Identification of local extinction and prediction of reignition in a spark-ignited sparse spray flame using data mining

Identification of local extinction and prediction of reignition in a spark-ignited sparse spray flame using data mining

Direct Numerical Simulations (DNS) of droplet fields which are ignited using a spark are investigated to deduce any behaviour that distinguishes between the cases where successful flame propagation occurs and where a flame ignites but subsequently extinguishes. At the instant the spark was deactivated, some of the studied cases displayed no local extinction, others showed some local extinction (one with reigni- tion and the rest with global extinction) and the rest showed global extinction. The gaseous field at this instant was analysed using the data mining technique the Gaussian Mixture Model on each case sepa- rately; this method groups data points, enabling distinction between the various behaviours. The results from this analysis showed that in the case with local extinction–reignition, the regions of space near the flame kernel which produced local quenching were caused by evaporating droplets. These regions of local quenching were relatively small compared to the strong flame front surrounding them; the regions of lo- cal quenching were also relatively far from the centre of the flame kernel. In contrast, in cases with local then global extinction, the droplets created regions which were extensions of the relatively-small flame front, and these regions behaved in a similar manner to the flame propagation. As a consequence, these cases were unable to support a self-sustaining flame. Such distinctive behaviour promises opportunities to detect situations where global extinction is imminent and implement appropriate control strategies to prevent global extinction.

File reproduced in accordance with the copyright policy of the publisher/author.