-
Transported Memory Networks accelerating Computational Fluid Dynamics
Authors:
Matthias Schulz,
Gwendal Jouan,
Daniel Berger,
Stefan Gavranovic,
Dirk Hartmann
Abstract:
In recent years, augmentation of differentiable PDE solvers with neural networks has shown promising results, particularly in fluid simulations. However, most approaches rely on convolutional neural networks and custom solvers operating on Cartesian grids with efficient access to cell data. This particular choice poses challenges for industrial-grade solvers that operate on unstructured meshes, wh…
▽ More
In recent years, augmentation of differentiable PDE solvers with neural networks has shown promising results, particularly in fluid simulations. However, most approaches rely on convolutional neural networks and custom solvers operating on Cartesian grids with efficient access to cell data. This particular choice poses challenges for industrial-grade solvers that operate on unstructured meshes, where access is restricted to neighboring cells only. In this work, we address this limitation using a novel architecture, named Transported Memory Networks. The architecture draws inspiration from both traditional turbulence models and recurrent neural networks, and it is fully compatible with generic discretizations. Our results show that it is point-wise and statistically comparable to, or improves upon, previous methods in terms of both accuracy and computational efficiency.
△ Less
Submitted 25 February, 2025;
originally announced February 2025.
-
Comparing cars with apples? Identifying the appropriate benchmark countries for relative ecological pollution rankings and international learning
Authors:
Dominik Hartmann,
Diogo Ferraz,
Mayra Bezerra,
Andreas Pyka,
Flavio L. Pinheiro
Abstract:
Research in Data Envelopment Analysis has created rankings of the ecological efficiency of countries' economies. At the same time, research in economic complexity has provided new methods to depict productive structures and has analyzed how economic diversification and sophistication affect environmental pollution indicators. However, no research so far has compared the ecological efficiency of co…
▽ More
Research in Data Envelopment Analysis has created rankings of the ecological efficiency of countries' economies. At the same time, research in economic complexity has provided new methods to depict productive structures and has analyzed how economic diversification and sophistication affect environmental pollution indicators. However, no research so far has compared the ecological efficiency of countries with similar productive structures and levels of economic complexity, combining the strengths of both approaches. In this article, we use data on 774 different types of exports, CO2 emissions, and the ecological footprint of 99 countries to create a relative ecological pollution ranking (REPR). Moreover, we use methods from network science to reveal a benchmark network of the best learning partners based on country pairs with a large extent of export similarity, yet significant differences in pollution values. This is important because it helps to reveal adequate benchmark countries for efficiency improvements and cleaner production, considering that countries may specialize in substantially different types of economic activities. Finally, the article (i) illustrates large efficiency improvements within current global output levels, (ii) helps to identify countries that can best learn from each other, and (iii) improves the information base in international negotiations for the sake of a clean global production system.
△ Less
Submitted 29 July, 2021;
originally announced July 2021.
-
COSI: From Calibrations and Observations to All-sky Images
Authors:
Andreas Zoglauer,
Thomas Siegert,
Alexander Lowell,
Brent Mochizuki,
Carolyn Kierans,
Clio Sleator,
Dieter H. Hartmann,
Hadar Lazar,
Hannah Gulick,
Jacqueline Beechert,
Jarred M. Roberts,
John A. Tomsick,
Mark D. Leising,
Nicholas Pellegrini,
Steven E. Boggs,
Terri J. Brandt
Abstract:
The soft MeV gamma-ray sky, from a few hundred keV up to several MeV, is one of the least explored regions of the electromagnetic spectrum. The most promising technology to access this energy range is a telescope that uses Compton scattering to detect the gamma rays. Going from the measured data to all-sky images ready for scientific interpretation, however, requires a well-understood detector set…
▽ More
The soft MeV gamma-ray sky, from a few hundred keV up to several MeV, is one of the least explored regions of the electromagnetic spectrum. The most promising technology to access this energy range is a telescope that uses Compton scattering to detect the gamma rays. Going from the measured data to all-sky images ready for scientific interpretation, however, requires a well-understood detector setup and a multi-step data-analysis pipeline. We have developed these capabilities for the Compton Spectrometer and Imager (COSI). Starting with a deep understanding of the many intricacies of the Compton measurement process and the Compton data space, we developed the tools to perform simulations that match well with instrument calibrations and to reconstruct the gamma-ray path in the detector. Together with our work to create an adequate model of the measured background while in flight, we are able to perform spectral and polarization analysis, and create images of the gamma-ray sky. This will enable future telescopes to achieve a deeper understanding of the astrophysical processes that shape the gamma-ray sky from the sites of star formation (26-Al map), to the history of core-collapse supernovae (e.g. 60-Fe map) and the distributions of positron annihilation (511-keV map) in our Galaxy.
△ Less
Submitted 25 February, 2021;
originally announced February 2021.
-
Machine Learning-Based Optimal Mesh Generation in Computational Fluid Dynamics
Authors:
Keefe Huang,
Moritz Krügener,
Alistair Brown,
Friedrich Menhorn,
Hans-Joachim Bungartz,
Dirk Hartmann
Abstract:
Computational Fluid Dynamics (CFD) is a major sub-field of engineering. Corresponding flow simulations are typically characterized by heavy computational resource requirements. Often, very fine and complex meshes are required to resolve physical effects in an appropriate manner. Since all CFD algorithms scale at least linearly with the size of the underlying mesh discretization, finding an optimal…
▽ More
Computational Fluid Dynamics (CFD) is a major sub-field of engineering. Corresponding flow simulations are typically characterized by heavy computational resource requirements. Often, very fine and complex meshes are required to resolve physical effects in an appropriate manner. Since all CFD algorithms scale at least linearly with the size of the underlying mesh discretization, finding an optimal mesh is key for computational efficiency.
One methodology used to find optimal meshes is goal-oriented adaptive mesh refinement. However, this is typically computationally demanding and only available in a limited number of tools. Within this contribution, we adopt a machine learning approach to identify optimal mesh densities. We generate optimized meshes using classical methodologies and propose to train a convolutional network predicting optimal mesh densities given arbitrary geometries. The proposed concept is validated along 2d wind tunnel simulations with more than 60,000 simulations. Using a training set of 20,000 simulations we achieve accuracies of more than 98.7%.
Corresponding predictions of optimal meshes can be used as input for any mesh generation and CFD tool. Thus without complex computations, any CFD engineer can start his predictions from a high quality mesh.
△ Less
Submitted 25 February, 2021;
originally announced February 2021.
-
A neural network multigrid solver for the Navier-Stokes equations
Authors:
Nils Margenberg,
Dirk Hartmann,
Christian Lessig,
Thomas Richter
Abstract:
We present the deep neural network multigrid solver (DNN-MG) that we develop for the instationary Navier-Stokes equations. DNN-MG improves computational efficiency using a judicious combination of a geometric multigrid solver and a recurrent neural network with memory. DNN-MG uses the multi-grid method to classically solve on coarse levels while the neural network corrects interpolated solutions o…
▽ More
We present the deep neural network multigrid solver (DNN-MG) that we develop for the instationary Navier-Stokes equations. DNN-MG improves computational efficiency using a judicious combination of a geometric multigrid solver and a recurrent neural network with memory. DNN-MG uses the multi-grid method to classically solve on coarse levels while the neural network corrects interpolated solutions on fine ones, thus avoiding the increasingly expensive computations that would have to be performed there. This results in a reduction in computation time through DNN-MG's highly compact neural network. The compactness results from its design for local patches and the available coarse multigrid solutions that provides a "guide" for the corrections. A compact neural network with a small number of parameters also reduces training time and data. Furthermore, the network's locality facilitates generalizability and allows one to use DNN-MG trained on one mesh domain also on different ones. We demonstrate the efficacy of DNN-MG for variations of the 2D laminar flow around an obstacle. For these, our method significantly improves the solutions as well as lift and drag functionals while requiring only about half the computation time of a full multigrid solution. We also show that DNN-MG trained for the configuration with one obstacle can be generalized to other time dependent problems that can be solved efficiently using a geometric multigrid method.
△ Less
Submitted 29 November, 2021; v1 submitted 26 August, 2020;
originally announced August 2020.
-
Validating the ASCOT modelling of NBI fast ions in Wendelstein 7-X stellarator
Authors:
S. Äkäslompolo,
P. Drewelow,
Y. Gao,
A. Ali,
C. Biedermann,
S. Bozhenkov,
C. P. Dhard,
M. Endler,
J. Fellinger,
O. P. Ford,
B. Geiger,
J. Geiger,
N. den Harderd,
D. Hartmann,
D. Hathiramani,
M. Isobe,
M. Jakubowski,
Y. Kazakov,
C. Killer,
S. Lazerson,
M. Mayerd,
P. McNeely,
D. Naujoks,
T. W. C. Neelis,
J. Kontula
, et al. (15 additional authors not shown)
Abstract:
The first fast ion experiments in Wendelstein 7-X were performed in 2018. They are one of the first steps in demonstrating the optimised fast ion confinement of the stellarator. The fast ions were produced with a neutral beam injection (NBI) system and detected with infrared cameras (IR), a fast ion loss detector (FILD), fast ion charge exchange spectroscopy (FIDA), and post-mortem analysis of pla…
▽ More
The first fast ion experiments in Wendelstein 7-X were performed in 2018. They are one of the first steps in demonstrating the optimised fast ion confinement of the stellarator. The fast ions were produced with a neutral beam injection (NBI) system and detected with infrared cameras (IR), a fast ion loss detector (FILD), fast ion charge exchange spectroscopy (FIDA), and post-mortem analysis of plasma facing components. The fast ion distribution function in the plasma and at the wall is being modelled with the ASCOT suite of codes. They calculate the ionisation of the injected neutrals and the consecutive slowing down process of the fast ions. The primary output of the code is the multidimensional fast ion distribution function within the plasma and the distribution of particle hit locations and velocities on the wall. Synthetic measurements based on ASCOT output are compared to experimental results to assess the validity of the modelling. This contribution presents an overview of the various fast ion measurements in 2018 and the current modelling status. The validation and data-analysis is on-going, but the wall load IR modelling already yield results that match with the experiments.
△ Less
Submitted 10 September, 2019; v1 submitted 18 June, 2019;
originally announced June 2019.
-
Validating fast-ion wall-load IR analysis-methods against W7-X NBI empty-torus experiment
Authors:
S. Äkäslompolo,
P. Drewelow,
Y. Gao,
A. Ali,
O. Asunta,
S. Bozhenkov,
J. Fellinger,
O. P. Ford,
N. den Harder,
D. Hartmann,
M. Jakubowski,
P. McNeely,
H. Niemann,
F. Pisano,
N. Rust,
A. Puig Sitjes,
M. Sleczka,
A. Spanier,
R. C. Wolf,
the W7-X Team
Abstract:
The first neutral beam injection (NBI) experiments in Wendelstein 7-X (W7-X) stellarator were conducted in the summer of 2018. The NBI system is used to heat the magnetically confined plasma by neutralising an accelerated hydrogen ion beam and directing it into the plasma, where the resulting energetic ions release their energy to heat the plasma. The modelling of the NBI fast ion experiments has…
▽ More
The first neutral beam injection (NBI) experiments in Wendelstein 7-X (W7-X) stellarator were conducted in the summer of 2018. The NBI system is used to heat the magnetically confined plasma by neutralising an accelerated hydrogen ion beam and directing it into the plasma, where the resulting energetic ions release their energy to heat the plasma. The modelling of the NBI fast ion experiments has commenced, including estimation of the shine-through and the orbit-losses. The stellarator has a wide-angle infra red (IR) imaging system to monitor the machine plasma facing component surface temperatures. This work validates the NBI model "Beamlet Based NBI (BBNBI)" and the newly written synthetic IR camera model. The validation is accomplished by comparing the measured and the synthetic IR camera measurements of an experiment where the NBI was injected into the vacuum vessel without a plasma. A good qualitative and quantitative match was found. This agreement is further supported by spectroscopic and calibration measurements of the NBI and and IR camera systems.
△ Less
Submitted 20 June, 2019; v1 submitted 6 June, 2019;
originally announced June 2019.
-
Cage occupancies of methane hydrates: Results from synchrotron X-ray diffraction and Raman spectroscopy
Authors:
Junfeng Qin,
Christiane D. Hartmann,
Werner F. Kuhs
Abstract:
An accurate knowledge of cage occupancy of methane is central for understanding the physical-chemical properties of gas hydrates, the actual inventory of natural gas in hydrate deposits and the description of gas exchange processes. Here we report the absolute cage occupancies, the cage occupancy ratios and hydration numbers of the synthetic CH4-H2O and CH4-D2O hydrates formed from the ice-gas sys…
▽ More
An accurate knowledge of cage occupancy of methane is central for understanding the physical-chemical properties of gas hydrates, the actual inventory of natural gas in hydrate deposits and the description of gas exchange processes. Here we report the absolute cage occupancies, the cage occupancy ratios and hydration numbers of the synthetic CH4-H2O and CH4-D2O hydrates formed from the ice-gas system under different pressures and temperatures. The results were obtained from Rietveld refinement using high-resolution synchrotron X-ray powder diffraction patterns and from Raman spectroscopic measurements. The small-cage occupancies of methane in the deuterated hydrates are found to be slightly higher than in the hydrogenated form, likely due to their different lattice constants. The CH4 occupancy in the small cages agrees fairly well with the predictions of CSMGem at the formation pressure of 3.5 MPa, but with the increasing formation pressure the disagreement grows up to 11 percent. While some deficiency of the prediction model cannot be excluded, the observed discrepancy may well be due to experimental difficulties of reaching true equilibrium at higher pressures. The experimentally determined large-to-small cage occupancy ratios of the synthetic and natural CH4 hydrates formed from the water-gas system are consistently higher than the results of CSMGem calculations. Possible reasons for these discrepancies will be discussed.
△ Less
Submitted 6 March, 2018;
originally announced March 2018.
-
Shooting High or Low: Do Countries Benefit from Entering Unrelated Activities?
Authors:
Flávio L. Pinheiro,
Aamena Alshamsi,
Dominik Hartmann,
Ron Boschma,
César A. Hidalgo
Abstract:
Countries tend to diversify their exports by entering products that are related to their current exports. Yet this average behavior is not representative of every diversification path. In this paper, we introduce a method to identify periods when countries enter unrelated products. We analyze the economic diversification paths of 93 countries between 1965 and 2014 and find that countries enter unr…
▽ More
Countries tend to diversify their exports by entering products that are related to their current exports. Yet this average behavior is not representative of every diversification path. In this paper, we introduce a method to identify periods when countries enter unrelated products. We analyze the economic diversification paths of 93 countries between 1965 and 2014 and find that countries enter unrelated products in only about 7.2% of all observations. We find that countries enter more unrelated products when they are at an intermediate level of economic development, and when they have higher levels of human capital. Finally, we ask whether countries entering more unrelated products grow faster than those entering only related products. The data shows that countries that enter more unrelated activities experience a small but significant increase in future economic growth, compared to countries with a similar level of income, human capital, capital stock per worker, and economic complexity.
△ Less
Submitted 20 March, 2018; v1 submitted 16 January, 2018;
originally announced January 2018.
-
The Research Space: using the career paths of scholars to predict the evolution of the research output of individuals, institutions, and nations
Authors:
Miguel R. Guevara,
Dominik Hartmann,
Manuel Aristarán,
Marcelo Mendoza,
César A. Hidalgo
Abstract:
In recent years scholars have built maps of science by connecting the academic fields that cite each other, are cited together, or that cite a similar literature. But since scholars cannot always publish in the fields they cite, or that cite them, these science maps are only rough proxies for the potential of a scholar, organization, or country, to enter a new academic field. Here we use a large d…
▽ More
In recent years scholars have built maps of science by connecting the academic fields that cite each other, are cited together, or that cite a similar literature. But since scholars cannot always publish in the fields they cite, or that cite them, these science maps are only rough proxies for the potential of a scholar, organization, or country, to enter a new academic field. Here we use a large dataset of scholarly publications disambiguated at the individual level to create a map of science-or research space-where links connect pairs of fields based on the probability that an individual has published in both of them. We find that the research space is a significantly more accurate predictor of the fields that individuals and organizations will enter in the future than citation based science maps. At the country level, however, the research space and citations based science maps are equally accurate. These findings show that data on career trajectories-the set of fields that individuals have previously published in-provide more accurate predictors of future research output for more focalized units-such as individuals or organizations-than citation based science maps.
△ Less
Submitted 14 April, 2016; v1 submitted 26 February, 2016;
originally announced February 2016.
-
Linking Economic Complexity, Institutions and Income Inequality
Authors:
D. Hartmann,
M. R. Guevara,
C. Jara-Figueroa,
M. Aristaran,
C. A. Hidalgo
Abstract:
A country's mix of products predicts its subsequent pattern of diversification and economic growth. But does this product mix also predict income inequality? Here we combine methods from econometrics, network science, and economic complexity to show that countries exporting complex products (as measured by the Economic Complexity Index) have lower levels of income inequality than countries exporti…
▽ More
A country's mix of products predicts its subsequent pattern of diversification and economic growth. But does this product mix also predict income inequality? Here we combine methods from econometrics, network science, and economic complexity to show that countries exporting complex products (as measured by the Economic Complexity Index) have lower levels of income inequality than countries exporting simpler products. Using multivariate regression analysis, we show that economic complexity is a significant and negative predictor of income inequality and that this relationship is robust to controlling for aggregate measures of income, institutions, export concentration, and human capital. Moreover, we introduce a measure that associates a product to a level of income inequality equal to the average GINI of the countries exporting that product (weighted by the share the product represents in that country's export basket). We use this measure together with the network of related products (or product space) to illustrate how the development of new products is associated with changes in income inequality. These findings show that economic complexity captures information about an economy's level of development that is relevant to the ways an economy generates and distributes its income. Moreover, these findings suggest that a country's productive structure may limit its range of income inequality. Finally, we make our results available through an online resource that allows for its users to visualize the structural transformation of over 150 countries and their associated changes in income inequality between 1963 and 2008.
△ Less
Submitted 4 January, 2017; v1 submitted 28 May, 2015;
originally announced May 2015.
-
Magnetic Field Generation in Core-Sheath Jets via the Kinetic Kelvin-Helmholtz Instability
Authors:
K. -I. Nishikawa,
P. E. Hardee,
I. Dutan,
J. Niemiec,
M. Medvedev,
Y. Mizuno,
A. Meli,
H. Sol,
B. Zhang,
M. Pohl,
D. H. Hartmann
Abstract:
We have investigated magnetic field generation in velocity shears via the kinetic Kelvin-Helmholtz instability (kKHI) using a relativistic plasma jet core and stationary plasma sheath. Our three-dimensional particle-in-cell simulations consider plasma jet cores with Lorentz factors of 1.5, 5, and 15 for both electron-proton and electron-positron plasmas. For electron-proton plasmas we find generat…
▽ More
We have investigated magnetic field generation in velocity shears via the kinetic Kelvin-Helmholtz instability (kKHI) using a relativistic plasma jet core and stationary plasma sheath. Our three-dimensional particle-in-cell simulations consider plasma jet cores with Lorentz factors of 1.5, 5, and 15 for both electron-proton and electron-positron plasmas. For electron-proton plasmas we find generation of strong large-scale DC currents and magnetic fields which extend over the entire shear-surface and reach thicknesses of a few tens of electron skin depths. For electron-positron plasmas we find generation of alternating currents and magnetic fields. Jet and sheath plasmas are accelerated across the shear surface in the strong magnetic fields generated by the kKHI. The mixing of jet and sheath plasmas generates transverse structure similar to that produced by the Weibel instability.
△ Less
Submitted 31 July, 2014; v1 submitted 20 May, 2014;
originally announced May 2014.
