Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “model structural error”

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 181 records · Page 10
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Characterization of oceanic post-cold frontal clouds and their model representation. Final report

Low-level clouds are ubiquitous over the oceans, and their impact on the Earth’ radiative budget is important. However, general circulation models (GCMs) that are used to represent our climate and its evolution still experience problems to represent their correct amount and radiative impact. In the extratropical latitudes (~30-60 N/S), cloud properties are strongly modulated by the occurrence of anti-cyclones (high pressure systems) and cyclones (low pressure systems). While deep clouds dominate in cyclones near the center and along the cold and warm fronts, low-level clouds populate the colder regions of the cyclones, and more specifically the region in the wake of the cold fronts, the post-cold frontal region (PCF). The GCMs mentioned above were found to underestimate the amount of these PCF clouds, and this caused errors in their representation of the amount of solar radiation at the surface of the southern hemisphere oceans, causing an excess in absorption and errors in long-term predictions. To better understand which processes are ill-represented in the GCMs to cause such issues, we used ground-based observations from the Eastern North Atlantic (ENA) ARM site and a high resolution regional model (called WRF) to examine the properties of PCF clouds. With the WRF model, we simulated the passage of a cold front at the ENA site using theoretically distinct representations of convection and the physics of the boundary layer. While these physical representations had little impact on the timing, structure and circulation of the cold front, we found a high sensitivity to how the large scale information was fed into the model. Errors rapidly developed if the incoming flow information was injected too far from the site, but 1000 km was found to be an optimal distance between the outer boundary and the ENA site for a realistic cold front passage. With this model, we then conducted a series of experiment where convection and boundary layer schemes were changed, and revealed that 1) the clouds were more sensitive to the convection than boundary layer schemes and 2) the interaction between the two has a large impact. Furthermore, changing the representation of the physics in the model had greater impacts on PCF clouds than changing the initial conditions. As large-scale models increase their spatial resolution, convection will eventually be resolved but it remains that their representation of the physics of the boundary layer will matter for the correct representation of shallow (low-level) clouds. Observations were used to find which large-scale information had a clear relationship with PCF cloud and precipitation properties, in view of helping with GCM parameterizations and evaluation. Both lower troposphere stability and surface forcing were found to be important for PCF clouds that experience more vigorous dynamics than their tropical or subtropical counterparts. For cloud base and top heights, a simple measure of the change in potential temperature between the surface and the 800 hPa was found to be a good predictor. This relationship was found to hold as well when using data from campaigns recently conducted in the southern ocean south of Tasmania. A recent GCM was found to succeed in representing similar relations. For cloud depth, water content, and different characteristics of precipitation both the temperature contrast between the surface and near-air and the near-surface wind speed had a strong relationship. To summarize, through observational constraints and regional model experiments, the project has provided guidance to evaluate GCMs and tools to improve their representation of low-level clouds in the extratropics. It has also highlighted differences in behavior between these clouds and those typically found in the subtropics, thus opening new avenues of research for shallow clouds in general.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2014

A New Perspective for the Calibration of Computational Predictor Models

This paper presents a framework for calibrating computational models using data from several and possibly dissimilar validation experiments. The offset between model predictions and observations, which might be caused by measurement noise, model-form uncertainty, and numerical error, drives the process by which uncertainty in the models parameters is characterized. The resulting description of uncertainty along with the computational model constitute a predictor model. Two types of predictor models are studied: Interval Predictor Models (IPMs) and Random Predictor Models (RPMs). IPMs use sets to characterize uncertainty, whereas RPMs use random vectors. The propagation of a set through a model makes the response an interval valued function of the state, whereas the propagation of a random vector yields a random process. Optimization-based strategies for calculating both types of predictor models are proposed. Whereas the formulations used to calculate IPMs target solutions leading to the interval value function of minimal spread containing all observations, those for RPMs seek to maximize the models’ ability to reproduce the distribution of observations. Regarding RPMs, we choose a structure for the random vector (i.e., the assignment of probability to points in the parameter space) solely dependent on the prediction error. As such, the probabilistic description of uncertainty is not a subjective assignment of belief, nor is it expected to asymptotically converge to a fixed value, but instead it is a description of the model’s ability to reproduce the experimental data. This framework enables evaluating the spread and distribution of the predicted response of target applications depending on the same parameters beyond the validation domain (i.e., roll-up and extrapolation).

42 ENGINEERING↗
DOE OSTI2014

TU-C-17A-04: BEST IN PHYSICS (THERAPY) - A Supervised Framework for Automatic Contour Assessment for Radiotherapy Planning of Head- Neck Cancer

Purpose: Precise contour delineation of tumor targets and critical structures from CT simulations is essential for accurate radiotherapy (RT) treatment planning. However, manual and automatic delineation processes can be error prone due to limitations in imaging techniques and individual anatomic variability. Tedious and laborious manual verification is hence needed. This study develops a general framework for automatically assessing RT contours for head-neck cancer patients using geometric attribute distribution models (GADMs). Methods: Geometric attributes (centroid and volume) were computed from physician-approved RT contours of 29 head-neck patients. Considering anatomical correlation between neighboring structures, the GADM for each attribute was trained to characterize intra- and interpatient structure variations using principal component analysis. Each trained GADM was scalable and deformable, but constrained by the principal attribute variations of the training contours. A new hierarchical model adaptation algorithm was utilized to assess the RT contour correctness for a given patient. Receiver operating characteristic (ROC) curves were employed to evaluate and tune system parameters for the training models. Results: Experiments utilizing training and non-training data sets with simulated contouring errors were conducted to validate the framework performance. Promising assessment results of contour normality/abnormality for the training contour-based data were achieved with excellent accuracy (0.99), precision (0.99), recall (0.83), and F-score (0.97), while corresponding values of 0.84, 0.96, 0.83, and 0.9 were achieved for the non-training data. Furthermore, the areas under the ROC curves were above 0.9, validating the accuracy of this test. Conclusion: The proposed framework can reliably identify contour normality/abnormality based upon intra- and inter-structure constraints derived from clinically-approved contours. It also allows physicians to analytically determine the system parameters to fit various clinic requirements (e.g. as-low-as-possible false positives). It has great potential for improving RT work flow. More geometric attributes and training sets will be investigated to improve framework performance in the future.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2019

SAD phasing of XFEL data depends critically on the error model

A nonlinear least-squares method for refining a parametric expression describing the estimated errors of reflection intensities in serial crystallographic (SX) data is presented. This approach, which is similar to that used in the rotation method of crystallographic data collection at synchrotrons, propagates error estimates from photon-counting statistics to the merged data. Here, it is demonstrated that the application of this approach to SX data provides better SAD phasing ability, enabling the autobuilding of a protein structure that had previously failed to be built. Estimating the error in the merged reflection intensities requires the understanding and propagation of all of the sources of error arising from the measurements. One type of error, which is well understood, is the counting error introduced when the detector counts X-ray photons. Thus, if other types of random errors (such as readout noise) as well as uncertainties in systematic corrections (such as from X-ray attenuation) are completely understood, they can be propagated along with the counting error, as appropriate. In practice, most software packages propagate as much error as they know how to model and then include error-adjustment terms that scale the error estimates until they explain the variance among the measurements. If this is performed carefully, then during SAD phasing likelihood-based approaches can make optimal use of these error estimates, increasing the chance of a successful structure solution. In serial crystallography, SAD phasing has remained challenging, with the few examples of de novo protein structure solution each requiring many thousands of diffraction patterns. Here, the effects of different methods of treating the error estimates are estimated and it is shown that using a parametric approach that includes terms proportional to the known experimental uncertainty, the reflection intensity and the squared reflection intensity to improve the error estimates can allow SAD phasing even from weak zinc anomalous signal.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2014

A New Perspective for the Calibration of Computational Predictor Models.

This paper presents a framework for calibrating computational models using data from sev- eral and possibly dissimilar validation experiments. The offset between model predictions and observations, which might be caused by measurement noise, model-form uncertainty, and numerical error, drives the process by which uncertainty in the models parameters is characterized. The resulting description of uncertainty along with the computational model constitute a predictor model. Two types of predictor models are studied: Interval Predictor Models (IPMs) and Random Predictor Models (RPMs). IPMs use sets to characterize uncer- tainty, whereas RPMs use random vectors. The propagation of a set through a model makes the response an interval valued function of the state, whereas the propagation of a random vector yields a random process. Optimization-based strategies for calculating both types of predictor models are proposed. Whereas the formulations used to calculate IPMs target solutions leading to the interval value function of minimal spread containing all observations, those for RPMs seek to maximize the models' ability to reproduce the distribution of obser- vations. Regarding RPMs, we choose a structure for the random vector (i.e., the assignment of probability to points in the parameter space) solely dependent on the prediction error. As such, the probabilistic description of uncertainty is not a subjective assignment of belief, nor is it expected to asymptotically converge to a fixed value, but instead it is a description of the model's ability to reproduce the experimental data. This framework enables evaluating the spread and distribution of the predicted response of target applications depending on the same parameters beyond the validation domain (i.e., roll-up and extrapolation).

Crespo, Luis Guillermo↗
DOE OSTI2024

Reduced‐Order Modeling for Linearized Representations of Microphysical Process Rates

Abstract Representing cloud microphysical processes in large scale atmospheric models is challenging because many processes depend on the details of the droplet size distribution (DSD, the spectrum of droplets with different sizes in a cloud). While full or partial statistical moments of droplet size distributions are the typical variables used in bulk models, prognostic moments are limited in their ability to represent microphysical processes across the range of conditions experienced in the atmosphere. Microphysical parameterizations employing prognostic moments are known to suffer from structural uncertainty in their representations of inherently higher dimensional cloud processes, which limit model fidelity and lead to forecasting errors. Here we investigate how data‐driven reduced‐order modeling can be used to learn predictors for microphysical process rates in bulk microphysics schemes in an unsupervised manner from higher dimensional bin distributions. Using simulations characteristic of marine stratiform clouds, we simultaneously learn lower dimensional representations of droplet size distributions and predict the evolution of the microphysical state of the system. Droplet collision‐coalescence, the main process for generating warm rain, is estimated to have an intrinsic dimension of three. This intrinsic dimension provides a lower limit on the number of degrees of freedom needed to accurately represent collision‐coalescence in models. We demonstrate how deep learning based reduced‐order modeling can be used to discover intrinsic coordinates describing the microphysical state of the system, where process rates such as collision‐coalescence are globally linearized. These implicitly learned representations of the DSD retain more information about the DSD than typical moment‐based representations.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2026

Modeling the behavior of concentrated aqueous HNO 3 using machine learning interatomic potentials

We develop two multi-defect machine learning interatomic potentials (MLIPs) trained at the BLYP-D2 and PBE-D3 density functional theories using the DeepMD-kit, allowing for the investigation of structural and thermodynamic properties of nitric acid over a wide range of concentrations via molecular dynamics (MD) simulations. We directly compute the degree of dissociation, α, and pK a from MD simulations, revealing that HNO 3 behaves as a weaker acid at higher concentrations, noting that our standard-state pK a value is in excellent agreement with the experimental one. In general, good agreement is observed with experimental results such as α and density outside the training dataset, with only modest deviations at low-to-medium concentrations. We benchmark our custom multi-defect DeepMD MLIPs against foundational models MACE-MP0 and MACE-OFF23. The foundation models capture some aspects of HNO 3 /NO 3 − solvation in concentrated nitric acid but show noticeable density errors and miss subtle structural features relevant to spectroscopy, whereas the bespoke DeepMD MLIPs yield more compact solvation shells, reproduce density-concentration trends, and run ∼12–15× faster than MACE-MP0. Although classical FFs are still more efficient and match experimental densities better, they lack chemical reactivity and thus cannot predict α or pK a , underscoring the need for system-specific reactive MLIPs beyond universal MLIPs.

Dinpajooh, Mohammadhasan [Pacific Northwest Nation↗
DOE OSTI2021

Structure Factors of Neutron Matter at Finite Temperature

Here, we compute continuum and infinite volume limit extrapolations of the structure factors of neutron matter at finite temperature and density. Using a lattice formulation of leading-order pionless effective field theory, we compute the momentum dependence of the structure factors at finite temperature and at densities beyond the reach of the virial expansion. The Tan contact parameter is computed and the result agrees with the high momentum tail of the vector structure factor. All errors, statistical and systematic, are controlled for. This calculation is a first step towards a model-independent understanding of the linear response of neutron matter at finite temperature.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
DOE OSTI2013

Feedback-Based, System-Level Properties of Vertebrate-Microbial Interactions

Improved characterization of infectious disease dynamics is required. To that end, three-dimensional (3D) data analysis of feedback-like processes may be considered. To detect infectious disease data patterns, a systems biology (SB) and evolutionary biology (EB) approach was evaluated, which utilizes leukocyte data structures designed to diminish data variability and enhance discrimination. Using data collected from one avian and two mammalian (human and bovine) species infected with viral, parasite, or bacterial agents (both sensitive and resistant to antimicrobials), four data structures were explored: (i) counts or percentages of a single leukocyte type, such as lymphocytes, neutrophils, or macrophages (the classic approach), and three levels of the SB/ EB approach, which assessed (ii) 2D, (iii) 3D, and (iv) multi-dimensional (rotating 3D) host-microbial interactions. In all studies, no classic data structure discriminated disease-positive (D+, or observations in which a microbe was isolated) from disease-negative (D–, or microbial-negative) groups: D+ and D– data distributions overlapped. In contrast, multi-dimensional analysis of indicators designed to possess desirable features, such as a single line of observations, displayed a continuous, circular data structure, whose abrupt inflections facilitated partitioning into subsets statistically significantly different from one another. In all studies, the 3D, SB/EB approach distinguished three (steady, positive, and negative) feedback phases, in which D– data characterized the steady state phase, and D+ data were found in the positive and negative phases. In humans, spatial patterns revealed false-negative observations and three malaria-positive data classes. In both humans and bovines, methicillin-resistant Staphylococcus aureus (MRSA) infections were discriminated from non-MRSA infections. More information can be extracted, from the same data, provided that data are structured, their 3D relationships are considered, and well-conserved (feedback-like) functions are estimated. Patterns emerging from such structures may distinguish well-conserved from recently developed host-microbial interactions. Applications include diagnosis, error detection, and modeling.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Experimental and theoretical study of energy transfer in a chromophore triad: What makes modeling dynamics successful?

Simulation of electronic dynamics in realistically large molecular systems is a demanding task that has not yet achieved the same level of quantitative prediction already realized for its static counterpart. This is particularly true for processes occurring beyond the Born–Oppenheimer regime. Non-adiabatic molecular dynamics (NAMD) simulations suffer from two convoluted sources of error: numerical algorithms for dynamics and electronic structure calculations. While the former has gained increasing attention, particularly addressing the validity of ad hoc methodologies, the effect of the latter remains relatively unexplored. Indeed, the required accuracy for electronic structure calculations to reach quantitative agreement with experiment in dynamics may be even more strict than that required for static simulations. In this study, we address this issue by modeling the electronic energy transfer in a donor–acceptor–donor (D–A–D) molecular light harvesting system using fewest switches surface hopping NAMD simulations. In the studied system, time-resolved experimental measurements deliver complete information on spectra and energy transfer rates. Subsequent modeling shows that the calculated electronic transition energies are “sufficiently good” to reproduce experimental spectra but produce over an order of magnitude error in simulated dynamical rates. We further perform simulations using artificially shifted energy gaps to investigate the complex relationship between transition energies and modeled dynamics to understand factors affecting non-radiative relaxation and energy transfer rates.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Quantification and prediction of error in MIMO tests

Dynamic testing of systems is most realistic of real-world conditions when multiple input and multiple output (MIMO) techniques are used. To replicate measured environmental conditions, a series of desired outputs (responses) on the system must be realized by inverting the frequency response functions (FRF) matrix for input estimation. System identification is dependent on the number of inputs and outputs, which affect the type of solutions. In input estimation, the auto spectra of the determined inputs are affected by both the auto spectra and cross spectra of the desired outputs. This research evaluates errors and investigates two sources in an experimental setting. The first significant error source is test-to-test variability in FRFs from system identification tests, quantified by magnitude variability from long-duration tests. The second significant error source is the realization of time histories from frequency domain inputs. The spectral content of the realized random process deviates from the desired content and is also quantified. The structure is a linear three-story frame fixed at the base; two inputs are provided uniaxially with the use of suspended electrodynamic shakers. The proposed model for error sources was found to be effective in predicting errors in a SISO (single-input single-output) and a square (2-input 2-output) MIMO test.

Yousef, Odey↗
DOE OSTI2017

Predicting Patient-specific Dosimetric Benefits of Proton Therapy for Skull-base Tumors Using a Geometric Knowledge-based Method

Purpose: To predict the organ at risk (OAR) dose levels achievable with proton beam therapy (PBT), solely based on the geometric arrangement of the target volume in relation to the OARs. A comparison with an alternative therapy yields a prediction of the patient-specific benefits offered by PBT. This could enable physicians at hospitals without proton capabilities to make a better-informed referral decision or aid patient selection in model-based clinical trials. Methods and Materials: Skull-base tumors were chosen to test the method, owing to their geometric complexity and multitude of nearby OARs. By exploiting the correlations between the dose and distance-to-target in existing PBT plans, the models were independently trained for 6 types of OARs: brainstem, cochlea, optic chiasm, optic nerve, parotid gland, and spinal cord. Once trained, the models could estimate the feasible dose–volume histogram and generalized equivalent uniform dose (gEUD) for OAR structures of new patients. The models were trained using 20 patients and validated using an additional 21 patients. Validation was achieved by comparing the predicted gEUD to that of the actual PBT plan. Results: The predicted and planned gEUD were in good agreement. Considering all OARs, the prediction error was +1.4 ± 5.1 Gy (mean ± standard deviation), and Pearson's correlation coefficient was 93%. By comparing with an intensity modulated photon treatment plan, the model could classify whether an OAR structure would experience a gain, with a sensitivity of 93% (95% confidence interval: 87%-97%) and specificity of 63% (95% confidence interval: 38%-84%). Conclusions: We trained and validated models that could quickly and accurately predict the patient-specific benefits of PBT for skull-base tumors. Similar models could be developed for other tumor sites. Such models will be useful when an estimation of the feasible benefits of PBT is desired but the experience and/or resources required for treatment planning are unavailable.

62 RADIOLOGY AND NUCLEAR MEDICINE↗
DOE OSTI2015

The Schottky barrier modulation at PtSi/Si interface by strain and structural deformation

We show, using density functional theory (DFT) calculations, that the Schottky barrier height (SBH) at the PtSi/Si interface can be lowered by uniaxial strain applied not only on Si but also on PtSi. The strain was applied to the (001) direction of Si and PtSi, which is normal for the interface. The SBH of the hole is lowered by 0.08 eV under 2% of tensile strain on Si and by 0.09 eV under 4 % of compressive strain on PtSi. Because the SBH at PtSi/Si contact is approximately 0.2 eV, this amount of reduction can significantly lower the resistance of the PtSi/Si contact; thus applying uniaxial strain on both PtSi and Si possibly enhances the performance of Schottky barrier field effect transistors. Theoretical models of SB formation and conventional structure model are evaluated. It is found that Pt penetration into Si stabilizes the interface and lowers the SBH by approximately 0.1 eV from the bulk-terminated interface model, which implies that conventionally used bulk-terminated interface models have significant errors. Among the theoretical models of SB formation, the model of strong Fermi level pining adequately explains the electron transfer phenomena and SBH, but it has limited ability to explain SBH changes induced by changes of interface structure.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Improved Treatment of 1–4 Interactions in Force Fields for Molecular Dynamics Simulations

Traditional force fields commonly use a combination of bonded torsional terms and empirically scaled nonbonded interactions to capture 1-4 energies and forces of atoms separated by three bonds in a molecule. While this approach can yield accurate torsional energy barriers, it often leads to inaccurate forces and erroneous geometries and creates an interdependence between dihedral terms and nonbonded interactions, complicating parametrization and reducing transferability. Here, in this paper, we demonstrate that 1-4 interactions can be accurately modeled using only bonded coupling terms, eliminating the need for arbitrarily scaled nonbonded interactions altogether. Furthermore, by leveraging the automated parametrization capabilities of the Q-Force toolkit, we efficiently determine the necessary coupling terms without the need for manual adjustment. Our approach is first validated on a range of small molecule systems, encompassing both flexible and rigid structures, and shows a significant improvement in force field accuracy, obtaining subkcal/mol mean absolute error for every molecule tested. We further extend the bonded-only model for 1-4 interactions to Amber ff14sb, CHARMM36, and OPLS-AA force fields to reproduce ab initio gas and implicit solvent ϕ,ψ surfaces of alanine dipeptide.

Abdullah, Aalim S. [University of California, Berk↗
DOE OSTI2022

Stress-constrained topology optimization of lattice-like structures using component-wise reduced order models

We report lattice-like structures can provide a combination of high stiffness with light weight that is useful in many applications, but a resolved finite element mesh of such structures results in a computationally expensive discretization. This computational expense may be particularly burdensome in many-query applications, such as optimization. We develop a stress-constrained topology optimization method for lattice-like structures that uses component-wise reduced order models as a cheap surrogate, providing accurate computation of stress fields while greatly reducing run time relative to a full order model. We demonstrate the ability of our method to produce large reductions in mass while respecting a constraint on the maximum stress in a pair of test problems. The ROM methodology provides a speedup of about 150x in forward solves compared to full order static condensation and provides a relative error of less than 5% in the relaxed stress.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2015

Anomalous small-angle X-ray scattering of nanoporous two-phase atomistic models for amorphous silicon–germanium alloys

The present work deals with a detailed analysis of the anomalous small-angle X-ray scattering in amorphous silicon–germanium alloy using the simulation technique. We envisage the nanoporous two-phase alloy model consisting in a mixture of Ge-rich and Ge-poor domains and voids at the nanoscale. By substituting Ge atoms for Si atoms in nanoporous amorphous silicon network, compositionally heterogeneous alloys are generated with various composition-contrasts between the two phases. After relaxing the as-generated structure, we compute its radial distribution function, and then we deduce by the Fourier transform technique its anomalous X-ray scattering pattern. Using a smoothing procedure, the computed X-ray scattering patterns are corrected for the termination errors due to the finite size of the model, allowing so a rigorous quantitative analysis of the anomalous small-angle scattering. Our simulation shows that, as expected, the anomalous small-angle X-ray scattering technique is a tool of choice for characterizing compositional heterogeneities coexisting with structural inhomogeneities in an amorphous alloy. Furthermore, the sizes of the compositional nanoheterogeneities, as measured by anomalous small-angle X-ray scattering technique, are X-ray energy independent. A quantitative analysis of the separated reduced anomalous small-angle X-ray scattering, as defined in this work, provided a good estimate of their size.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Accurate Prediction of HSE06 Band Structures for a Diverse Set of Materials Using Δ-Learning

Here we used machine learning (ML) to accurately predict eigenvalues of the hybrid HSE06 functional using eigenvalues computed by the less computationally expensive PBE functional and associated electronic features based on the k-point resolved atomic band character. The ML model was trained by using eigenvalues from only one k-point for each of the 168 compounds in the training set. The HSE06 eigenvalues across all k-points were then predicted for a separate set of 169 compounds with a mean absolute error (MAE) of 0.13 eV, representing a significant improvement over the error of PBE-computed eigenvalues relative to that of HSE06 (MAE = 0.96 eV). These accurately predicted eigenvalues result in remarkably accurate predictions for the band structures, projected density of states, and band gaps, even though the model was not explicitly trained on these other properties. Finally, we demonstrate that our ML model has a similar accuracy for both ternary and quaternary compounds well outside the initial training set and on systems with 112 and 160 atoms, demonstrating its potential to rapidly predict HSE06-quality electronic structures of complex materials that are practically unfeasible for HSE06.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

A Fast Time-Stepping Strategy for Dynamical Systems Equipped with a Surrogate Model

Simulation of complex dynamical systems arising in many applications is computationally challenging due to their size and complexity. Model order reduction, machine learning, and other types of surrogate modeling techniques offer cheaper and simpler ways to describe the dynamics of these systems but are inexact and introduce additional approximation errors. In order to overcome the computational difficulties of the full complex models, on one hand, and the limitations of surrogate models, on the other, this work proposes a new accelerated time-stepping strategy that combines information from both. This approach is based on the multirate infinitesimal general-structure additive Runge--Kutta framework. The inexpensive surrogate model is integrated with a small time step to guide the solution trajectory, and the full model is treated with a large time step to occasionally correct for the surrogate model error and ensure convergence. Here, we provide a theoretical error analysis, and several numerical experiments, to show that this approach can be significantly more efficient than using only the full or only the surrogate model for the integration.

Surrogate models↗