arXiv subjects

Dimitrios Loukrezis

Publications and source records attributed to Dimitrios Loukrezis.

30 records · Page 2Linked to original sources

arXiv2021

Grassmannian diffusion maps based surrogate modeling via geometric harmonics

In this paper, a novel surrogate model based on the Grassmannian diffusion maps (GDMaps) and utilizing geometric harmonics is developed for predicting the response of engineering systems and complex physical phenomena. The method utilizes the GDMaps to obtain a low-dimensional representation of the underlying behavior of physical/mathematical systems with respect to uncertainties in the input parameters. Using this representation, geometric harmonics, an out-of-sample function extension technique, is employed to create a global map from the space of input parameters to a Grassmannian diffusion manifold. Geometric harmonics is also employed to locally map points on the diffusion manifold onto the tangent space of a Grassmann manifold. The exponential map is then used to project the points in the tangent space onto the Grassmann manifold, where reconstruction of the full solution is performed. The performance of the proposed surrogate modeling is verified with three examples. The first problem is a toy example used to illustrate the development of the technique. In the second example, errors associated with the various mappings employed in the technique are assessed by studying response predictions of the electric potential of a dielectric cylinder in a homogeneous electric field. The last example applies the method for uncertainty prediction in the strain field evolution in a model amorphous material using the shear transformation zone (STZ) theory of plasticity. In all examples, accurate predictions are obtained, showing that the present technique is a strong candidate for the application of uncertainty quantification in large-scale models.

physics.data-an↗

arXiv2021

Manifold learning-based polynomial chaos expansions for high-dimensional surrogate models

In this work we introduce a manifold learning-based method for uncertainty quantification (UQ) in systems describing complex spatiotemporal processes. Our first objective is to identify the embedding of a set of high-dimensional data representing quantities of interest of the computational or analytical model. For this purpose, we employ Grassmannian diffusion maps, a two-step nonlinear dimension reduction technique which allows us to reduce the dimensionality of the data and identify meaningful geometric descriptions in a parsimonious and inexpensive manner. Polynomial chaos expansion is then used to construct a mapping between the stochastic input parameters and the diffusion coordinates of the reduced space. An adaptive clustering technique is proposed to identify an optimal number of clusters of points in the latent space. The similarity of points allows us to construct a number of geometric harmonic emulators which are finally utilized as a set of inexpensive pre-trained models to perform an inverse map of realizations of latent features to the ambient space and thus perform accurate out-of-sample predictions. Thus, the proposed method acts as an encoder-decoder system which is able to automatically handle very high-dimensional data while simultaneously operating successfully in the small-data regime. The method is demonstrated on two benchmark problems and on a system of advection-diffusion-reaction equations which model a first-order chemical reaction between two species. In all test cases, the proposed method is able to achieve highly accurate approximations which ultimately lead to the significant acceleration of UQ tasks.

physics.data-an↗

arXiv2021

Tensor-train approximation of the chemical master equation and its application for parameter inference

In this work, we perform Bayesian inference tasks for the chemical master equation in the tensor-train format. The tensor-train approximation has been proven to be very efficient in representing high dimensional data arising from the explicit representation of the chemical master equation solution. An additional advantage of representing the probability mass function in the tensor train format is that parametric dependency can be easily incorporated by introducing a tensor product basis expansion in the parameter space. Time is treated as an additional dimension of the tensor and a linear system is derived to solve the chemical master equation in time. We exemplify the tensor-train method by performing inference tasks such as smoothing and parameter inference using the tensor-train framework. A very high compression ratio is observed for storing the probability mass function of the solution. Since all linear algebra operations are performed in the tensor-train format, a significant reduction of the computational time is observed as well.

stat.CO↗

arXiv2021

Local field reconstruction from rotating coil measurements in particle accelerator magnets

In this paper a general approach to reconstruct three dimensional field solutions in particle accelerator magnets from distributed magnetic measurements is presented. To exploit the locality of the measurement operation a special discretization of the Laplace equation is used. Extracting the coefficients of the field representations yields an inverse problem which is solved by Bayesian inversion. This allows not only to pave the way for uncertainty quantification, but also to derive a suitable regularization. The approach is applied to rotating coil measurements and can be extended to any other measurement procedure.

physics.acc-ph↗

arXiv2020

Data-Driven Solvers for Strongly Nonlinear Material Response

This work presents a data-driven magnetostatic finite-element solver that is specifically well-suited to cope with strongly nonlinear material responses. The data-driven computing framework is essentially a multiobjective optimization procedure matching the material operation points as closely as possible to given material data while obeying Maxwell's equations. Here, the framework is extended with heterogeneous (local) weighting factors - one per finite element - equilibrating the goal function locally according to the material behavior. This modification allows the data-driven solver to cope with unbalanced measurement data sets, i.e. data sets suffering from unbalanced space filling. This occurs particularly in the case of strongly nonlinear materials, which constitute problematic cases that hinder the efficiency and accuracy of standard data-driven solvers with a homogeneous (global) weighting factor. The local weighting factors are embedded in the distance-minimizing data-driven algorithm used for noiseless data, likewise for the maximum entropy data-driven algorithm used for noisy data. Numerical experiments based on a quadrupole magnet model with a soft magnetic material show that the proposed modification results in major improvements in terms of solution accuracy and solver efficiency. For the case of noiseless data, local weighting factors improve the convergence of the data-driven solver by orders of magnitude. When noisy data are considered, the convergence rate of the data-driven solver is doubled.

physics.comp-ph↗

arXiv2020

Magnetic Field Simulation with Data-Driven Material Modeling

This paper developes a data-driven magnetostatic finite-element (FE) solver which directly exploits measured material data instead of a material curve constructed from it. The distances between the field solution and the measurement points are minimized while enforcing Maxwell's equations. The minimization problem is solved by employing the Lagrange multiplier approach. The procedure wraps the FE method within an outer data-driven iteration. The method is capable of considering anisotropic materials and is adapted to deal with models featuring a combination of exact material knowledge and measured material data. Thereto, three approaches with an increasing level of intrusivity according to the FE formulation are proposed. The numerical results for a quadrupole-magnet model show that data-driven field simulation is feasible and affordable and overcomes the need of modeling the material law.

physics.comp-ph↗

arXiv2020

Enhanced adaptive surrogate models with applications in uncertainty quantification for nanoplasmonics

We propose an efficient surrogate modeling technique for uncertainty quantification. The method is based on a well-known dimension-adaptive collocation scheme. We improve the scheme by enhancing sparse polynomial surrogates with conformal maps and adjoint error correction. The methodology is applied to Maxwell's source problem with random input data. This setting comprises many applications of current interest from computational nanoplasmonics, such as grating couplers or optical waveguides. Using a non-trivial benchmark model we show the benefits and drawbacks of using enhanced surrogate models through various numerical studies. The proposed strategy allows us to conduct a thorough uncertainty analysis, taking into account a moderately large number of random parameters.

cs.CE↗

arXiv2020

Approximation and Uncertainty Quantification of Systems with Arbitrary Parameter Distributions using Weighted Leja Interpolation

Approximation and uncertainty quantification methods based on Lagrange interpolation are typically abandoned in cases where the probability distributions of one or more {system} parameters are not normal, uniform, or closely related {distributions}, due to the computational issues that arise when one wishes to define interpolation nodes for general distributions. This paper examines the use of the recently introduced weighted Leja nodes for that purpose. Weighted Leja interpolation rules are presented, along with a dimension-adaptive sparse interpolation algorithm, to be employed in the case of high-dimensional input uncertainty. The performance and reliability of the suggested approach is verified by four numerical experiments, where the respective models feature extreme value and truncated normal parameter distributions. Furthermore, the suggested approach is compared with a well-established polynomial chaos method and found to be either comparable or superior in terms of approximation and statistics estimation accuracy.

math.NA↗

arXiv2019

Robust Adaptive Least Squares Polynomial Chaos Expansions in High-Frequency Applications

We present an algorithm for computing sparse, least squares-based polynomial chaos expansions, incorporating both adaptive polynomial bases and sequential experimental designs. The algorithm is employed to approximate stochastic high-frequency electromagnetic models in a black-box way, in particular, given only a dataset of random parameter realizations and the corresponding observations regarding a quantity of interest, typically a scattering parameter. The construction of the polynomial basis is based on a greedy, adaptive, sensitivity-related method. The sequential expansion of the experimental design employs different optimality criteria, with respect to the algebraic form of the least squares problem. We investigate how different conditions affect the robustness of the derived surrogate models, that is, how much the approximation accuracy varies given different experimental designs. It is found that relatively optimistic criteria perform on average better than stricter ones, yielding superior approximation accuracies for equal dataset sizes. However, the results of strict criteria are significantly more robust, as reduced variations regarding the approximation accuracy are obtained, over a range of experimental designs. Two criteria are proposed for a good accuracy-robustness trade-off.

cs.CE↗

arXiv2019

Adaptive Sparse Polynomial Chaos Expansions via Leja Interpolation

This work suggests an interpolation-based stochastic collocation method for the non-intrusive and adaptive construction of sparse polynomial chaos expansions (PCEs). Unlike pseudo-spectral projection and regression-based stochastic collocation methods, the proposed approach results in PCEs featuring one polynomial term per collocation point. Moreover, the resulting PCEs are interpolating, i.e., they are exact on the interpolation nodes/collocation points. Once available, an interpolating PCE can be used as an inexpensive surrogate model, or be post-processed for the purposes of uncertainty quantification and sensitivity analysis. The main idea is conceptually simple and relies on the use of Leja sequence points as interpolation nodes. Using Newton-like, hierarchical basis polynomials defined upon Leja sequences, a sparse-grid interpolation can be derived, the basis polynomials of which are unique in terms of their multivariate degrees. A dimension-adaptive scheme can be employed for the construction of an anisotropic interpolation. Due to the degree uniqueness, a one-to-one transform to orthogonal polynomials of the exact same degrees is possible and shall result in an interpolating PCE. However, since each Leja node defines a unique Newton basis polynomial, an implicit one-to-one map between Leja nodes and orthogonal basis polynomials exists as well. Therefore, the in-between steps of hierarchical interpolation and basis transform can be discarded altogether, and the interpolating PCE can be computed directly. For directly computed, adaptive, anisotropic interpolating PCEs, the dimension-adaptive algorithm is modified accordingly. A series of numerical experiments verify the suggested approach in both low and moderately high-dimensional settings, as well as for various input distributions.

math.NA↗

arXiv2019

Influence of spatial dispersion on surface plasmons, nanoparticles and grating couplers

Recent experiments have shown that spatial dispersion may have a conspicuous impact on the response of plasmonic structures. This suggests that in some cases the Drude model should be replaced by more advanced descriptions that take spatial dispersion into account, like the hydrodynamic model. Here we show that nonlocality in the metallic response affects surface plasmons propagating at the interface between a metal and a dielectric with high permittivity. As a direct consequence, any nanoparticle with a radius larger than 20 nm can be expected to be sensitive to spatial dispersion whatever its size. The same behavior is expected for a simple metallic grating allowing the excitation of surface plasmons, just as in Wood's famous experiments. Importantly, our work suggests that for any plasmonic structure in a high permittivity dielectric, nonlocality should be taken into account.

physics.optics↗

arXiv2018

Robust Shape Optimization of Electric Devices Based on Deterministic Optimization Methods and Finite Element Analysis With Affine Decomposition and Design Elements

In this paper, gradient-based optimization methods are combined with finite-element modeling for improving electric devices. Geometric design parameters are considered by affine decomposition of the geometry or by the design element approach, both of which avoid remeshing. Furthermore, it is shown how to robustify the optimization procedure, i.e., how to deal with uncertainties on the design parameters. The overall procedure is illustrated by an academic example and by the example of a permanent-magnet synchronous machine. The examples show the advantages of deterministic optimization compared to standard and popular stochastic optimization procedures such as, e.g., particle swarm optimization.

math.OC↗