Local curvature measurements of a lean, partially premixed swirl-stabilised flame

Abstract

A swirl-stabilised, lean, partially premixed combustor operating at atmospheric conditions has been used to investigate the local curvature distributions in lifted, stable and thermoacoustically oscillating CH 4 -air partially premixed flames for bulk cold-flow Reynolds numbers of 15,000 and 23,000. Single-shot OH planar laser-induced fluorescence has been used to capture instantaneous images of these three different flame types. Use of binary thresholding to identify the reactant and product regions in the OH planar laser-induced fluorescence images, in order to extract accurate flame-front locations, is shown to be unsatisfactory for the examined flames. The Canny-Deriche edge detection filter has also been examined and is seen to still leave an unacceptable quantity of artificial flame-fronts. A novel approach has been developed for image analysis where a combination of a non-linear diffusion filter, Sobel gradient and thresholdbased curve elimination routines have been used to extract traces of the flame-front to obtain local curvature distributions. A visual comparison of the effectiveness of flamefront identification is made between the novel approach, the threshold binarisation filter and the Canny-Deriche filter. The novel approach appears to most accurately identify the flame-fronts. Example histograms of the curvature for six flame conditions and of the total image area are presented and are found to have a broader range of local flame curvatures for increasing bulk Reynolds numbers. Significantly positive values of mean curvature and marginally positive values of skewness of the histogram have been measured for one lifted flame case, but this is generally accounted for by the effect of flame brush curvature. The mean local flame-front curvature reduces with increasing axial distance from the burner exit plane for all flame types. These changes are more pronounced in the lifted flames but are marginal for the thermoacoustically oscillating flames. It is concluded that additional fuel mixture fraction and velocimetry studies are required to examine whether processes such as the degree of partial-premixedness close to the burner exit plane, the velocity field and the turbulence field have a strong correlation with the curvature characteristics of the investigated flames.

Effects of mixture composition and turbulence intensity on flame front structure and burning velocities of premixed turbulent hydrocarbon/air Bunsen flames

“…This observation indicates that the flame front curvature distributions are dominant by the large-scale structures and the effect of Kolmogorov-scale eddies are minimal. This observation is in agreement with previous measurements of Yuen and Gülder [87], whereas Kostiuk et al [86], Haq et al [77], Gashi et al [85], and Bayley et al [84] observed that the flame front curvature distribution becomes wider by increasing the turbulence intensity. It is observed that the flame front curvature distributions for lean/stoichiometric and rich mixtures are identical when the unstrained premixed laminar burning velocity, non-dimensional bulk flow velocity, non-dimensional turbulence intensity, and non-dimensional longitudinal integral length scale were kept constant, see, for example, Flames M9 and M16.…”

“…These distributions are observed to be Gaussian, and they are symmetrical about zero flame front curvature. Similar observations were previously reported in the literature for premixed turbulent Bunsen flames, see, for example, [34,[53][54][55]73,80,81], premixed turbulent low-swirl stabilized flames, see, for example, [62,82,83], partially premixed turbulent low-swirl stabilized flames, see, for example, [84], premixed turbulent propagating flame kernels, see, for example, [77,85], premixed turbulent V-shaped flames, see, for example, [36], and premixed turbulent opposed streams, see, for example, [86]. It is observed that the standard deviation of the distributions does not change with increasing equivalence ratio, see, for example, Flames E1-E8.…”

“…The flame front curvature results in the current study are limited to two-dimensional measurements. In addition, the flame front curvature normal to the measurement plane is not identified, and the measured curvature will be scarcely the case of the principle flame front curvature [84]. Gashi et al [85] compared the twoand three-dimensional probability density functions of the flame front curvature for premixed turbulent propagating flame kernels using direct numerical simulation.…”

The structure of turbulent flames in fractal- and regular-grid-generated turbulence

“…The extracted contours were then represented parametrically in Cartesian coordinates, x(s) and y(s), as a function 14 of the path length parameter, s, of the contour and all subsequent flame contour quantities such as flame front normals and flame front curvatures, , were calculated from the parameterized contours. We used the processing procedure described in [41] for the extraction of the flame curvature, which results in the error of the local curvature value being within 10%.…”

Flame front detection and curvature calculation using level set

“…Some traditional methods usually use edge detection operators [2], [8] to detect flame front. They always have the problems of low efficiency for noisy flame images and still suffer from the difficulties of determining the suitable global or local thresholds from intensity inhomogeneous images.…”