scholarly journals Nonlinear information fusion algorithms for data-efficient multi-fidelity modelling

2017 ◽  
Vol 473 (2198) ◽  
pp. 20160751 ◽  
Author(s):  
P. Perdikaris ◽  
M. Raissi ◽  
A. Damianou ◽  
N. D. Lawrence ◽  
G. E. Karniadakis
Keyword(s):  
Information Fusion ◽  
Low Cost ◽  
Computational Cost ◽  
Practical Interest ◽  
Gaussian Process Regression ◽  
Benchmark Problems ◽  
High Fidelity ◽  
Small Set ◽  
Variable Fidelity ◽  
Dynamics Simulations

Multi-fidelity modelling enables accurate inference of quantities of interest by synergistically combining realizations of low-cost/low-fidelity models with a small set of high-fidelity observations. This is particularly effective when the low- and high-fidelity models exhibit strong correlations, and can lead to significant computational gains over approaches that solely rely on high-fidelity models. However, in many cases of practical interest, low-fidelity models can only be well correlated to their high-fidelity counterparts for a specific range of input parameters, and potentially return wrong trends and erroneous predictions if probed outside of their validity regime. Here we put forth a probabilistic framework based on Gaussian process regression and nonlinear autoregressive schemes that is capable of learning complex nonlinear and space-dependent cross-correlations between models of variable fidelity, and can effectively safeguard against low-fidelity models that provide wrong trends. This introduces a new class of multi-fidelity information fusion algorithms that provide a fundamental extension to the existing linear autoregressive methodologies, while still maintaining the same algorithmic complexity and overall computational cost. The performance of the proposed methods is tested in several benchmark problems involving both synthetic and real multi-fidelity datasets from computational fluid dynamics simulations.

Inverse surrogate modeling for low-cost geometry scaling of microwave and antenna structures

2016 ◽  
Vol 33 (4) ◽  
pp. 1095-1113 ◽  
Author(s):  
Slawomir Koziel ◽  
Adrian Bekasiewicz
Keyword(s):  
Low Cost ◽  
Computational Cost ◽  
Surrogate Modeling ◽  
High Fidelity ◽  
Modeling Framework ◽  
Simulation Accuracy ◽  
Content Type ◽  
Scaling Process ◽  
Variable Fidelity ◽  
Em Simulation

Purpose – The purpose of this paper is to investigate strategies for expedited dimension scaling of electromagnetic (EM)-simulated microwave and antenna structures, exploiting the concept of variable-fidelity inverse surrogate modeling. Design/methodology/approach – A fast inverse surrogate modeling technique is described for dimension scaling of microwave and antenna structures. The model is established using reference designs obtained for cheap underlying low-fidelity model and corrected to allow structure scaling at high accuracy level. Numerical and experimental case studies are provided demonstrating feasibility of the proposed approach. Findings – It is possible, by appropriate combination of surrogate modeling techniques, to establish an inverse model for explicit determination of geometry dimensions of the structure at hand so as to re-design it for various operating frequencies. The scaling process can be concluded at a low computational cost corresponding to just a few evaluations of the high-fidelity computational model of the structure. Research limitations/implications – The present study is a step toward development of procedures for rapid dimension scaling of microwave and antenna structures at high-fidelity EM-simulation accuracy. Originality/value – The proposed modeling framework proved useful for fast geometry scaling of microwave and antenna structures, which is very laborious when using conventional methods. To the authors’ knowledge, this is one of the first attempts to surrogate-assisted dimension scaling of microwave components at the EM-simulation level.

scholarly journals Multi-Fidelity Optimization of a Composite Airliner Wing Subject to Structural and Aeroelastic Constraints

Aerospace ◽  
2021 ◽  
Vol 8 (12) ◽  
pp. 398
Author(s):  
Angelos Kafkas ◽  
Spyridon Kilimtzidis ◽  
Athanasios Kotzakolios ◽  
Vassilis Kostopoulos ◽  
George Lampeas
Keyword(s):  
Low Cost ◽  
Computational Cost ◽  
Design Stage ◽  
Computational Time ◽  
High Fidelity ◽  
Full Potential ◽  
Optimized Design ◽  
Structural Representation ◽  
Optimization Framework ◽  
Variable Fidelity

Efficient optimization is a prerequisite to realize the full potential of an aeronautical structure. The success of an optimization framework is predominately influenced by the ability to capture all relevant physics. Furthermore, high computational efficiency allows a greater number of runs during the design optimization process to support decision-making. The efficiency can be improved by the selection of highly optimized algorithms and by reducing the dimensionality of the optimization problem by formulating it using a finite number of significant parameters. A plethora of variable-fidelity tools, dictated by each design stage, are commonly used, ranging from costly high-fidelity to low-cost, low-fidelity methods. Unfortunately, despite rapid solution times, an optimization framework utilizing low-fidelity tools does not necessarily capture the physical problem accurately. At the same time, high-fidelity solution methods incur a very high computational cost. Aiming to bridge the gap and combine the best of both worlds, a multi-fidelity optimization framework was constructed in this research paper. In our approach, the low-fidelity modules and especially the equivalent-plate methodology structural representation, capable of drastically reducing the associated computational time, form the backbone of the optimization framework and a MIDACO optimizer is tasked with providing an initial optimized design. The higher fidelity modules are then employed to explore possible further gains in performance. The developed framework was applied to a benchmark airliner wing. As demonstrated, reasonable mass reduction was obtained for a current state of the art configuration.

Variable Fidelity Modeling in Closed Loop Dynamical Systems

2014 ◽  
Author(s):  
Matthew A. Williams ◽  
Andrew G. Alleyne
Keyword(s):  
Dynamical Systems ◽  
Closed Loop ◽  
System Development ◽  
Computational Cost ◽  
Computational Effort ◽  
High Fidelity ◽  
Vapor Compression System ◽  
Variable Fidelity ◽  
Time Required ◽  
Fidelity Model

In the early stages of control system development, designers often require multiple iterations for purposes of validating control designs in simulation. This has the potential to make high fidelity models undesirable due to increased computational complexity and time required for simulation. As a solution, lower fidelity or simplified models are used for initial designs before controllers are tested on higher fidelity models. In the event that unmodeled dynamics cause the controller to fail when applied on a higher fidelity model, an iterative approach involving designing and validating a controller’s performance may be required. In this paper, a switched-fidelity modeling formulation for closed loop dynamical systems is proposed to reduce computational effort while maintaining elevated accuracy levels of system outputs and control inputs. The effects on computational effort and accuracy are investigated by applying the formulation to a traditional vapor compression system with high and low fidelity models of the evaporator and condenser. This sample case showed the ability of the switched fidelity framework to closely match the outputs and inputs of the high fidelity model while decreasing computational cost by 32% from the high fidelity model. For contrast, the low fidelity model decreases computational cost by 48% relative to the high fidelity model.

Variable-Fidelity Multidisciplinary Design Optimization Based on Analytical Target Cascading Framework

2012 ◽  
Vol 544 ◽  
pp. 49-54 ◽  
Author(s):  
Jun Zheng ◽  
Hao Bo Qiu ◽  
Xiao Lin Zhang
Keyword(s):  
Multidisciplinary Design Optimization ◽  
Model Building ◽  
Computational Cost ◽  
Simulation Models ◽  
Multidisciplinary Design ◽  
High Fidelity ◽  
Analytical Target Cascading ◽  
Approximation Model ◽  
Target Cascading ◽  
Variable Fidelity

ATC provides a systematic approach in solving decomposed large scale systems that has solvable subsystems. However， complex engineering system usually has a high computational cost , which result in limiting real-life applications of ATC based on high-fidelity simulation models. To address these problems, this paper aims to develop an efficient approximation model building techniques under the analytical target cascading (ATC) framework, to reduce computational cost associated with multidisciplinary design optimization problems based on high-fidelity simulations. An approximation model building techniques is proposed: approximations in the subsystem level are based on variable-fidelity modeling (interaction of low- and high-fidelity models). The variable-fidelity modeling consists of computationally efficient simplified models (low-fidelity) and expensive detailed (high-fidelity) models. The effectiveness of the method for modeling under the ATC framework using variable-fidelity models is studied. Overall results show the methods introduced in this paper provide an effective way of improving computational efficiency of the ATC method based on variable-fidelity simulation models.

scholarly journals Linking Gaussian process regression with data-driven manifold embeddings for nonlinear data fusion

2019 ◽  
Vol 9 (3) ◽  
pp. 20180083 ◽  
Author(s):  
Seungjoon Lee ◽  
Felix Dietrich ◽  
George E. Karniadakis ◽  
Ioannis G. Kevrekidis
Keyword(s):  
Gaussian Process ◽  
Information Fusion ◽  
Gaussian Process Regression ◽  
Statistical Modelling ◽  
Training Data ◽  
High Fidelity ◽  
Several Variables ◽  
Mathematical Algorithms ◽  
Data Points ◽  
Fidelity Model

In statistical modelling with Gaussian process regression, it has been shown that combining (few) high-fidelity data with (many) low-fidelity data can enhance prediction accuracy, compared to prediction based on the few high-fidelity data only. Such information fusion techniques for multi-fidelity data commonly approach the high-fidelity model f h ( t ) as a function of two variables ( t , s ), and then use f l ( t ) as the s data. More generally, the high-fidelity model can be written as a function of several variables ( t , s 1 , s 2 ….); the low-fidelity model f l and, say, some of its derivatives can then be substituted for these variables. In this paper, we will explore mathematical algorithms for multi-fidelity information fusion that use such an approach towards improving the representation of the high-fidelity function with only a few training data points. Given that f h may not be a simple function—and sometimes not even a function—of f l , we demonstrate that using additional functions of t , such as derivatives or shifts of f l , can drastically improve the approximation of f h through Gaussian processes. We also point out a connection with ‘embedology’ techniques from topology and dynamical systems. Our illustrative examples range from instructive caricatures to computational biology models, such as Hodgkin–Huxley neural oscillations.

Fast multi-objective design optimization of microwave and antenna structures using data-driven surrogates and domain segmentation

2019 ◽  
Vol 37 (2) ◽  
pp. 753-788
Author(s):  
Slawomir Koziel ◽  
Adrian Bekasiewicz
Keyword(s):  
Design Optimization ◽  
Design Problem ◽  
Low Cost ◽  
Computational Cost ◽  
Search Space ◽  
Population Based ◽  
Design Framework ◽  
Content Type ◽  
Multi Objective ◽  
Variable Fidelity

Purpose The purpose of this paper is to investigate the strategies and algorithms for expedited design optimization of microwave and antenna structures in multi-objective setup. Design/methodology/approach Formulation of the multi-objective design problem-oriented toward execution of the population-based metaheuristic algorithm within the segmented search space is investigated. Described algorithmic framework exploits variable fidelity modeling, physics- and approximation-based representation of the structure and model correction techniques. The considered approach is suitable for handling various problems pertinent to the design of microwave and antenna structures. Numerical case studies are provided demonstrating the feasibility of the segmentation-based framework for the design of real-world structures in setups with two and three objectives. Findings Formulation of appropriate design problem enables identification of the search space region containing Pareto front, which can be further divided into a set of compartments characterized by small combined volume. Approximation model of each segment can be constructed using a small number of training samples and then optimized, at a negligible computational cost, using population-based metaheuristics. Introduction of segmentation mechanism to multi-objective design framework is important to facilitate low-cost optimization of many-parameter structures represented by numerically expensive computational models. Further reduction of the design cost can be achieved by enforcing equal-volumes of the search space segments. Research limitations/implications The study summarizes recent advances in low-cost multi-objective design of microwave and antenna structures. The investigated techniques exceed capabilities of conventional design approaches involving direct evaluation of physics-based models for determination of trade-offs between the design objectives, particularly in terms of reliability and reduction of the computational cost. Studies on the scalability of segmentation mechanism indicate that computational benefits of the approach decrease with the number of search space segments. Originality/value The proposed design framework proved useful for the rapid multi-objective design of microwave and antenna structures characterized by complex and multi-parameter topologies, which is extremely challenging when using conventional methods driven by population-based metaheuristics algorithms. To the authors knowledge, this is the first work that summarizes segmentation-based approaches to multi-objective optimization of microwave and antenna components.

Fast tolerance-aware design optimization of miniaturized microstrip couplers using variable-fidelity EM simulations and response features

2019 ◽  
Vol 36 (9) ◽  
pp. 2983-2995
Author(s):  
Slawomir Koziel ◽  
Adrian Bekasiewicz
Keyword(s):  
Monte Carlo ◽  
Statistical Analysis ◽  
Low Cost ◽  
Computational Cost ◽  
Data Driven ◽  
Yield Optimization ◽  
Content Type ◽  
Frequency Responses ◽  
Training Samples ◽  
Variable Fidelity

Purpose This paper aims to investigate the strategy for low-cost yield optimization of miniaturized microstrip couplers using variable-fidelity electromagnetic (EM) simulations. Design/methodology/approach Usefulness of data-driven models constructed from structure frequency responses formulated in the form of suitably defined characteristic points for statistical analysis is investigated. Reformulation of the characteristics leads to a less nonlinear functional landscape and reduces the number of training samples required for accurate modeling. Further reduction of the cost associated with construction of the data-driven model, is achieved using variable-fidelity methods. Numerical case study is provided demonstrating feasibility of the feature-based modeling for low cost statistical analysis and yield optimization. Findings It is possible, through reformulation of the structure frequency responses in the form of suitably defined feature points, to reduce the number of training samples required for its data-driven modeling. The approximation model can be used as an accurate evaluation engine for a low-cost Monte Carlo analysis. Yield optimization can be realized through minimization of yield within the data-driven model bounds and subsequent model re-set around the optimized design. Research limitations/implications The investigated technique exceeds capabilities of conventional Monte Carlo-based approaches for statistical analysis in terms of computational cost without compromising its accuracy with respect to the conventional EM-based Monte Carlo. Originality/value The proposed tolerance-aware design approach proved useful for rapid yield optimization of compact microstrip couplers represented using EM-simulation models, which is extremely challenging when using conventional approaches due to tremendous number of EM evaluations required for statistical analysis.

Parameterization of Divalent Cations for Coarse-Grained Simulations

2020 ◽  
Author(s):  
Florencia Klein ◽  
Daniela Cáceres-Rojas ◽  
Monica Carrasco ◽  
Juan Carlos Tapia ◽  
Julio Caballero ◽  
...  
Keyword(s):  
Molecular Dynamics ◽  
Metal Ions ◽  
Molecular Dynamics Simulations ◽  
Divalent Cations ◽  
Computational Cost ◽  
Data Bank ◽  
Coarse Grained ◽  
Interaction Parameters ◽  
Dynamics Simulations ◽  
Dynamical Description

<p>Although molecular dynamics simulations allow for the study of interactions among virtually all biomolecular entities, metal ions still pose significant challenges to achieve an accurate structural and dynamical description of many biological assemblies. This is particularly the case for coarse-grained (CG) models. Although the reduced computational cost of CG methods often makes them the technique of choice for the study of large biomolecular systems, the parameterization of metal ions is still very crude or simply not available for the vast majority of CG- force fields. Here, we show that incorporating statistical data retrieved from the Protein Data Bank (PDB) to set specific Lennard-Jones interactions can produce structurally accurate CG molecular dynamics simulations. Using this simple approach, we provide a set of interaction parameters for Calcium, Magnesium, and Zinc ions, which cover more than 80% of the metal-bound structures reported on the PDB. Simulations performed using the SIRAH force field on several proteins and DNA systems show that using the present approach it is possible to obtain non-bonded interaction parameters that obviate the use of topological constraints. </p>

Active Learning Accelerates Ab Initio Molecular Dynamics on Pericyclic Reactive Energy Surfaces

2020 ◽  
Author(s):  
Shi Jun Ang ◽  
Wujie Wang ◽  
Daniel Schwalbe-Koda ◽  
Simon Axelrod ◽  
Rafael Gomez-Bombarelli
Keyword(s):  
Molecular Dynamics ◽  
Potential Energy ◽  
Transition State ◽  
Potential Energy Surface ◽  
Ab Initio ◽  
Energy Surface ◽  
Computational Cost ◽  
Reactive Trajectories ◽  
Dynamics Simulations ◽  
Energy Surfaces

<div>Modeling dynamical effects in chemical reactions, such as post-transition state bifurcation, requires <i>ab initio</i> molecular dynamics simulations due to the breakdown of simpler static models like transition state theory. However, these simulations tend to be restricted to lower-accuracy electronic structure methods and scarce sampling because of their high computational cost. Here, we report the use of statistical learning to accelerate reactive molecular dynamics simulations by combining high-throughput ab initio calculations, graph-convolution interatomic potentials and active learning. This pipeline was demonstrated on an ambimodal trispericyclic reaction involving 8,8-dicyanoheptafulvene and 6,6-dimethylfulvene. With a dataset size of approximately</div><div>31,000 M062X/def2-SVP quantum mechanical calculations, the computational cost of exploring the reactive potential energy surface was reduced by an order of magnitude. Thousands of virtually costless picosecond-long reactive trajectories suggest that post-transition state bifurcation plays a minor role for the reaction in vacuum. Furthermore, a transfer-learning strategy effectively upgraded the potential energy surface to higher</div><div>levels of theory ((SMD-)M06-2X/def2-TZVPD in vacuum and three other solvents, as well as the more accurate DLPNO-DSD-PBEP86 D3BJ/def2-TZVPD) using about 10% additional calculations for each surface. Since the larger basis set and the dynamic correlation capture intramolecular non-covalent interactions more accurately, they uncover longer lifetimes for the charge-separated intermediate on the more accurate potential energy surfaces. The character of the intermediate switches from entropic to thermodynamic upon including implicit solvation effects, with lifetimes increasing with solvent polarity. Analysis of 2,000 reactive trajectories on the chloroform PES shows a qualitative agreement with the experimentally-reported periselectivity for this reaction. This overall approach is broadly applicable and opens a door to the study of dynamical effects in larger, previously-intractable reactive systems.</div>

Message Passing Neural Networks for Partial Charge Assignment to Metal-Organic Frameworks

2020 ◽  
Author(s):  
Ali Raza ◽  
Arni Sturluson ◽  
Cory Simon ◽  
Xiaoli Fern
Keyword(s):  
Electronic Structure ◽  
Message Passing ◽  
Gas Adsorption ◽  
Metal Organic Frameworks ◽  
Computational Cost ◽  
Electronic Structure Calculations ◽  
High Fidelity ◽  
Partial Charges ◽  
Metal Organic ◽  
Structure Calculations

Virtual screenings can accelerate and reduce the cost of discovering metal-organic frameworks (MOFs) for their applications in gas storage, separation, and sensing. In molecular simulations of gas adsorption/diffusion in MOFs, the adsorbate-MOF electrostatic interaction is typically modeled by placing partial point charges on the atoms of the MOF. For the virtual screening of large libraries of MOFs, it is critical to develop computationally inexpensive methods to assign atomic partial charges to MOFs that accurately reproduce the electrostatic potential in their pores. Herein, we design and train a message passing neural network (MPNN) to predict the atomic partial charges on MOFs under a charge neutral constraint. A set of ca. 2,250 MOFs labeled with high-fidelity partial charges, derived from periodic electronic structure calculations, serves as training examples. In an end-to-end manner, from charge-labeled crystal graphs representing MOFs, our MPNN machine-learns features of the local bonding environments of the atoms and learns to predict partial atomic charges from these features. Our trained MPNN assigns high-fidelity partial point charges to MOFs with orders of magnitude lower computational cost than electronic structure calculations. To enhance the accuracy of virtual screenings of large libraries of MOFs for their adsorption-based applications, we make our trained MPNN model and MPNN-charge-assigned computation-ready, experimental MOF structures publicly available.<br>

Sign in / Sign up

Export Citation Format

Share Document