Data driven investigation of local combustion mode in heterogeneous combustion

Abstract

Moderate or Intense Low-oxygen dilution (MILD) combustion was studied by utilizing data driven techniques. It was found that MILD combustion yields relatively small scalar gradients whereas intense reaction zones were predominantly observed. This could be caused due to local ignition and flame interactions in the combustion field. Such zones are identified in direct numerical simulation (DNS) field with appropriate thresholding of scalar gradient and reaction rate. Then, a data driven model for predicting local combustion modes was proposed and validated for large eddy simulation (LES) context, using filtered DNS data at various filter sizes.