Axel Coussement - Academia.edu (original) (raw)

Uploads

Papers by Axel Coussement

Research paper thumbnail of Nox Formation in Hydrogen-Rich Fuel Blends (H2/Ch4/Co) Flames Doped with Aromatic Additive in a Semi-Industrial Scale Furnace

Research paper thumbnail of Reduced-Order Modeling of Reacting Flows Using Data-Driven Approaches

Lecture notes in energy, 2023

Data-driven modeling of complex dynamical systems is becoming increasingly popular across various... more Data-driven modeling of complex dynamical systems is becoming increasingly popular across various domains of science and engineering. This is thanks to advances in numerical computing, which provides high fidelity data, and to algorithm development in data science and machine learning. Simulations of multicomponent reacting flows can particularly profit from data-based reduced-order modeling (ROM). The original system of coupled partial differential equations that describes a reacting flow is often large due to high number of chemical species involved. While the datasets from reacting flow simulation have high state-space dimensionality, they also exhibit attracting low-dimensional manifolds (LDMs). Data-driven approaches can be used to obtain and parameterize these LDMs. Evolving the reacting system using a smaller number of parameters can yield substantial model reduction and savings in computational cost. In this chapter, we review recent advances in ROM of turbulent reacting flows. We demonstrate the entire ROM workflow with a particular focus on obtaining the training datasets and data science and machine learning techniques such as dimensionality reduction and nonlinear regression. We present recent results from ROM-based simulations of experimentally measured Sandia flames D and F. We also delineate a few remaining challenges and possible future directions to address them. This chapter is accompanied by illustrative examples using the recently developed Python software, PCAfold. The software can be used to obtain, analyze and improve low-dimensional data representations. The examples provided herein can be helpful to students and researchers learning to apply dimensional

Research paper thumbnail of Reduced-order modeling of supersonic fuel–air mixing in a multi-strut injection scramjet engine using machine learning techniques

Research paper thumbnail of Mini-symposium on Safety related ignition processes

Research paper thumbnail of Local manifold learning and its link to domain-based physics knowledge

arXiv (Cornell University), Jul 1, 2022

In many reacting flow systems, the thermo-chemical state-space is known or assumed to evolve clos... more In many reacting flow systems, the thermo-chemical state-space is known or assumed to evolve close to a low-dimensional manifold (LDM). Various approaches are available to obtain those manifolds and subsequently express the original high-dimensional space with fewer parameterizing variables. Principal component analysis (PCA) is one of the dimensionality reduction methods that can be used to obtain LDMs. PCA does not make prior assumptions about the parameterizing variables and retrieves them empirically from the training data. In this paper, we show that PCA applied in local clusters of data (local PCA) is capable of detecting the intrinsic parameterization of the thermo-chemical state-space. We first demonstrate that utilizing three common combustion models of varying complexity: the Burke-Schumann model, the chemical equilibrium model and the homogeneous reactor. Parameterization of these models is known a priori which allows for benchmarking with the local PCA approach. We further extend the application of local PCA to a more challenging case of a turbulent non-premixed-heptane/air jet flame for which the parameterization is no longer obvious. Our results suggest that meaningful parameterization can be obtained also for more complex datasets. We show that local PCA finds variables that can be linked to local stoichiometry, reaction progress and soot formation processes.

Research paper thumbnail of Adaptive digital twins of combustion systems using sparse sensing strategies

Proceedings of the Combustion Institute

This work proposes to implement a sparse sensing framework to build a hybrid numerical-experiment... more This work proposes to implement a sparse sensing framework to build a hybrid numerical-experimental Digital Twin of a practical combustion system. The goal is to find the optimal sensor placement that minimizes the prediction error, and to predict the distribution of reacting scalars using few measurements. Three-dimensional CFD simulations with detailed chemistry were used to build the design space by varying the fuel composition (from pure methane to pure hydrogen), the equivalence ratio (from 0.7 to 1) and the air velocity. The Proper Orthogonal Decomposition (POD) was applied to the numerical data to find a tailored basis for dimensionality reduction. Then, the QR decomposition with column pivoting was applied to the tailored basis to find the optimal sensor placement. Finally, the model was employed to predict the three-dimensional temperature distribution in the unexplored part of the design space, using the experimental samples as input. The optimal placement of the sensors provides valuable information on the key locations and features, which can then be used in the design of reactor network models, for example. Also, the results show that the hybrid Digital Twin could predict an adjusted temperature distribution which reduces the error with the experimental measurements, when compared to the original CFD temperature distribution.

Research paper thumbnail of Dimensionality reduction and unsupervised classification for high-fidelity reacting flow simulations

Proceedings of the Combustion Institute

Research paper thumbnail of Principal component analysis based combustion model in the context of a lifted methane/air flame: Sensitivity to the manifold parameters and subgrid closure

Research paper thumbnail of Predicting the composition of unconventional fuels

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of NO FORMATION IN PREMIXED FLAT H2/CH4/CO/O2/N2 and H2/CH4/CO/O2/N2/C6H6 FLAMES AT LOW PRESSURE

Research paper thumbnail of Nitrogen Oxides Emission Minimization of Coke Batteries

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Three-dimensional boundary conditions for reactive compressible viscous flows

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Reduced‐order modelling of turbulent reacting flows

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Reduced-order modeling of reacting flows

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Multicomponent real gas 3-D-NSCBC for direct numerical simulation of reactive compressible viscous flows

Journal of Computational Physics, 2013

The topic of this paper is to propose an extension of the classical one-dimensional Navier-Stokes... more The topic of this paper is to propose an extension of the classical one-dimensional Navier-Stokes boundary conditions (1-D-NSCBC) for real gases initially developed by Okong'o and Bellan [1] to a 3-D-NSCBC formulation based on the work of Lodato et al. [2] and Coussement et al. [3]. All the differences due to the real gas formulation compared to the perfect gas formulation proposed in [3] are emphasized. A new way of determining the pressure relaxation coefficient is introduced for handling transcritical flows crossing the boundary. The real gas 3-D-NSCBC are then challenged on several test cases: a two-dimensional subsonic vortex convection, a subsonic supercritical bubble convection and a flame vortex interaction. All these test cases are performed by direct numerical simulation of multicomponent flows. It shows the stability of the boundary conditions without creating any numerical artifact.

Research paper thumbnail of Local manifold learning and its link to domain-based physics knowledge

Applications in Energy and Combustion Science

Research paper thumbnail of Digital Twin for Experimental Data Fusion Applied to a Semi-Industrial Furnace Fed with H2-Rich Fuel Mixtures

Energies

The objective of this work is to build a Digital Twin of a semi-industrial furnace using Gaussian... more The objective of this work is to build a Digital Twin of a semi-industrial furnace using Gaussian Process Regression coupled with dimensionality reduction via Proper Orthogonal Decomposition. The Digital Twin is capable of integrating different sources of information, such as temperature, chemiluminescence intensity and species concentration at the outlet. The parameters selected to build the design space are the equivalence ratio and the benzene concentration in the fuel stream. The fuel consists of a H2/CH4/CO blend, doped with a progressive addition of C6H6. It is an H2-rich fuel mixture, representing a surrogate of a more complex Coke Oven Gas industrial mixture. The experimental measurements include the flame temperature distribution, measured on a 6×8 grid using an air-cooled suction pyrometer, spatially resolved chemiluminescence measurements of OH* and CH*, and the species concentration (i.e., NO, NO2, CO, H2O, CO2, O2) measured in the exhaust gases. The GPR-based Digital Tw...

Research paper thumbnail of Decarbonisation potential of dimethyl ether/hydrogen mixtures in a flameless furnace: Reactive structures and pollutant emissions

International Journal of Hydrogen Energy

Research paper thumbnail of Characterization of unconventional fuel from car residue processing: uncertainty quantification of composition prediction

Research paper thumbnail of Evaluation of the non premixed Filtered Tabulated Chemistry for LES model on a turbulent non premixed piloted methane jet flame

The performance of the non premixed Filtered Tabulated Chemistry for LES (FTACLES) model is asses... more The performance of the non premixed Filtered Tabulated Chemistry for LES (FTACLES) model is assessed on the piloted partially premixed methane/air Sandia flame-D configuration. The non premixed FTACLES model has been developed taking into consideration non premixed combustion in LES context, with either fully resolved wrinkling or in laminar condition. One of the aspects that distinguishes it from the analogous premixed model is the fact that not only the closure of the filtered progress variable but as well of the filtered mixture fraction is addressed. It has been shown how following this approach the filtered flame structure and thickness can be adequately described as function of the filtered mixture fraction, the filtered progress variable and the filter size.

Research paper thumbnail of Nox Formation in Hydrogen-Rich Fuel Blends (H2/Ch4/Co) Flames Doped with Aromatic Additive in a Semi-Industrial Scale Furnace

Research paper thumbnail of Reduced-Order Modeling of Reacting Flows Using Data-Driven Approaches

Lecture notes in energy, 2023

Data-driven modeling of complex dynamical systems is becoming increasingly popular across various... more Data-driven modeling of complex dynamical systems is becoming increasingly popular across various domains of science and engineering. This is thanks to advances in numerical computing, which provides high fidelity data, and to algorithm development in data science and machine learning. Simulations of multicomponent reacting flows can particularly profit from data-based reduced-order modeling (ROM). The original system of coupled partial differential equations that describes a reacting flow is often large due to high number of chemical species involved. While the datasets from reacting flow simulation have high state-space dimensionality, they also exhibit attracting low-dimensional manifolds (LDMs). Data-driven approaches can be used to obtain and parameterize these LDMs. Evolving the reacting system using a smaller number of parameters can yield substantial model reduction and savings in computational cost. In this chapter, we review recent advances in ROM of turbulent reacting flows. We demonstrate the entire ROM workflow with a particular focus on obtaining the training datasets and data science and machine learning techniques such as dimensionality reduction and nonlinear regression. We present recent results from ROM-based simulations of experimentally measured Sandia flames D and F. We also delineate a few remaining challenges and possible future directions to address them. This chapter is accompanied by illustrative examples using the recently developed Python software, PCAfold. The software can be used to obtain, analyze and improve low-dimensional data representations. The examples provided herein can be helpful to students and researchers learning to apply dimensional

Research paper thumbnail of Reduced-order modeling of supersonic fuel–air mixing in a multi-strut injection scramjet engine using machine learning techniques

Research paper thumbnail of Mini-symposium on Safety related ignition processes

Research paper thumbnail of Local manifold learning and its link to domain-based physics knowledge

arXiv (Cornell University), Jul 1, 2022

In many reacting flow systems, the thermo-chemical state-space is known or assumed to evolve clos... more In many reacting flow systems, the thermo-chemical state-space is known or assumed to evolve close to a low-dimensional manifold (LDM). Various approaches are available to obtain those manifolds and subsequently express the original high-dimensional space with fewer parameterizing variables. Principal component analysis (PCA) is one of the dimensionality reduction methods that can be used to obtain LDMs. PCA does not make prior assumptions about the parameterizing variables and retrieves them empirically from the training data. In this paper, we show that PCA applied in local clusters of data (local PCA) is capable of detecting the intrinsic parameterization of the thermo-chemical state-space. We first demonstrate that utilizing three common combustion models of varying complexity: the Burke-Schumann model, the chemical equilibrium model and the homogeneous reactor. Parameterization of these models is known a priori which allows for benchmarking with the local PCA approach. We further extend the application of local PCA to a more challenging case of a turbulent non-premixed-heptane/air jet flame for which the parameterization is no longer obvious. Our results suggest that meaningful parameterization can be obtained also for more complex datasets. We show that local PCA finds variables that can be linked to local stoichiometry, reaction progress and soot formation processes.

Research paper thumbnail of Adaptive digital twins of combustion systems using sparse sensing strategies

Proceedings of the Combustion Institute

This work proposes to implement a sparse sensing framework to build a hybrid numerical-experiment... more This work proposes to implement a sparse sensing framework to build a hybrid numerical-experimental Digital Twin of a practical combustion system. The goal is to find the optimal sensor placement that minimizes the prediction error, and to predict the distribution of reacting scalars using few measurements. Three-dimensional CFD simulations with detailed chemistry were used to build the design space by varying the fuel composition (from pure methane to pure hydrogen), the equivalence ratio (from 0.7 to 1) and the air velocity. The Proper Orthogonal Decomposition (POD) was applied to the numerical data to find a tailored basis for dimensionality reduction. Then, the QR decomposition with column pivoting was applied to the tailored basis to find the optimal sensor placement. Finally, the model was employed to predict the three-dimensional temperature distribution in the unexplored part of the design space, using the experimental samples as input. The optimal placement of the sensors provides valuable information on the key locations and features, which can then be used in the design of reactor network models, for example. Also, the results show that the hybrid Digital Twin could predict an adjusted temperature distribution which reduces the error with the experimental measurements, when compared to the original CFD temperature distribution.

Research paper thumbnail of Dimensionality reduction and unsupervised classification for high-fidelity reacting flow simulations

Proceedings of the Combustion Institute

Research paper thumbnail of Principal component analysis based combustion model in the context of a lifted methane/air flame: Sensitivity to the manifold parameters and subgrid closure

Research paper thumbnail of Predicting the composition of unconventional fuels

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of NO FORMATION IN PREMIXED FLAT H2/CH4/CO/O2/N2 and H2/CH4/CO/O2/N2/C6H6 FLAMES AT LOW PRESSURE

Research paper thumbnail of Nitrogen Oxides Emission Minimization of Coke Batteries

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Three-dimensional boundary conditions for reactive compressible viscous flows

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Reduced‐order modelling of turbulent reacting flows

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Reduced-order modeling of reacting flows

info:eu-repo/semantics/nonPublishe

Research paper thumbnail of Multicomponent real gas 3-D-NSCBC for direct numerical simulation of reactive compressible viscous flows

Journal of Computational Physics, 2013

The topic of this paper is to propose an extension of the classical one-dimensional Navier-Stokes... more The topic of this paper is to propose an extension of the classical one-dimensional Navier-Stokes boundary conditions (1-D-NSCBC) for real gases initially developed by Okong'o and Bellan [1] to a 3-D-NSCBC formulation based on the work of Lodato et al. [2] and Coussement et al. [3]. All the differences due to the real gas formulation compared to the perfect gas formulation proposed in [3] are emphasized. A new way of determining the pressure relaxation coefficient is introduced for handling transcritical flows crossing the boundary. The real gas 3-D-NSCBC are then challenged on several test cases: a two-dimensional subsonic vortex convection, a subsonic supercritical bubble convection and a flame vortex interaction. All these test cases are performed by direct numerical simulation of multicomponent flows. It shows the stability of the boundary conditions without creating any numerical artifact.

Research paper thumbnail of Local manifold learning and its link to domain-based physics knowledge

Applications in Energy and Combustion Science

Research paper thumbnail of Digital Twin for Experimental Data Fusion Applied to a Semi-Industrial Furnace Fed with H2-Rich Fuel Mixtures

Energies

The objective of this work is to build a Digital Twin of a semi-industrial furnace using Gaussian... more The objective of this work is to build a Digital Twin of a semi-industrial furnace using Gaussian Process Regression coupled with dimensionality reduction via Proper Orthogonal Decomposition. The Digital Twin is capable of integrating different sources of information, such as temperature, chemiluminescence intensity and species concentration at the outlet. The parameters selected to build the design space are the equivalence ratio and the benzene concentration in the fuel stream. The fuel consists of a H2/CH4/CO blend, doped with a progressive addition of C6H6. It is an H2-rich fuel mixture, representing a surrogate of a more complex Coke Oven Gas industrial mixture. The experimental measurements include the flame temperature distribution, measured on a 6×8 grid using an air-cooled suction pyrometer, spatially resolved chemiluminescence measurements of OH* and CH*, and the species concentration (i.e., NO, NO2, CO, H2O, CO2, O2) measured in the exhaust gases. The GPR-based Digital Tw...

Research paper thumbnail of Decarbonisation potential of dimethyl ether/hydrogen mixtures in a flameless furnace: Reactive structures and pollutant emissions

International Journal of Hydrogen Energy

Research paper thumbnail of Characterization of unconventional fuel from car residue processing: uncertainty quantification of composition prediction

Research paper thumbnail of Evaluation of the non premixed Filtered Tabulated Chemistry for LES model on a turbulent non premixed piloted methane jet flame

The performance of the non premixed Filtered Tabulated Chemistry for LES (FTACLES) model is asses... more The performance of the non premixed Filtered Tabulated Chemistry for LES (FTACLES) model is assessed on the piloted partially premixed methane/air Sandia flame-D configuration. The non premixed FTACLES model has been developed taking into consideration non premixed combustion in LES context, with either fully resolved wrinkling or in laminar condition. One of the aspects that distinguishes it from the analogous premixed model is the fact that not only the closure of the filtered progress variable but as well of the filtered mixture fraction is addressed. It has been shown how following this approach the filtered flame structure and thickness can be adequately described as function of the filtered mixture fraction, the filtered progress variable and the filter size.