On the improvement of two-dimensional curvature computation

Measurement of curvature of the flamefront of premixed turbulent flames is important for the validation of numerical models for combustion. In this work, curvature is measured from contours that outline the flamefront, which are generated from laser-induced fluorescence images. The contours are inherently digitized, resulting in pixelation effects that lead to difficulties in computing curvature of the flamefront accurately. A common approach is to fit functions locally to short sections along the flame contour, and this approach is also followed in this work; the method helps smoothen the pixelation before curvature is measured. However, the length and degree of the polynomial, and hence the amount of smoothing, must be correctly set in order to maximize the precision and accuracy of the curvature measurements. Other researchers have applied polynomials of different orders and over different segment lengths to circles of known curvature as a test to determine the appropriate choice of polynomial; it is shown here that this method results in a sub-optimal choice of polynomial function. Here, we determine more suitable polynomial functions through use of a circle whose radius is sinusoidally modulated. We show that this leads to a more consistent and reliable choice for the local polynomial functions fitted to experimental data. A polynomial function thus determined is then applied to flame contour data to measure curvature of experimentally acquired flame contours. The results show that there is an enhancement in local flame speed at sections of the flamefront with a non-zero curvature, and this agrees with numerical models.