Unsupervised analysis of experiments of laminar flame propagation in a spherical enclosure

The paper illustrates the methodology developed for unsupervised analysis to be conducted on high-definition, high sampling rate image sequences collected in experiments with a single spark ignition optically accessible spherical bomb. Images recorded along the line-of-sight were first processed to identify the reaction front, and then analyzed by means of a two-dimensional numerical estimation technique. The laminar flame front is detected by making use of the concept of "scalar dissipation rate" basing on flame luminosity data, i.e. the square of the gradient of flame luminosity. The new scalar field is then tracked to derive the time history of the flame radius. In order to extract the Region Of Interest from the images, masking techniques are employed, whereas signal-to-noise ratio is improved by means of data binning. The proposed automatic, non-intrusive method proves effective in providing a fast characterization of the flame propagation phenomenon in terms of apparent velocity.