Analysis of flame detection data from multiple-ion probes using feature extraction

Abstract

The multiple-ion-probe method detects flames using a plurality of ion probes installed on the wall of a combustion chamber and by reconstructing the dynamic behavior of the flame front along the wall. Although this method is effective only close to the wall, it can indirectly aid in visualizing flame propagation behavior. Because this method has an extremely high time resolution, it can accurately measure high-speed phenomena such as knocking in a spark-ignition engine. This study aims to establish a method to automatically determine the characteristics of combustion for each mixture using the acquired data from a multiple-ion-probe measurement system; various propagating flames in combustible mixtures with different compositions were used. The combustible mixture was composed of methane or LPG as the fuel and argon or nitrogen as the diluent. A stoichiometric mixture of fuel and oxygen was diluted by changing the diluent ratio to prepare fourteen types of mixtures for investigation. From the obtained individual experimental data, twelve types of scalar features were extracted and compared with the experimental data. We compared the scalar features extracted from the output data of the multiple-ion probe with the different combustion modes obtained in the series of experiments; we found that a few of the features strongly responded to specific propagation states. Thus, we confirmed that by using the features, it is possible to determine the characteristics of the propagation states; such characteristics are difficult to determine merely from the flame surface shape reconstructed from the multiple-ion-probe data.