by the Geothermal Office of the Department of Energy and the Sanford Underground Research Facility and was prepared by LLNL under Contract DE-AC52-07NA27344.

Morris, JosephP.↗

DOE OSTI2021

Discrete ion stochastic continuum overdamped solvent algorithm for modeling electrolytes

In this paper we develop a methodology for the mesoscale simulation of strong electrolytes. The methodology is an extension of the fluctuating immersed-boundary approach that treats a solute as discrete Lagrangian particles that interact with Eulerian hydrodynamic and electrostatic fields. In both algorithms the immersed-boundary method of Peskin is used for particle-field coupling. Hydrodynamic interactions are taken to be overdamped, with thermal noise incorporated using the fluctuating Stokes equation, including a "dry diffusion" Brownian motion to account for scales not resolved by the coarse-grained model of the solvent. Long-range electrostatic interactions are computed by solving the Poisson equation, with short-range corrections included using an immersed-boundary variant of the classical particle-particle particle-mesh technique. Also included is a short-range repulsive force based on the Weeks-Chandler-Andersen potential. This methodology is validated by comparison to Debye-Hückel theory for ion-ion pair correlation functions, and Debye-Hückel-Onsager theory for conductivity, including the Wien effect for strong electric fields. In each case, good agreement is observed, provided that hydrodynamic interactions at the typical ion-ion separation are resolved by the fluid grid.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2022

Learning functional priors and posteriors from data and physics

In this work, we develop a new Bayesian framework based on deep neural networks to be able to extrapolate in space-time using historical data and to quantify uncertainties arising from both noisy and gappy data in physical problems. Specifically, the proposed approach has two stages: (1) prior learning and (2) posterior estimation. At the first stage, we employ the physics-informed Generative Adversarial Networks (PI-GAN) to learn a functional prior either from a prescribed function distribution, e.g., Gaussian process, or from historical data and physics. At the second stage, we employ the Hamiltonian Monte Carlo (HMC) method to estimate the posterior in the latent space of PI-GANs. In addition, we use two different approaches to encode the physics: (1) automatic differentiation, used in the physicsinformed neural networks (PINNs) for scenarios with explicitly known partial differential equations (PDEs), and (2) operator regression using the deep operator network (DeepONet) for PDE-agnostic scenarios. We then test the proposed method for (1) meta-learning for one-dimensional regression, and forward/inverse PDE problems (combined with PINNs); (2) PDE-agnostic physical problems (combined with DeepONet), e.g., fractional diffusion as well as saturated stochastic (100-dimensional) flows in heterogeneous porous media; and (3) spatial-temporal regression problems, i.e., inference of a marine riser displacement field using experimental data from the Norwegian Deepwater Programme (NDP). The results demonstrate that the proposed approach can provide accurate predictions as well as uncertainty quantification given very limited scattered and noisy data, since historical data could be available to provide informative priors. In summary, the proposed method is capable of learning flexible functional priors, e.g., both Gaussian and non-Gaussian process, and can be readily extended to big data problems by enabling mini-batch training using stochastic HMC or normalizing flows since the latent space is generally characterized as low dimensional.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

DOE OSTI2022

A survey of unsupervised learning methods for high-dimensional uncertainty quantification in black-box-type problems

Constructing surrogate models for uncertainty quantification (UQ) on complex partial differential equations (PDEs) having inherently high-dimensional O(10 n ), n ≥ 2, stochastic inputs (e.g., forcing terms, boundary conditions, initial conditions) poses tremendous challenges. The “curse of dimensionality” can be addressed with suitable unsupervised learning techniques used as a pre-processing tool to encode inputs onto lower-dimensional subspaces while retaining its structural information and meaningful properties. In this work, we review and investigate thirteen dimension reduction methods including linear and nonlinear, spectral, blind source separation, convex and non-convex methods and utilize the resulting embeddings to construct a mapping to quantities of interest via polynomial chaos expansions (PCE). Here, we refer to the general proposed approach as manifold PCE (m-PCE), where manifold corresponds to the latent space resulting from any of the studied dimension reduction methods. To investigate the capabilities and limitations of these methods we conduct numerical tests for three physics-based systems (treated as black-boxes) having high-dimensional stochastic inputs of varying complexity modeled as both Gaussian and non-Gaussian random fields to investigate the effect of the intrinsic dimensionality of input data. We demonstrate both the advantages and limitations of the unsupervised learning methods and we conclude that a suitable m-PCE model provides a cost-effective approach compared to alternative algorithms proposed in the literature, including recently proposed expensive deep neural network-based surrogates and can be readily applied for high-dimensional UQ in stochastic PDEs.

42 ENGINEERING↗