arXiv subjects

Andreas Kronenburg

Publications and source records attributed to Andreas Kronenburg.

6 recordsLinked to original sources

arXiv2026

Large-eddy simulation of moderately dense evaporating sprays with particle-informed super-resolution

In large-eddy simulation (LES) of dense sprays or sprays with pronounced clustering, evaporation rates can be inaccurate when the mesh is too coarse to provide realistic boundary conditions for the widely employed single droplet evaporation model. This is especially relevant to liquid spray combustion in practical applications. Deep learning-based super-resolution (SR) has recently emerged as a promising method for LES subgrid-scale modeling, capable of enhancing flow field resolution. This technique appears well-suited to reconstruct the local gas fields within the inter-droplet space that can be used to correct the evaporation rates. However, it has not yet been applied for this purpose. This paper presents an innovative SR approach $-$ particle-informed super-resolution (PISR) $-$ that approximates high-resolution flow fields for improved evaporation computation. It is validated with a priori, a posteriori and generalization tests on moderately dense sprays. The results show that PISR-LES can closely replicate the evaporation rates computed in a carrier-phase direct numerical simulation (CP-DNS), significantly reducing the discrepancy in the fuel mass fraction field between LES and CP-DNS. Furthermore, the PISR model exhibits robust generalization to cases unseen in training when varying air temperature, droplet diameter, turbulent Reynolds number and spray pattern.

physics.flu-dyn↗

arXiv2026

A convolutional autoencoder and neural ODE surrogate modeling framework applied to transient counterflow flames

A novel convolutional autoencoder and neural ODE (CAE-NODE) framework is proposed for a reduced-order model (ROM) applied to transient 2D counterflow flames, as an extension of AE-NODE methods in homogeneous reactive systems to spatially resolved flows. The multidimensional thermochemical fields (256 x 256 grid, 21 variables) obtained from direct numerical simulations (DNS) are used in training the CAE, where convolutional layers learned the underlying spatial correlations, allowing the CAE to construct an unsupervised 3D latent manifold that is physically meaningful, smooth, and continuous in time. This results in a compression ratio of over 400,000 times. The NODE then subsequently learns the continuous-time dynamics on the latent manifold, enabling the prediction of the full temporal evolution of the flames by integrating forward in time from an initial condition. The results demonstrate that the CAE-NODE can accurately capture the entire transient process, including ignition, flame propagation, and the gradual transition to a non-premixed condition, with excellent agreement with the DNS, while adhering to conservation principles at virtually no computational cost compared to the reference DNS. Predictions remain accurate at strain rates outside the training range. Moreover, despite being unsupervised, the learned latent manifold is highly correlated with the flame-state descriptors such as the progress variable, mixture fraction, and the scalar dissipation rate. This study, for the first time, highlights the potential of CAE-NODE for surrogate modeling of unsteady dynamics of multi-dimensional reacting flows.

physics.flu-dyn↗

arXiv2025

Super-resolution of turbulent velocity fields in two-way coupled particle-laden flows

This paper introduces a deep learning-based super-resolution (SR) framework specifically developed for accurately reconstructing high-resolution velocity fields in two-way coupled particle-laden turbulent flows. Leveraging conditional generative adversarial networks (cGANs), the generator network architecture incorporates explicit conditioning on physical parameters, such as effective particle mass density and subgrid kinetic energy, while the discriminator network is conditioned on low-resolution data as well as high-frequency content of the input data. High-fidelity direct numerical simulation (DNS) datasets, covering a range of particle Stokes numbers, particle mass loadings, and carrier gas turbulence regimes, including forced- and decaying-turbulence, serve as training and testing datasets. Extensive validation studies, including detailed analyses of energy spectra, probability density functions (PDFs), vorticity distributions, and wavelet-based decomposition demonstrate the model's accuracy and generalization capabilities across different particle parameters. The results show that the network utilizes particle data, mainly in the reconstruction of high-frequency details modulated by particles. Additionally, systematic assessment of the model's performance in capturing previously unseen flow regimes further validates its predictive capabilities.

physics.flu-dyn↗

arXiv2025

Super-resolution of turbulent velocity and scalar fields using different scalar distributions

In recent years, sub-grid models for turbulent mixing have been developed by data-driven methods for large eddy simulation (LES). Super-resolution is a data-driven deconvolution technique in which deep convolutional neural networks are trained using direct numerical simulation (DNS) data to learn mappings between the input data from a low resolution domain to the super-resolved high resolution output domain. While the technique has been of a great success in a-priori tests, the assessment of its generalization capabilities is required for further a-posteriori applications. In this study we assess the generalization capability of a super-resolution generative adversarial network (GAN) in reconstructing scalars with different distributions. Forced turbulence mixing DNS data with a fixed Reynolds number but different bulk scalar distributions, are generated and used as training/testing datasets. The results show that the velocity vector field can be reconstructed well, but the model fails to super-resolve the scalars from out-of-sample distributions. Including two extreme mixture fraction distributions, namely double Pareto and semi-Gaussian, in the training dataset significantly improves the performance of the model, not only for those distributions, but also for previously unseen bimodal distributions.

physics.flu-dyn↗

arXiv2024

Carrier-phase DNS of ignition and combustion of iron particles in a turbulent mixing layer

Three-dimensional CP-DNS of reacting iron particle dust clouds in a turbulent mixing layer are conducted. The simulation approach considers the Eulerian transport equations for the reacting gas phase and resolves all scales of turbulence, whereas the particle boundary layers are modelled employing the Lagrangian point-particle framework for the dispersed phase. The CP-DNS employs an existing sub-model for iron particle combustion that considers the oxidation of iron to FeO and that accounts for both diffusion- and kinetically-limited combustion. At first, the particle sub-model is validated against experimental results for single iron particle combustion considering various particle diameters and ambient oxygen concentrations. Subsequently, the CP-DNS approach is employed to predict iron particle cloud ignition and combustion in a turbulent mixing layer. The upper stream of the mixing layer is initialised with cold particles in air, while the lower stream consists of hot air flowing in the opposite direction. Simulation results show that turbulent mixing induces heating, ignition and combustion of the iron particles. Significant increases in gas temperature and oxygen consumption occur mainly in regions where clusters of iron particles are formed. Over the course of the oxidation, the particles are subjected to different rate-limiting processes. While initially particle oxidation is kinetically-limited it becomes diffusion-limited for higher particle temperatures and peak particle temperatures are observed near the fully-oxidised particle state. Comparing the present non-volatile iron dust flames to general trends in volatile-containing solid fuel flames, non-vanishing particles at late simulation times and a stronger limiting effect of the local oxygen concentration on particle conversion is found for the present iron dust flames in shear-driven turbulence.

physics.flu-dyn↗

arXiv2014

Modelling nano-particle agglomeration using local interactions

Nano-particle agglomeration plays an important role in processes such as spray drying and particle flame synthesis. These processes have in common that nano-particles collide at low concentrations and get irreversibly linked at the point of contact due to plastic deformation. In this paper, we investigate several models of irreversible connections, which require only local interactions between the colliding nano-particles and thus allow for scalable simulations. The models investigated here connect the particles upon collision by non-bonded strongly attractive interactions, bonded interactions or by binding agents placed at the point of contact. Models using spherically symmetric interactions form compact agglomerates and are therefore unsuitable to study agglomeration. In contrast, models that are either based on both central and angular potentials (type one) or on binding agents (type two) efficiently prevent restructuring of the agglomerates, and are therefore useful for modeling contacts formed by plastic deformation. Moreover, both types of models allow to control the rigidity and by that the degree of restructuring. The first type of model is computationally more efficient at low fractional dimensions of the aggregates, while the second gives easy access to local shear forces, which is important when breaking of agglomerates is to be considered. As example applications, we reproduce the well-known diffusion-limited agglomeration (DLA) and report results on soot aggregation.

cond-mat.soft↗