SGS Reaction rate modelling for MILD combustion based on machine-learning combustion mode classification: Development and a priori study

A neural network (NN) aided model is proposed for the filtered reaction rate in moderate or intense low-oxygen dilution (MILD) combustion. The framework of present based on partially stirred reactor (PaSR) approach, and fraction reactive structure appearing PaSR predicted using different NN’s, to consider both premixed non-premixed conditions while allowing use imbalanced training data between combustion direct numerical simulation (DNS) data. key ingredient local mode prediction performed by another NN, which developed a previous study. trained was then assessed two unknown DNS cases, yields much higher level (more MILD condition) Karlovitz number than cases used as performance assessment has been carried out means Pearson’s correlation coefficient mean squared error. For zeroth-order approximated rate, with target values shows relatively high values, suggesting that trend field, approximation, well correlated actual field. This suggests equation promising if appropriately predicted, objective On other hand, substantially lower error observed range filter sizes approximation. can account SGS contribution reasonably well.