Flame front reconstruction and volume estimation through computational geometry: a case study on machine vision applied to combustion systems

Abstract

A computationally-supported experimental procedure to estimate the primary dimensions of diffusion flames, using volume reconstruction from thermal imagery, is presented. The experimental setup uses a 4 × 16.94 mm radial distribution gas-burner, with a 0.8 mm nozzle diameter, a thermal imaging camera and a proprietary image processing algorithm. Flame thermal imagery was captured, using four different fuel loads, 350, 650, 950 and 1200 cc/min, from two different visualisation planes, 0° and 90°. The images were visually and qualitatively processed leaving aside the temperature measurement and favouring instead a non-dimensional temperature gradient, . Corresponding flame front structures were estimated and reconstructed employing computational geometry. The height and diameter magnitudes were measured indirectly through a reference length. The results show that at the flame front structure separates itself from the background noise. Furthermore, when compared against available benchmarks, at and , the resulting flame coincides with the luminous and continuous flame heights, respectively. This approach yields maximum relative error of 36.54% and 18.91% for both compared geometries. When compared to image convolution and spatial density clustering procedures, this approach reduces the maximum error obtained by 47%. Based on this information, the methodology presented is considered suitable for dimensioning diffusion flames, thus, proposed as an estimation tool for the design and manufacturing of gas-fuelled appliances/devices.