Lukas Berger
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Lukas Berger has a total of 75 co-authors. Here are the top co-authors by number of shared publications:
- Heinz Pitsch
RWTH Aachen University
28 shared publications - Antonio Attili
University of Edinburgh
18 shared publications - Pasquale Eduardo Lapenna
Department of Mechanical and Aerospace Engineering, Sapienza University of Rome, Rome, Italy
9 shared publications - Francesco Creta
Department of Mechanical and Aerospace Engineering, Sapienza University of Rome, Rome, Italy
8 shared publications Heinz Pitsch
8 shared publicationsKonstantin Kleinheinz
8 shared publications
Latest Publications
- Open Access
- 16 October 2026
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 42
Uncovering flame physics with machine learning: Application to the reaction rate in hydrogen flames
- 16 October 2026
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 42
Open AccessUncovering flame physics with machine learning: Application to the reaction rate in hydrogen flames
- 12 February 2026
- Published by Elsevier in Combustion and Flame
- Vol. 286
Generalizable deep learning for sub-filter PDF of thermo-diffusively unstable lean hydrogen flames at varying Ka numbers
- 12 February 2026
- Published by Elsevier in Combustion and Flame
- Vol. 286
Generalizable deep learning for sub-filter PDF of thermo-diffusively unstable lean hydrogen flames at varying Ka numbers
- 1 May 2025
- Published by Springer Nature
- p. 105-156
Laminare Wasserstoffflammen
- Open Access
- 24 March 2025
- Published by Springer Nature in Flow, Turbulence and Combustion
- Vol. 115 (3) , 1261-1290
Extrapolation Performance of Convolutional Neural Network-Based Combustion Models for Large-Eddy Simulation: Influence of Reynolds Number, Filter Kernel and Filter Size
The extrapolation performance of Convolutional Neural Network (CNN)-based models for Large-Eddy Simulations (LES) has been investigated in the context of turbulent premixed combustion. The study utilises a series of Direct Numerical Simulation (DNS) datasets of turbulent premixed methane/air and hydrogen/air jet flames to train the CNN models. The methane/air flames, which are characterised by increasing Reynolds numbers, are used to model the subgrid-scale flame wrinkling. The hydrogen/air flame, exhibiting complex thermodiffusive instability, is employed to test the ability of the CNN-based combustion models to predict the filtered progress variable source term. This study focuses on the influence of varying training Reynolds numbers, filter sizes, and filter kernels to evaluate the performance of the CNN models to out-of-sample conditions, i.e., not seen during training. The objectives of this study are threefold: (i) analyse the performance of CNN models at different Reynolds numbers compared to the one trained with; (ii) analyse the performance of CNN models at different filter sizes compared to the one trained with; (iii) assess the influence of using different filter kernels (i.e., Gaussian and box filter kernels) between training and testing, to emulate a posteriori applications. The results demonstrate that the CNN models show good extrapolation performance when the training Reynolds number is sufficiently high. Vice versa, when CNN models are trained on low-Reynolds-number flame data, their performance degrades as they are applied to flames with progressively higher Reynolds numbers. When these CNN models are tested on datasets with filter sizes not included in the training process, they exhibit sufficient interpolation capabilities, the extrapolation performance is less precise but still satisfactory overall. This indicates that CNN models can be effectively trained using data filtered with a limited range of filter sizes and then successfully applied across a broader spectrum of filter sizes. Furthermore, when CNNs trained on box-filtered data are applied to Gaussian-filtered data, or vice versa, the models perform well for smaller filter sizes. However, as the filter size increases, the accuracy of the predictions diminishes. Interestingly, increasing the quantity of training data does not significantly enhance model performance. Yet, when training data are distributed with greater weighting towards larger filter sizes, the model’s overall performance improves. This suggests that the strategic selection and weighting of training data can lead to more robust generalization across different filter conditions.
- 24 March 2025
- Published by Springer Nature in Flow, Turbulence and Combustion
- Vol. 115 (3) , 1261-1290
Open AccessExtrapolation Performance of Convolutional Neural Network-Based Combustion Models for Large-Eddy Simulation: Influence of Reynolds Number, Filter Kernel and Filter Size
The extrapolation performance of Convolutional Neural Network (CNN)-based models for Large-Eddy Simulations (LES) has been investigated in the context of turbulent premixed combustion. The study utilises a series of Direct Numerical Simulation (DNS) datasets of turbulent premixed methane/air and hydrogen/air jet flames to train the CNN models. The methane/air flames, which are characterised by increasing Reynolds numbers, are used to model the subgrid-scale flame wrinkling. The hydrogen/air flame, exhibiting complex thermodiffusive instability, is employed to test the ability of the CNN-based combustion models to predict the filtered progress variable source term. This study focuses on the influence of varying training Reynolds numbers, filter sizes, and filter kernels to evaluate the performance of the CNN models to out-of-sample conditions, i.e., not seen during training. The objectives of this study are threefold: (i) analyse the performance of CNN models at different Reynolds numbers compared to the one trained with; (ii) analyse the performance of CNN models at different filter sizes compared to the one trained with; (iii) assess the influence of using different filter kernels (i.e., Gaussian and box filter kernels) between training and testing, to emulate a posteriori applications. The results demonstrate that the CNN models show good extrapolation performance when the training Reynolds number is sufficiently high. Vice versa, when CNN models are trained on low-Reynolds-number flame data, their performance degrades as they are applied to flames with progressively higher Reynolds numbers. When these CNN models are tested on datasets with filter sizes not included in the training process, they exhibit sufficient interpolation capabilities, the extrapolation performance is less precise but still satisfactory overall. This indicates that CNN models can be effectively trained using data filtered with a limited range of filter sizes and then successfully applied across a broader spectrum of filter sizes. Furthermore, when CNNs trained on box-filtered data are applied to Gaussian-filtered data, or vice versa, the models perform well for smaller filter sizes. However, as the filter size increases, the accuracy of the predictions diminishes. Interestingly, increasing the quantity of training data does not significantly enhance model performance. Yet, when training data are distributed with greater weighting towards larger filter sizes, the model’s overall performance improves. This suggests that the strategic selection and weighting of training data can lead to more robust generalization across different filter conditions.
- Open Access
- 5 February 2025
- Published by Elsevier in Combustion and Flame
- Vol. 274
Interactions of preferential diffusion and mixture inhomogeneities in hydrogen and iso-octane flame kernels under engine conditions
- 5 February 2025
- Published by Elsevier in Combustion and Flame
- Vol. 274
Open AccessInteractions of preferential diffusion and mixture inhomogeneities in hydrogen and iso-octane flame kernels under engine conditions
- Top Cited PapersOpen Access
- 24 January 2025
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1003, A33
LES combustion model for premixed turbulent hydrogen flames with thermodiffusive instabilities: a priori and a posteriori analysis
Premixed hydrogen flames are prone to thermodiffusive instabilities due to strong differential diffusion effects. Reproducing these instabilities in large eddy simulations (LES), where their effects are only partially resolved, is challenging. Combustion models that account for differential diffusion effects have been developed for laminar flames, but to use them in LES, models for the turbulence/flame subfilter interactions are required. Modelling of the subfilter interactions is particularly challenging as instabilities synergistically interact with turbulence resulting in a strong enhancement of the turbulent flame speed. In this work, a combustion model for LES, which accounts for thermodiffusive instabilities and their interactions with turbulence, is presented. In the first part, an a priori analysis based on a direct numerical simulation (DNS) of a turbulent hydrogen/air jet flame is discussed. Progress variable, progress variable variance and mixture fraction are rigorously identified as suitable model input parameters, and an LES combustion model based on pre-tabulated unstretched premixed flamelets with varying equivalence ratio is formulated. Subfilter closure is achieved via a presumed probability density function and a significant reduction of modelling errors is achieved with the presented model. In the second part, LES of the DNS configuration are performed for an a posteriori analysis. The presented combustion model shows significant improvements in predicting the flame length and local phenomena, such as super-adiabatic temperature, compared with combustion models that either neglect differential diffusion effects or consider these effects but neglect the subfilter closure. Two variants of the model formulation with a water- or hydrogen-based progress variable have been tested, yielding overall similar predictions.
- 24 January 2025
- Published by Cambridge University Press (CUP) in Journal of Fluid Mechanics
- Vol. 1003, A33
Top Cited PapersOpen AccessLES combustion model for premixed turbulent hydrogen flames with thermodiffusive instabilities: a priori and a posteriori analysis
Premixed hydrogen flames are prone to thermodiffusive instabilities due to strong differential diffusion effects. Reproducing these instabilities in large eddy simulations (LES), where their effects are only partially resolved, is challenging. Combustion models that account for differential diffusion effects have been developed for laminar flames, but to use them in LES, models for the turbulence/flame subfilter interactions are required. Modelling of the subfilter interactions is particularly challenging as instabilities synergistically interact with turbulence resulting in a strong enhancement of the turbulent flame speed. In this work, a combustion model for LES, which accounts for thermodiffusive instabilities and their interactions with turbulence, is presented. In the first part, an a priori analysis based on a direct numerical simulation (DNS) of a turbulent hydrogen/air jet flame is discussed. Progress variable, progress variable variance and mixture fraction are rigorously identified as suitable model input parameters, and an LES combustion model based on pre-tabulated unstretched premixed flamelets with varying equivalence ratio is formulated. Subfilter closure is achieved via a presumed probability density function and a significant reduction of modelling errors is achieved with the presented model. In the second part, LES of the DNS configuration are performed for an a posteriori analysis. The presented combustion model shows significant improvements in predicting the flame length and local phenomena, such as super-adiabatic temperature, compared with combustion models that either neglect differential diffusion effects or consider these effects but neglect the subfilter closure. Two variants of the model formulation with a water- or hydrogen-based progress variable have been tested, yielding overall similar predictions.
- Open Access
- 31 December 2024
- Published by Elsevier in Combustion and Flame
- Vol. 273
Comprehensive linear stability analysis for intrinsic instabilities in premixed ammonia/hydrogen/air flames
- 31 December 2024
- Published by Elsevier in Combustion and Flame
- Vol. 273
Open AccessComprehensive linear stability analysis for intrinsic instabilities in premixed ammonia/hydrogen/air flames
- Open Access
- 16 August 2024
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 40 (1-4)
Data-driven modeling of resolved and filtered thermo-diffusively unstable hydrogen–air flames
- 16 August 2024
- Published by Elsevier in Proceedings of the Combustion Institute
- Vol. 40 (1-4)
Open AccessData-driven modeling of resolved and filtered thermo-diffusively unstable hydrogen–air flames
- Open Access
- 9 August 2024
- Published by Elsevier in Combustion and Flame
- Vol. 269
Efficient and accurate calculation of dispersion relations for intrinsically unstable premixed flames
- 9 August 2024
- Published by Elsevier in Combustion and Flame
- Vol. 269
Open AccessEfficient and accurate calculation of dispersion relations for intrinsically unstable premixed flames