Efficient and Scalable Initialization of Partitioned Coupled Simulations with preCICE

preCICE is an open-source library, that provides comprehensive functionality to couple independent parallelized solver codes to establish a partitioned multi-physics multi-code simulation environment. For data communication between the respective executables at runtime, it implements a peer-to-peer concept, which renders the computational cost of the coupling per time step negligible compared to the typical run time of the coupled codes. To initialize the peer-to-peer coupling, the mesh partitions of the respective solvers need to be compared to determine the point-to-point communication channels between the processes of both codes. This initialization effort can become a limiting factor, if we either reach memory limits or if we have to re-initialize communication relations in every time step. In this contribution, we remove two remaining bottlenecks: (i) We base the neighborhood search between mesh entities of two solvers on a tree data structure to avoid quadratic complexity, and (ii) we replace the sequential gather-scatter comparison of both mesh partitions by a two-level approach that first compares bounding boxes around mesh partitions in a sequential manner, subsequently establishes pairwise communication between processes of the two solvers, and finally compares mesh partitions between connected processes in parallel. We show, that the two-level initialization method is fives times faster than the old one-level scheme on 24,567 CPU-cores using a mesh with 628,898 vertices. In addition, the two-level scheme is able to handle much larger computational meshes, since the central mesh communication of the one-level scheme is replaced with a fully point-to-point mesh communication scheme.

Download Full-text

Fast Conjugate Heat Transfer Simulation of Long Transient Flexible Operations Using Adaptive Time Stepping

Journal of Turbomachinery ◽

10.1115/1.4040997 ◽

2018 ◽

Vol 140 (9) ◽

Cited By ~ 2

Author(s):

R. Maffulli ◽

L. He ◽

P. Stein ◽

G. Marinescu

Keyword(s):

Heat Transfer ◽

Conjugate Heat Transfer ◽

Time Derivative ◽

Computational Cost ◽

Strong Impact ◽

Time Step ◽

Time Stepping ◽

Adaptive Time ◽

Fully Coupled ◽

Adaptive Time Stepping

The emerging renewable energy market calls for more advanced prediction tools for turbine transient operations in fast startup/shutdown cycles. Reliable numerical analysis of such transient cycles is complicated by the disparity in time scales of the thermal responses in fluid and solid domains. Obtaining fully coupled time-accurate unsteady conjugate heat transfer (CHT) results under these conditions would require to march in both domains using the time-step dictated by the fluid domain: typically, several orders of magnitude smaller than the one required by the solid. This requirement has strong impact on the computational cost of the simulation as well as being potentially detrimental to the accuracy of the solution due to accumulation of round-off errors in the solid. A novel loosely coupled CHT methodology has been recently proposed, and successfully applied to both natural and forced convection cases that remove these requirements through a source-term based modeling (STM) approach of the physical time derivative terms in the relevant equations. The method has been shown to be numerically stable for very large time steps with adequate accuracy. The present effort is aimed at further exploiting the potential of the methodology through a new adaptive time stepping approach. The proposed method allows for automatic time-step adjustment based on estimating the magnitude of the truncation error of the time discretization. The developed automatic time stepping strategy is applied to natural convection cases under long (2000 s) transients: relevant to the prediction of turbine thermal loads during fast startups/shutdowns. The results of the method are compared with fully coupled unsteady simulations showing comparable accuracy with a significant reduction of the computational costs.

Download Full-text

Score matching filters

10.5194/egusphere-egu21-16125 ◽

2021 ◽

Author(s):

Marie Turčičová ◽

Jan Mandel ◽

Kryštof Eben

Keyword(s):

Kalman Filter ◽

Covariance Matrix ◽

Markov Random Fields ◽

Computational Cost ◽

Time Step ◽

Sample Covariance ◽

Monte Carlo Approximation ◽

Gauss Markov Random Fields ◽

Ensemble Filtering ◽

Non Gaussian

A widely popular group of data assimilation methods in meteorological and geophysical sciences is formed by filters based on Monte-Carlo approximation of the traditional Kalman filter, e.g. Ensemble Kalman filter (EnKF), Ensemble square-root filter and others. Due to the computational cost, ensemble size is usually small compared to the dimension of the state vector. Traditional EnKF implicitly uses the sample covariance which is a poor estimate of the background covariance matrix - singular and contaminated by spurious correlations. We focus on modelling the background covariance matrix by means of a linear model for its inverse. This is particularly useful in Gauss-Markov random fields (GMRF), where the inverse covariance matrix has a banded structure. The parameters of the model are estimated by the score matching method which provides estimators in a closed form, cheap to compute. The resulting estimate is a key component of the proposed ensemble filtering algorithms. Under the assumption that the state vector is a GMRF in every time-step, the Score matching filter with Gaussian resampling (SMF-GR) gives in every time-step a consistent (in the large ensemble limit) estimator of mean and covariance matrix of the forecast and analysis distribution. Further, we propose a filtering method called Score matching ensemble filter (SMEF), based on regularization of the EnKF. This filter performs well even for non-Gaussian systems with non-linear dynamics. The performance of both filters is illustrated on a simple linear convection model and Lorenz-96.

Download Full-text

Finite element method completely implemented for graphic processor units using parallel algorithm libraries

The International Journal of High Performance Computing Applications ◽

10.1177/1094342017694703 ◽

2017 ◽

Vol 33 (1) ◽

pp. 53-66 ◽

Cited By ~ 1

Author(s):

Franz Pichler ◽

Gundolf Haase

Keyword(s):

Finite Element ◽

Graphics Processing Unit ◽

Computational Cost ◽

Processing Unit ◽

Time Step ◽

Device Architecture ◽

Transient Problems ◽

Speed Up ◽

Automotive Batteries ◽

Graphics Processing

A finite element code is developed in which all of the computationally expensive steps are performed on a graphics processing unit via the THRUST and the PARALUTION libraries. The code focuses on the simulation of transient problems where the repeated computations per time-step create the computational cost. It is used to solve partial and ordinary differential equations as they arise in thermal-runaway simulations of automotive batteries. The speed-up obtained by utilizing the graphics processing unit for every critical step is compared against the single core and the multi-threading solutions which are also supported by the chosen libraries. This way a high total speed-up on the graphics processing unit is achieved without the need for programming a single classical Compute Unified Device Architecture kernel.

Download Full-text

SpaceWire

Electromagnetic Compatibility for Space Systems Design - Advances in Computer and Electrical Engineering ◽

10.4018/978-1-5225-5415-8.ch002 ◽

2018 ◽

pp. 38-70

Author(s):

Anargyros T. Baklezos ◽

Christos N. Capsalis

Keyword(s):

System Performance ◽

High Speed ◽

Data Communication ◽

Space Missions ◽

Development Cost ◽

Communication Links ◽

Recent Developments ◽

High Speed Data ◽

Electromagnetic Behavior ◽

Point To Point

SpaceWire is a point-to-point bit shipping protocol for high-speed data communication links and networks providing equipment compatibility and seamless component reusability. It has found great application in many space missions reducing the development cost, offering architectural flexibility and improving reliability. This chapter delves into the standard describing the SpaceWire, focusing on the lower levels that play a key role in the electromagnetic behavior of the system and concern cable assemblies, shielding, bonding, and grounding. Findings regarding emissions affecting spacecraft components are presented as well as other EMC issues that have an impact on the system performance. Recent developments and upcoming updates to the standard are also presented and discussed.

Download Full-text

A Sequential Approach to Numerical Simulations of Solidification with Domain and Time Decomposition

Applied Sciences ◽

10.3390/app9101972 ◽

2019 ◽

Vol 9 (10) ◽

pp. 1972 ◽

Cited By ~ 5

Author(s):

Elzbieta Gawronska

Keyword(s):

Parallel Computing ◽

Computational Cost ◽

Fixed Time ◽

Simulation Software ◽

Computational Method ◽

Computational Time ◽

Time Step ◽

Sequential Approach ◽

Time Partitioning ◽

Physical Phenomena

Progress in computational methods has been stimulated by the widespread availability of cheap computational power leading to the improved precision and efficiency of simulation software. Simulation tools become indispensable tools for engineers who are interested in attacking increasingly larger problems or are interested in searching larger phase space of process and system variables to find the optimal design. In this paper, we show and introduce a new approach to a computational method that involves mixed time stepping scheme and allows to decrease computational cost. Implementation of our algorithm does not require a parallel computing environment. Our strategy splits domains of a dynamically changing physical phenomena and allows to adjust the numerical model to various sub-domains. We are the first (to our best knowledge) to show that it is possible to use a mixed time partitioning method with various combination of schemes during binary alloys solidification. In particular, we use a fixed time step in one domain, and look for much larger time steps in other domains, while maintaining high accuracy. Our method is independent of a number of domains considered, comparing to traditional methods where only two domains were considered. Mixed time partitioning methods are of high importance here, because of natural separation of domain types. Typically all important physical phenomena occur in the casting and are of high computational cost, while in the mold domains less dynamic processes are observed and consequently larger time step can be chosen. Finally, we performed series of numerical experiments and demonstrate that our approach allows reducing computational time by more than three times without losing the significant precision of results and without parallel computing.

Download Full-text

A New Accelerated Approach for Spectral Radiation Calculation Using K Distribution and Discrete Ordinate Methods With Application in Industrial Combustion Systems

ASME 2019 Heat Transfer Summer Conference ◽

10.1115/ht2019-3636 ◽

2019 ◽

Author(s):

Andrew M. Feldick ◽

Gopalendu Pal

Keyword(s):

Computational Cost ◽

Computational Time ◽

Discrete Ordinates ◽

Discrete Ordinate ◽

Spectral Radiation ◽

Time Step ◽

Scattering Coefficients ◽

Coupled Equations ◽

Spectral Models ◽

Radiation Time

Abstract The introduction of higher fidelity spectral models into a Discrete Ordinates Method (DOM) RTE solver introduces the challenge of solving the N(N+2) coupled equations in intensity over many spectral points. The inability to store intensity fields leads to a nonlinear increase in computational cost as compared to basic gray models, as the solution in an evolving field must be recalculated at each radiation time step. In this paper an approximate initialization approach is used to a reconstructed values of the intensities. This approach is particularly well suited to spectrally reordered methods, as the boundary conditions and scattering coefficients are gray. This approach leads to more tractable computational time, and is demonstrated using on two industrial scale flames.

Download Full-text

A Higher-Order Closure Model with an Explicit PBL Top

Journal of the Atmospheric Sciences ◽

10.1175/2009jas3205.1 ◽

2010 ◽

Vol 67 (3) ◽

pp. 834-850 ◽

Cited By ~ 4

Author(s):

Cara-Lyn Lappen ◽

David Randall ◽

Takanobu Yamaguchi

Keyword(s):

Large Scale ◽

Computational Cost ◽

Higher Order ◽

Entrainment Rate ◽

Time Step ◽

Vertical Coordinate ◽

Second Moments ◽

Scale Models ◽

Vertical Grid ◽

Flux Model

Abstract In 2001, the authors presented a higher-order mass-flux model called “assumed distributions with higher-order closure” (ADHOC 1), which represents the large eddies of the planetary boundary layer (PBL) in terms of an assumed joint distribution of the vertical velocity and scalars. In a subsequent version (ADHOC 2) the authors incorporated vertical momentum fluxes and second moments involving pressure perturbations into the framework. These versions of ADHOC, as well as all other higher-order closure models, are not suitable for use in large-scale models because of the high vertical and temporal resolution that is required. This high resolution is needed mainly because higher-order closure (HOC) models must resolve discontinuities at the PBL top, which can occur anywhere on a model’s Eulerian vertical grid. This paper reports the development of ADHOC 3, in which the computational cost of the model is reduced by introducing the PBL depth as an explicit prognostic variable. ADHOC 3 uses a stretched vertical coordinate that is attached to the PBL top. The discontinuous jumps at the PBL top are “hidden” in the layer edge that represents the PBL top. This new HOC model can use much coarser vertical resolution and a longer time step and is thus suitable for use in large-scale models. To predict the PBL depth, an entrainment parameterization is needed. In the development of the model, the authors have been led to a new view of the old problem of entrainment parameterization. The relatively detailed information available in the HOC model is used to parameterize the entrainment rate. The present approach thus borrows ideas from mixed-layer modeling to create a new, more economical type of HOC model that is better suited for use as a parameterization in large-scale models.

Download Full-text

A RANS numerical model for cross-shore beach profile evolution

10.5194/egusphere-egu2020-10501 ◽

2020 ◽

Author(s):

Julio Garcia-Maribona ◽

Javier L. Lara ◽

Maria Maza ◽

Iñigo J. Losada

Keyword(s):

Sediment Transport ◽

Numerical Model ◽

Transport Model ◽

Numerical Models ◽

Computational Cost ◽

Physical Modelling ◽

Computational Effort ◽

Beach Profile ◽

Time Step ◽

Wave Conditions

The evolution of the cross-shore beach profile is tightly related to the evolution of the coastline in both small and large time scales. Bathymetry changes in extreme maritime events can also have important effects on coastal infrastructures such as geotechnical failures of foundations or the modification of the incident wave conditions towards a more unfavourable situation.The available strategies to study the evolution of beach profiles can be classified in analytical, physical and numerical modelling. Analytical solutions are fast, but too simplistic for many applications. Physical modelling provides trustworthy results and can be applied to a wide variety of configurations, however, they are costly and time-consuming compared to analytical strategies. Finally, &#160;numerical approaches offer different balances between cost and precision depending on the particular model.Some numerical models provide greater precision in the beach profile evolution, but incurring in a prohibitive computational cost for many applications. In contrast, the less expensive ones assume simplifications which do not allow to correctly reproduce significant phenomena of the near-shore hydrodynamics such as wave breaking or undertow currents, neither to predict important features of the beach profile like breaker bars.In this work, a new numerical model is developed to reproduce the main features of the beach profile and hydrodynamics while maintaining an affordable computational cost. In addition, it is intended to reduce to the minimum the number of coefficients that the user has to provide to make the model more predictive.The model consists of two main modules. Firstly, the already existing 2D RANS numerical model IH2VOF is used to compute the hydrodynamics. Secondly, the sediment transport model modifies the bathymetry according to the obtained hydrodynamics. The new bathymetry is then considered in the hydrodynamic model to account for it in the next time step.The sediment transport module considers bedload and suspended transports separately. The former is obtained with empirical formulae. In the later,the distribution of sediment concentration in the domain is obtained by solving an advective-diffusive transport equation. Then, the sedimentation and erosion rates are obtained along the seabed. Once these contributions are calculated, a sediment balance is performed in every seabed segment to determine the variation in its level.With the previously described strategy, the resulting model is able to predict not only the seabed changes due to different wave conditions, but also the influence of this new bathymetry in the hydrodynamics, capturing features such as the generation of a breaker bar, displacement of the breaking point or variation of the run-up over the beach profile. To validate the model, the numerical results are compared to experimental data.An important novelty of the present model is the computational effort required to perform the simulations, which is significantly smaller than the one associated to existing models able to reproduce the same phenomena.

Download Full-text

Entropy Ensemble Filter: Does information content assessment of bootstrapped training datasets before model training lead to better trade-off between ensemble size and predictive performance?

10.5194/egusphere-egu2020-1963 ◽

2020 ◽

Author(s):

Hossein Foroozand ◽

Steven V. Weijs

Keyword(s):

Machine Learning ◽

Computational Cost ◽

Predictive Performance ◽

Original Data ◽

Training Data ◽

Computational Time ◽

Limiting Factor ◽

Ensemble Size ◽

Content Assessment ◽

Model Training

Machine learning is the fast-growing branch of data-driven models, and its main objective is to use computational methods to become more accurate in predicting outcomes without being explicitly programmed. In this field, a way to improve model predictions is to use a large collection of models (called ensemble) instead of a single one. Each model is then trained on slightly different samples of the original data, and their predictions are averaged. This is called bootstrap aggregating, or Bagging, and is widely applied. A recurring question in previous works was: how to choose the ensemble size of training data sets for tuning the weights in machine learning? The computational cost of ensemble-based methods scales with the size of the ensemble, but excessively reducing the ensemble size comes at the cost of reduced predictive performance. The choice of ensemble size was often determined based on the size of input data and available computational power, which can become a limiting factor for larger datasets and complex models&#8217; training. In this research, it is our hypothesis that if an ensemble of artificial neural networks (ANN) models or any other machine learning technique uses the most informative ensemble members for training purpose rather than all bootstrapped ensemble members, it could reduce the computational time substantially without negatively affecting the performance of simulation.

Download Full-text

Combination of aerosol sectional scheme and modal scheme in NorESM: Sensitivities to emissions

10.5194/egusphere-egu2020-9740 ◽

2020 ◽

Author(s):

Sara Marie Blichner ◽

Moa K. Sporre ◽

Risto Makkonen ◽

Terje K. Berntsen

Keyword(s):

Size Range ◽

Cloud Condensation Nuclei ◽

Model Development ◽

Computational Cost ◽

Particle Formation ◽

Complex Nature ◽

Earth System ◽

Time Step ◽

Median Diameter ◽

Aerosol Module

Cloud-aerosol interactions give rise to much of the uncertainty in estimates of climate forcing, climate sensitivity and thus also future climate predictions. Furthermore, the modelled concentration of cloud condensation nuclei (CCN) in the past, present and future are highly dependent on the how the models represent new particle formation (NPF) &#8211; a process which is both poorly understood theoretically and difficult to model due to its complex nature. Global modellers in particular have to prioritize between theoretical accuracy and keeping computational costs low. A common approach in these models is to use a modal scheme to parameterize the sizedistribution of the aerosols, while sectional schemes are in general considered closer to first principals.To better capture the dynamics of early growth in the Norwegian Earth System Model (NorESM), we have implemented a sectional scheme for the smallest particles (currently 5 - 39 nm), which proceeds to feed particles into the original modal scheme (Kirkev&#229;g et al, 2018) after growth. The sectional scheme includes two species, H2SO4 and low volatile organics and has 5 bins. The motivation is: (1) In the original scheme in NorESM, newly formed particles are added to the smallest mode which has a number median diameter of 23.6 nm. The survival of particles from NPF (formed at ~4 nm diameter) to this mode is calculated based on Lehtinen et al (2007). Thus it does not take into account dynamics within this size range, i.e. competition for condensing vapours and growth of particles over more than one time step. (2) Including a sectional scheme in this range adds precision for this crucial stage of growth while keeping the computational cost low due to the limited number of species involved (currently 2 in the model). (3) A sectional scheme within this size range is an interesting alternative to a nucleation mode, which is known to have problems with moving particles to larger sizes at the same time as adding newly formed particles.We present several sensitivity tests which investigate the response of the model to changes in emissions of SO2 and biogenic volatile organic compounds and nucleation parameterizations, with and without the sectional scheme. Our results in particular show that in the globally averaged boundary layer, the sectional scheme drastically reduces the number of particles that survive to the modal scheme compared the original model, while more particles survive in remote regions. On the other hand, the sectional scheme is less sensitive to the choice in NPF/nucleation parameterization.&#160;References:&#160;Lehtinen, Kari E. J., Miikka Dal Maso, Markku Kulmala, and Veli-Matti Kerminen. "Estimating Nucleation Rates from Apparent Particle Formation Rates and Vice Versa: Revised Formulation of the Kerminen&#8211;Kulmala Equation." Journal of Aerosol Science 38, no. 9 (September 1, 2007): 988&#8211;94. https://doi.org/10.1016/j.jaerosci.2007.06.009.Kirkev&#229;g, A., A. Grini, D. Olivi&#233;, &#216;. Seland, K. Alterskj&#230;r, M. Hummel, I.H.H Karset, A. Lewinschal, X. Liu, R. Makkonen, I. Bethke, J. Griesfeller, M. Schulz and T. Iversen. "A production-tagged aerosol module for Earth system models, OsloAero5.3 &#8211; extensions and updates for CAM5.3-Oslo." Geoscientific Model Development 11. no. 10 (October, 2018): 3945--3982. https://doi.org/10.5194/gmd-11-3945-2018

Download Full-text