Improvement of the weighted multi-point (WMP) radiation model for diffusive flames by the application of a set of stochastic optimisation algorithms

Abstract The weighted multi-point source model (WMP) has been proposed to model flame radiation from far to near distances, encompassing regions where the single point (SP) model cannot predict it adequately. To develop its formulation, the WMP has been studied as an inverse problem, being optimised to minimise the error between experimental and numerical data. Efficient optimisation methods are thus necessary, and a performance study applied to the WMP model is essential for its development. This study aims to evaluate which types of stochastic algorithms are the best for this problem and to improve the current radiation model comparing the results to those by previous studies. Five algorithms with different characteristics are chosen and tuned for the WMP problem with a design of experiments (DoE) methodology, and applied to each configuration of the model. The best performance was shown by the grey wolf optimiser (GWO), allying stability to fast convergence. The best solution improved previous results by 23.7%, and was also 82.72% better than the solution calculated with the SP model, and 76.24% better than the one calculated with the canonical WMP. A trend in radiation emission distribution is observed with results by previous studies, guiding better formulations in weight distribution.