Harald Köstler
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Harald Köstler has a total of 105 co-authors. Here are the top co-authors by number of shared publications:
- Ulrich Rüde
CERFACS
9 shared publications - Christoph Alt
Paderborn University, Paderborn, Germany and Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
6 shared publications - Samuel Kemmler
Bundesanstalt für Materialforschung und -prüfung (BAM)
5 shared publications - Vadym Aizinger
University of Bayreuth, Bayreuth, Germany
5 shared publications - Axel Klawonn
University of Cologne
4 shared publications - Christian Plessl
Paderborn University, Paderborn, Germany
4 shared publications
Latest Publications
- Open Access
- 21 September 2026
- Published by Elsevier in Computers and Geotechnics
- Vol. 203
Fully-resolved micromechanical study of suction bucket installation regimes with emphasis on piping
- 21 September 2026
- Published by Elsevier in Computers and Geotechnics
- Vol. 203
Open AccessFully-resolved micromechanical study of suction bucket installation regimes with emphasis on piping
- Open Access
- 15 July 2026
- Published by Elsevier in SSRN Electronic Journal
Scaling and Performance Engineering for Wind Park Simulations on Non-Uniform Meshes in WALBERLA-WIND
High-fidelity simulations are central to modern wind-energy research, since they enable the studyof atmospheric turbulence, turbine interactions, and full wind -farm performance. Modeling thesemultiscale effects requires the simultaneous simulation of many turbines in a computational domainspanning hundreds of rotor diameters, thus pushing the limits of current high-performance computingsystems. We evaluate the performance of WALBERLA-WIND, a lattice-Boltzmann -based simulatorfor wind-farms using actuator -model representations of the turbines. In order to resolve the wakedynamics at acceptable computational cost, we employ non -uniform meshes to refine resolution at theatmospheric boundary layer and the turbines. We present single -GPU results on three architectures,demonstrating consistently high aggregate memory -bandwidth utilization. Extensive weak and strongscaling studies on three HPC systems show excellent efficiency up to 4096 GPUs, simulating 8192 windturbines in parallel. These results demonstrate the suitability of WALBERLA-WIND for exascale-classwind-farm simulations when high throughput and energy efficiency are essential.
- 15 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessScaling and Performance Engineering for Wind Park Simulations on Non-Uniform Meshes in WALBERLA-WIND
High-fidelity simulations are central to modern wind-energy research, since they enable the studyof atmospheric turbulence, turbine interactions, and full wind -farm performance. Modeling thesemultiscale effects requires the simultaneous simulation of many turbines in a computational domainspanning hundreds of rotor diameters, thus pushing the limits of current high-performance computingsystems. We evaluate the performance of WALBERLA-WIND, a lattice-Boltzmann -based simulatorfor wind-farms using actuator -model representations of the turbines. In order to resolve the wakedynamics at acceptable computational cost, we employ non -uniform meshes to refine resolution at theatmospheric boundary layer and the turbines. We present single -GPU results on three architectures,demonstrating consistently high aggregate memory -bandwidth utilization. Extensive weak and strongscaling studies on three HPC systems show excellent efficiency up to 4096 GPUs, simulating 8192 windturbines in parallel. These results demonstrate the suitability of WALBERLA-WIND for exascale-classwind-farm simulations when high throughput and energy efficiency are essential.
- Open Access
- 28 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Multi-Device Shallow Water Simulations on CPUs, GPUs, and FPGAs with SYCL
- 28 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-12
Open AccessMulti-Device Shallow Water Simulations on CPUs, GPUs, and FPGAs with SYCL
- Open Access
- 5 June 2026
- Published by SAGE Publications in The International Journal of High Performance Computing Applications
- Vol. 40 (6) , 866-881
Towards exascale simulations of granular fluidization in offshore wind turbine foundations
Fully-resolved micromechanical simulations coupling the Lattice Boltzmann Method (LBM) with the Discrete Element Method (DEM) provide high-fidelity insights into granular fluidization. However, the substantial computational demands of such simulations require efficient implementations on supercomputing architectures. This work presents a comprehensive performance analysis of a fully-resolved LBM-DEM model to study granular fluidization, i.e., piping erosion, during the installation of an offshore caisson foundation. The performance evaluation focuses on real-world workloads rather than simplified benchmark problems to obtain realistic performance insights. The study considers a diverse range of state-of-the-art HPC hardware architectures, namely the LUMI and MareNostrum 5 EuroHPC supercomputers, both CPU and GPU partitions. The results demonstrate that GPU-based systems generally outperform the CPU-based systems. Strong and weak scaling analyses were conducted with up to 512 nodes, and parallel efficiencies reached up to 92%. Nonetheless, the results also indicate that atomic add operations on GPUs can become a bottleneck for the parallel efficiency at large scales. Moreover, the study reveals a close link between physical variations of the setup and significant scaling implications, underscoring the need to consider physical model characteristics in scaling assessments.
- 5 June 2026
- Published by SAGE Publications in The International Journal of High Performance Computing Applications
- Vol. 40 (6) , 866-881
Open AccessTowards exascale simulations of granular fluidization in offshore wind turbine foundations
Fully-resolved micromechanical simulations coupling the Lattice Boltzmann Method (LBM) with the Discrete Element Method (DEM) provide high-fidelity insights into granular fluidization. However, the substantial computational demands of such simulations require efficient implementations on supercomputing architectures. This work presents a comprehensive performance analysis of a fully-resolved LBM-DEM model to study granular fluidization, i.e., piping erosion, during the installation of an offshore caisson foundation. The performance evaluation focuses on real-world workloads rather than simplified benchmark problems to obtain realistic performance insights. The study considers a diverse range of state-of-the-art HPC hardware architectures, namely the LUMI and MareNostrum 5 EuroHPC supercomputers, both CPU and GPU partitions. The results demonstrate that GPU-based systems generally outperform the CPU-based systems. Strong and weak scaling analyses were conducted with up to 512 nodes, and parallel efficiencies reached up to 92%. Nonetheless, the results also indicate that atomic add operations on GPUs can become a bottleneck for the parallel efficiency at large scales. Moreover, the study reveals a close link between physical variations of the setup and significant scaling implications, underscoring the need to consider physical model characteristics in scaling assessments.
- Open Access
- 19 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 163-171
The Energy-Oriented Centre of Excellence
EoCoE is a EuroHPC Centre of Excellence contributing to Europe transition to sustainable energy through High Performance Computing. Now in its third phase, the project harnesses exascale technologies, massively parallel simulations, and advanced computational methods to tackle some of the most pressing challenges in renewable energy. At the heart of EoCoE are five lighthouse applications spanning four strategic domains: fusion, photovoltaics, water, and wind energy. By enhancing these applications, making them exascale-ready, and developing and optimising the underlying software tools and mathematical libraries, the project contributes to both the green energy transition and the advancement of scientific computing methodologies in Europe. This paper presents the five lighthouse applications and highlights the key achievements in performance, portability, and productivity. It further describes the development of enabling tools and libraries, and outlines the project broader efforts to maximise scientific, technological, and societal impact.
- 19 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 163-171
Open AccessThe Energy-Oriented Centre of Excellence
EoCoE is a EuroHPC Centre of Excellence contributing to Europe transition to sustainable energy through High Performance Computing. Now in its third phase, the project harnesses exascale technologies, massively parallel simulations, and advanced computational methods to tackle some of the most pressing challenges in renewable energy. At the heart of EoCoE are five lighthouse applications spanning four strategic domains: fusion, photovoltaics, water, and wind energy. By enhancing these applications, making them exascale-ready, and developing and optimising the underlying software tools and mathematical libraries, the project contributes to both the green energy transition and the advancement of scientific computing methodologies in Europe. This paper presents the five lighthouse applications and highlights the key achievements in performance, portability, and productivity. It further describes the development of enabling tools and libraries, and outlines the project broader efforts to maximise scientific, technological, and societal impact.
- Open Access
- 13 May 2026
- Published by The Open Journal in Journal of Open Source Software
- Vol. 11 (121)
MaterForge: Materials Formulation Engine with Python
- 13 May 2026
- Published by The Open Journal in Journal of Open Source Software
- Vol. 11 (121)
Open AccessMaterForge: Materials Formulation Engine with Python
- 1 April 2026
- Published by Springer Nature
- p. 213-222
Patch-Level Brain Tumor Sub-region Classification Using Foundation Models Under Long-Tailed Data Distributions
- 1 April 2026
- Published by Springer Nature
- p. 213-222
Patch-Level Brain Tumor Sub-region Classification Using Foundation Models Under Long-Tailed Data Distributions
- Open Access
PMRT: A Training Recipe for Fast, 3D High-Resolution Aerodynamic Prediction
The aerodynamic optimization of cars requires close collaboration between aerodynamicists and stylists, while slow, expensive simulations remain a bottleneck. Surrogate models have been shown to accurately predict aerodynamics within the design space for which they were trained. However, many of these models struggle to scale to higher resolutions because of the 3D nature of the problem and data scarcity. We propose Progressive Multi-Resolution Training (PMRT), a probabilistic multi-resolution training schedule that enables training a U-Net to predict the drag coefficient ($c_d$) and high-resolution velocity fields (512 x 128 x 128) in 24 hours on a single NVIDIA H100 GPU, 7x cheaper than the high-resolution-only baseline, with similar accuracy. PMRT samples batches from three resolutions based on probabilities that change during training, starting with an emphasis on lower resolutions and gradually shifting toward higher resolutions. Since this is a training methodology, it can be adapted to other high-resolution-focused backbones. We also show that a single model can be trained across five datasets from different solvers, including a real-world dataset, by conditioning on the simulation parameters. In the DrivAerML dataset, our models achieve a $c_d$ $R^2$ of 0.975, matching literature baselines at a fraction of the training cost.Open AccessPMRT: A Training Recipe for Fast, 3D High-Resolution Aerodynamic Prediction
The aerodynamic optimization of cars requires close collaboration between aerodynamicists and stylists, while slow, expensive simulations remain a bottleneck. Surrogate models have been shown to accurately predict aerodynamics within the design space for which they were trained. However, many of these models struggle to scale to higher resolutions because of the 3D nature of the problem and data scarcity. We propose Progressive Multi-Resolution Training (PMRT), a probabilistic multi-resolution training schedule that enables training a U-Net to predict the drag coefficient ($c_d$) and high-resolution velocity fields (512 x 128 x 128) in 24 hours on a single NVIDIA H100 GPU, 7x cheaper than the high-resolution-only baseline, with similar accuracy. PMRT samples batches from three resolutions based on probabilities that change during training, starting with an emphasis on lower resolutions and gradually shifting toward higher resolutions. Since this is a training methodology, it can be adapted to other high-resolution-focused backbones. We also show that a single model can be trained across five datasets from different solvers, including a real-world dataset, by conditioning on the simulation parameters. In the DrivAerML dataset, our models achieve a $c_d$ $R^2$ of 0.975, matching literature baselines at a fraction of the training cost. - Open Access
ChannelFlow-Tools: A Standardized Dataset Creation Pipeline for 3D Obstructed Channel Flows
We present ChannelFlow-Tools, a configuration-driven framework that standardizes the end-to-end path from programmatic CAD solid generation to ML-ready inputs and targets for 3D obstructed channel flows. The toolchain integrates geometry synthesis with feasibility checks, signed distance field (SDF) voxelization, automated solver orchestration on HPC (waLBerla LBM), and Cartesian resampling to co-registered multi-resolution tensors. A single Hydra/OmegaConf configuration governs all stages, enabling deterministic reproduction and controlled ablations. As a case study, we generate 10k+ scenes spanning Re=100-15000 with diverse shapes and poses. An end-to-end evaluation of storage trade-offs directly from the emitted artifacts, a minimal 3D U-Net at 128x32x32, and example surrogate models with dataset size illustrate that the standardized representations support reproducible ML training. ChannelFlow-Tools turns one-off dataset creation into a reproducible, configurable pipeline for CFD surrogate modeling.
Open AccessChannelFlow-Tools: A Standardized Dataset Creation Pipeline for 3D Obstructed Channel Flows
We present ChannelFlow-Tools, a configuration-driven framework that standardizes the end-to-end path from programmatic CAD solid generation to ML-ready inputs and targets for 3D obstructed channel flows. The toolchain integrates geometry synthesis with feasibility checks, signed distance field (SDF) voxelization, automated solver orchestration on HPC (waLBerla LBM), and Cartesian resampling to co-registered multi-resolution tensors. A single Hydra/OmegaConf configuration governs all stages, enabling deterministic reproduction and controlled ablations. As a case study, we generate 10k+ scenes spanning Re=100-15000 with diverse shapes and poses. An end-to-end evaluation of storage trade-offs directly from the emitted artifacts, a minimal 3D U-Net at 128x32x32, and example surrogate models with dataset size illustrate that the standardized representations support reproducible ML training. ChannelFlow-Tools turns one-off dataset creation into a reproducible, configurable pipeline for CFD surrogate modeling.