Super-grid Linear Eddy Model (SG-LEM): Efficient mode- and regime-independent combustion closure for Large Eddy Simulation (LES)
Doktorsavhandling, 2025
Practical combustion devices, such as ICEs (Internal Combustion Engines), often exhibit characteristics of premixed and non-premixed modes, as well as spatial variation in combustion regimes. Such ‘mixed-mode’/‘multi-regime’ scenarios are a challenge for combustion modelling as models often rely upon assumptions regarding the mode, as well as on the fast vs. slow nature of the underlying chemistry. Newer combustion technologies like ‘lean burn’ for gas turbines introduce further challenges such as ‘blowout’ and complex ignition sequences that require adequate description of turbulence/chemistry interaction (TCI) for accurate simulation. Hence, there is a great need for a mode- and regime-independent combustion model that accounts for TCI for the simulation of next-generation combustion devices. The Linear Eddy Model (LEM) is one that meets these criteria. When used with Large Eddy Simulation (LES), a technique called LES-LEM, it is able to simulate all the processes of reactive flow at their respective scales: large-and small-scale advection, molecular diffusion, and chemical reactions. LES-LEM is computationally intensive as each LES cell is embedded with a highly resolved LEM domain which requires numerical time-advancement, and so the technique finds limited use in industrial simulations.
This dissertation introduces a novel variant of LES-LEM that uses coarse-graining of the LES mesh, and resulting down-scaling of the number of embedded LEM domains, to significantly reduce compute times while still producing LES quality temperature and concentration fields by means of a presumed PDF (probability density function) mapping closure. Large-scale transport between neighbouring LEM domains is simulated through a novel Lagrangian ‘splicing’ scheme. The method, termed super-grid LES-LEM (or SG-LEM), is validated for three distinct flame cases: a premixed backward-facing case, the well-studied Volvo Validation Rig, and a mixed-mode flame produced by the Darmstadt burner. The work also identifies key drawbacks resulting from the effects of coarse-graining on large-scale transport. However, mitigation methods are introduced and tested which also leads to suggestions for future implementations. Overall, SG-LEM is a significant step in the practical application of LES-LEM, and its many benefits, to real-world technical flames.
This dissertation introduces a novel variant of LES-LEM that uses coarse-graining of the LES mesh, and resulting down-scaling of the number of embedded LEM domains, to significantly reduce compute times while still producing LES quality temperature and concentration fields by means of a presumed PDF (probability density function) mapping closure. Large-scale transport between neighbouring LEM domains is simulated through a novel Lagrangian ‘splicing’ scheme. The method, termed super-grid LES-LEM (or SG-LEM), is validated for three distinct flame cases: a premixed backward-facing case, the well-studied Volvo Validation Rig, and a mixed-mode flame produced by the Darmstadt burner. The work also identifies key drawbacks resulting from the effects of coarse-graining on large-scale transport. However, mitigation methods are introduced and tested which also leads to suggestions for future implementations. Overall, SG-LEM is a significant step in the practical application of LES-LEM, and its many benefits, to real-world technical flames.
Lagrangian splicing
mesh coarse-graining
mapping closure
mixed-mode flames
Linear Eddy Model
Large Eddy Simulation
HC1, Hörsalsvägen
Opponent: Dr David Lignell, Brigham Young University, USA
Författare
Abhilash Murlidharan Menon
Chalmers, Mekanik och maritima vetenskaper, Energiomvandling och framdrivningssystem
A super-grid approach for LES combustion closure using the Linear Eddy Model
Combustion Theory and Modelling,;Vol. 28(2024)p. 99-126
Artikel i vetenskaplig tidskrift
Investigation of coarse-graining parameters for super-grid LEM closure applied to LES of practical bluff-body flames
Combustion Theory and Modelling,;Vol. In Press(2024)
Artikel i vetenskaplig tidskrift
Assessing the Multi-Regime Capability of the Super-Grid Linear Eddy Model (SG-LEM) Using the Darmstadt Multi-Regime Burner
Flow, Turbulence and Combustion,;Vol. In Press(2024)
Artikel i vetenskaplig tidskrift
The modern world is powered by combustion devices such as internal combustion engines (ICEs), gas turbines and furnaces. They convert the chemical energy stored in fuels into useful forms of energy such as heat, electricity and movement. The design of these devices is difficult and engineers often use simulations to achieve an acceptable level of performance and/or meet emission standards. ‘Computational Fluid Dynamics’ (CFD) simulations can generate useful data, e.g., heat release and pollutant concentrations, during the combustion process. ‘Combustion models’ provide data such as chemical reaction rates and burning temperatures that are needed for CFD.
This thesis lies in the field of ‘combustion modelling’, which deals with development and testing of combustion models for CFD. Standard combustion models lend themselves to either ‘premixed’ combustion (e.g., in a carburettor-fed petrol engine) or ‘non-premixed’ combustion (e.g., in a direct-injection Diesel engine). In reality, however, devices often operate in ‘mixed-mode’ conditions, exhibiting regions of premixed and non-premixed characteristics. This has spurred the development of more sophisticated ‘mode- independent’ models, in an effort to provide more accurate performance and pollutant prediction. One such is the Linear Eddy Model (LEM), which is highly accurate but is computationally expensive for CFD. This study proposes techniques to improve the computational efficiency of LEM, using existing ideas such as ‘mesh agglomeration’ in novel ways. The work encompasses method development, validation using experimental data, and finally, identifies means to improve the robustness and general usability of the proposed method.
This thesis lies in the field of ‘combustion modelling’, which deals with development and testing of combustion models for CFD. Standard combustion models lend themselves to either ‘premixed’ combustion (e.g., in a carburettor-fed petrol engine) or ‘non-premixed’ combustion (e.g., in a direct-injection Diesel engine). In reality, however, devices often operate in ‘mixed-mode’ conditions, exhibiting regions of premixed and non-premixed characteristics. This has spurred the development of more sophisticated ‘mode- independent’ models, in an effort to provide more accurate performance and pollutant prediction. One such is the Linear Eddy Model (LEM), which is highly accurate but is computationally expensive for CFD. This study proposes techniques to improve the computational efficiency of LEM, using existing ideas such as ‘mesh agglomeration’ in novel ways. The work encompasses method development, validation using experimental data, and finally, identifies means to improve the robustness and general usability of the proposed method.
Ämneskategorier (SSIF 2025)
Strömningsmekanik
Infrastruktur
C3SE (-2020, Chalmers Centre for Computational Science and Engineering)
ISBN
978-91-8103-166-9
Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 5624
Utgivare
Chalmers
HC1, Hörsalsvägen
Opponent: Dr David Lignell, Brigham Young University, USA