Large Eddy Simulation Of Turbulent Flames Research Articles

Towards a generalised artificial neural network for sub-grid filtered density function closure in turbulent combustion

An artificial neural network (ANN) is trained on moderate or intense low-oxygen dilution (MILD) combustion to predict the sub-grid filtered density function (FDF) in large eddy simulation (LES). For wide usability, a new quantity ψ is calculated by scaling the filtered mixture fraction using the stoichiometric value and a logarithmic function, which is used as an input feature of the ANN. Self-testing on MILD combustion is first conducted for validation and determining the best layout of ANN. An ANN with 4 hidden layers and the activation function of rectified linear unit (ReLU) has the highest accuracy. This ANN is subsequently tested on five different premixed combustion cases. Overall predictions of the progress variable FDF are satisfactory, and the filtered chemical source term modelled by the FDFs is in good agreement with the DNS data for each case. However, an underprediction of the filtered reaction rate occurs for the hydrogen–air flames, and the ANN accuracy is lower than that of a presumed β-FDF approach. DNS data of one premixed flame are then added to the training datasets, and newly trained ANNs are tested on remaining premixed flame cases. When the testing cases are similar to the added training dataset, improved predictions of marginal FDFs are observed, and the modelled progress variable source terms show good comparisons with the corresponding DNS data. Compared to the presumed β-FDF method, the new ANN shows an improved accuracy and serves as a viable alternative for FDF-based combustion modelling in the LES of turbulent flames.

Effect of multiscalar subfilter PDF models in LES of turbulent flames with inhomogeneous inlets

To apply reduced-order manifold combustion models in Large Eddy Simulations (LES) of systems with multiple inlets, it is necessary to incorporate more than one mixture fraction. As a result, the joint subfilter Probability Density Function (PDF) of the mixture fractions must be modeled. To determine the sensitivity of the predicted turbulent flame structure to these models, LES calculations of the Sydney piloted jet burner with inhomogeneous inlets have been performed using the Dirichlet, Connor-Mosimann (CM), and Beta-Delta (BD) distributions. A memory-efficient convolution-on-the-fly approach was applied to enable use of more complex subfilter PDF models, such as the CM distribution. Strong sensitivities in the LES predictions were observed to the choice of subfilter PDF model. The Dirichlet distribution was found to be ill-suited to this partially premixed turbulent flame because as a subfilter PDF model it implicitly assumes symmetric mixing in mixture fraction space, which is inconsistent with partial premixing. Good agreement with the experimental data was obtained using both the CM and BD distributions, which are mathematically consistent with partial premixing. The CM distribution is the most general of the considered distributions and gives the best predictions, but this comes with increased complexity and computational cost.

Scalar energy fluctuations in Large-Eddy Simulation of turbulent flames: Statistical budgets and mesh quality criterion

Large-Eddy Simulation (LES) provides space-filtered quantities to compare with measurements, which usually have been obtained using a different filtering operation; hence, numerical and experimental results can be examined side-by-side in a statistical sense only. Instantaneous, space-filtered and statistically time-averaged signals feature different characteristic length-scales, which can be combined in dimensionless ratios. From two canonical manufactured turbulent solutions, a turbulent flame and a passive scalar turbulent mixing layer, the critical values of these ratios under which measured and computed variances (resolved plus sub-grid scale) can be compared without resorting to additional residual terms are first determined. It is shown that actual Direct Numerical Simulation can hardly accommodate a sufficiently large range of length-scales to perform statistical studies of LES filtered reactive scalar-fields energy budget based on sub-grid scale variances; an estimation of the minimum Reynolds number allowing for such DNS studies is given. From these developments, a reliability mesh criterion emerges for scalar LES and scaling for scalar sub-grid scale energy is discussed.