arXiv subjects

Christos E. Frouzakis

Publications and source records attributed to Christos E. Frouzakis.

8 recordsLinked to original sources

arXiv2026

Super-resolution of turbulent reacting flows on complex meshes using graph neural networks

State-of-the-art deep learning models have been extensively utilized to reconstruct small-scale structures from coarse-grained data in turbulent flows. However, their application has predominantly been restricted to structured uniform meshes, limiting their applicability to data associated with complex geometries that are typically simulated on structured non-uniform or unstructured meshes. Machine learning (ML) models based on graph neural networks (GNNs), known for their ability to process unstructured data, offer a promising alternative. In this study, we leverage the inherent flexibility of GNNs featuring message passing layers to develop a methodology for reconstructing unresolved small-scale structures from low-resolution data on complex meshes. The accuracy of the proposed approach is demonstrated using two cases: a reacting channel flow on a structured non-uniform mesh, and a reacting hydrogen fueled internal combustion (IC) engine featuring an unstructured mesh. Evaluation of results based on visual agreement, statistical metrics, and cumulative error reduction indicates the effectiveness of the method in accurately reconstructing fine-scale features. Overall, this study provides a pathway for integrating data-driven small-scale reconstruction and subgrid-scale modeling to enhance the accuracy of coarse-grained simulations on complex meshes.

physics.flu-dyn↗

arXiv2025

Analysis of In-cylinder Flow Structures and Turbulence in a Laboratory Scale Engine using Direct Numerical Simulations

In-cylinder flow structures and turbulence characteristics are investigated using direct numerical simulations (DNS) in a laboratory-scale engine at technically relevant engine speeds (1500 and 2500 rpm at full load). The data is computed for 12 compression-expansion cycles at each engine speed with initial conditions derived from precursor large eddy simulations (LES) validated against experimental data. Analysis of the tumble ratio indicates significant cycle-to-cycle variation, with lower variability found at higher engine speed. The process of tumble breakdown, quantified by the evolution of mean and turbulent kinetic energy, reveals distinct features between operating conditions, with delayed turbulence peaks observed at lower engine speed. Analysis of the Reynolds stress tensor demonstrates higher stress values at higher engine speed, with pronounced anisotropy near the walls and higher values around the tumble vortex core. Examination of the anisotropic Reynolds stress invariants through Lumley triangles reveals predominantly isotropic turbulence during mid-compression, transitioning to distinct anisotropic states near top dead center (TDC). The lower engine speed exhibits a stronger tendency toward one-component turbulence due to partial tumble dissipation, while the higher speed maintains a more balanced anisotropy. These findings extend previous numerical and experimental studies on turbulence development and in-cylinder flow structures during the compression stroke, providing insights for improving turbulence modeling in practical engine simulations.

physics.flu-dyn↗

arXiv2025

Direct Numerical Simulation of Hydrogen Combustion in a Real-Size IC Engine

This study presents the first Direct Numerical Simulation (DNS) of hydrogen combustion in a real-size internal combustion engine, investigating the complex dynamics of ignition, flame propagation, and flame-wall interaction under engine-relevant conditions. The simulation focuses on ultra-lean hydrogen operation at equivalence ratio $ϕ=0.4$ and 800 rpm, utilizing a state-of-the-art spectral element solver optimized for GPU architectures. The computational domain encompasses the full engine geometry. Results highlight the strong coupling between the flame dynamics and the coherent flow structures during early flame kernel development, while differential diffusion effects lead to increased reactivity at positive flame curvatures, a phenomenon that has only been studied in canonical configurations of freely propagating hydrogen/air flames. As the flame approaches the walls, distinct behavior is observed during head-on and side-wall quenching scenarios, characterized by different spatial distributions of wall heat flux. The findings provide insights into hydrogen combustion in real engines, essential for the development of clean and efficient hydrogen-fueled powertrains.

physics.flu-dyn↗

arXiv2025

KinetiX: A performance portable code generator for chemical kinetics and transport properties

We present KinetiX, a software toolkit to generate computationally efficient fuel-specific routines for the chemical source term, thermodynamic and mixture-averaged transport properties for use in combustion simulation codes. The C++ routines are designed for high-performance execution on both CPU and GPU architectures. On CPUs, chemical kinetics computations are optimized by eliminating redundant operations and using data alignment and loops with trivial access patterns that enable auto-vectorization, reducing the latency of complex mathematical operations. On GPUs, performance is improved by loop unrolling, reducing the number of costly exponential evaluations and limiting the number of live variables for better register usage. The accuracy of the generated routines is checked against reference values computed using Cantera and the maximum relative errors are below $10^{-7}$. We evaluate the performance of the kernels on some of the latest CPU and GPU architectures from AMD and NVIDIA, i.e., AMD EPYC 9653, AMD MI250X, and NVIDIA H100. The routines generated by KinetiX outperform the general-purpose Cantera library, achieving speedups of up to 2.4x for species production rates and 3.2x for mixture-averaged transport properties on CPUs. Compared to the routines generated by PelePhysics (CEPTR), KinetiX achieves speedups of up to 2.6x on CPUs and 1.7x on GPUs for the species production rates kernel on a single-threaded basis.

physics.comp-ph↗

arXiv2025

Deciphering boundary layer dynamics in high-Rayleigh-number convection using 3360 GPUs and a high-scaling in-situ workflow

Turbulent heat and momentum transfer processes due to thermal convection cover many scales and are of great importance for several natural and technical flows. One consequence is that a fully resolved three-dimensional analysis of these turbulent transfers at high Rayleigh numbers, which includes the boundary layers, is possible only using supercomputers. The visualization of these dynamics poses an additional hurdle since the thermal and viscous boundary layers in thermal convection fluctuate strongly. In order to track these fluctuations continuously, data must be tapped at high frequency for visualization, which is difficult to achieve using conventional methods. This paper makes two main contributions in this context. First, it discusses the simulations of turbulent Rayleigh-Bénard convection up to Rayleigh numbers of $Ra=10^{12}$ computed with NekRS on GPUs. The largest simulation was run on 840 nodes with 3360 GPU on the JUWELS Booster supercomputer. Secondly, an in-situ workflow using ASCENT is presented, which was successfully used to visualize the high-frequency turbulent fluctuations.

physics.flu-dyn↗

arXiv2024

Flow reconstruction in time-varying geometries using graph neural networks

The paper presents a Graph Attention Convolutional Network (GACN) for flow reconstruction from very sparse data in time-varying geometries. The model incorporates a feature propagation algorithm as a preprocessing step to handle extremely sparse inputs, leveraging information from neighboring nodes to initialize missing features. In addition, a binary indicator is introduced as a validity mask to distinguish between the original and propagated data points, enabling more effective learning from sparse inputs. Trained on a unique data set of Direct Numerical Simulations (DNS) of a motored engine at a technically relevant operating condition, the GACN shows robust performance across different resolutions and domain sizes and can effectively handle unstructured data and variable input sizes. The model is tested on previously unseen DNS data as well as on an experimental data set from Particle Image Velocimetry (PIV) measurements that were not considered during training. A comparative analysis shows that the GACN consistently outperforms both a conventional Convolutional Neural Network (CNN) and cubic interpolation methods on the DNS and PIV test sets by achieving lower reconstruction errors and better capturing fine-scale turbulent structures. In particular, the GACN effectively reconstructs flow fields from domains up to 14 times larger than those observed during training, with the performance advantage increasing for larger domains.

cs.LG↗

arXiv2024

nekCRF: A next generation high-order reactive low Mach flow solver for direct numerical simulations

Exascale computing enables high-fidelity simulations of chemically reactive flows in practical geometries and conditions, and paves the way for valuable insights that can optimize combustion processes, ultimately reducing emissions and improving fuel combustion efficiency. However, this requires software that can fully leverage the capabilities of current high performance computing systems. The paper introduces nekCRF, a high-order reactive low Mach flow solver specifically designed for this purpose. Its capabilities and efficiency are showcased on the pre-exascale system JUWELS Booster, a GPU-based supercomputer at the Jülich Supercomputing Centre including a validation across diverse cases of varying complexity.

physics.comp-ph↗

arXiv2003

A route to computational chaos revisited: noninvertibility and the breakup of an invariant circle

In a one-parameter study of a noninvertible family of maps of the plane arising in the context of a numerical integration scheme, Lorenz studied a sequence of transitions from an attracting fixed point to "computational chaos." As part of the transition sequence, he proposed the following as a possible scenario for the breakup of an invariant circle: the invariant circle develops regions of increasingly sharper curvature until at a critical parameter value it develops cusps; beyond this parameter value, the invariant circle fails to persist, and the system exhibits chaotic behavior on an invariant set with loops [Lorenz, 1989]. We investigate this problem in more detail and show that the invariant circle is actually destroyed in a global bifurcation before it has a chance to develop cusps. Instead, the global unstable manifolds of saddle-type periodic points are the objects which develop cusps and subsequently "loops" or "antennae." The one-parameter study is better understood when embedded in the full two-parameter space and viewed in the context of the two-parameter Arnold horn structure. Certain elements of the interplay of noninvertibility with this structure, the associated invariant circles, periodic points and global bifurcations are examined.

math.DS↗