Data driven forecast of concurrent flame spread in micro-gravity

Abstract

This paper presents a methodology that combines data assimilation and physical modelling of flame spread for fire growth forecast. The concurrent flame spreading over a flat solid sample in the absence of buoyant flows is considered as a simple canonical fire spread configuration. To predict flame spread rate and flame length evolution, the analytical solution to the two-dimensional boundary layer non-premixed combustion problem is combined with a CFD model which delivers the structure of the flow away from the fuel surface. In the process, two coefficients which intervene in the gas and solid heat transfer equations are assimilated using the pyrolysis length data as a flame spread rate surrogate input to absorb the shortcomings of the modelling. The robustness of the overall method is evaluated through convergence assessment of these two assimilated variables for different initial guesses, and for different values of the input invariants. Validation over large-scale microgravity data provides confidence in this approach, and demonstrates its potential to deliver flame spread predictions from initial measurements at a reduced computational cost. However, discrepancies between flame length measurements and predictions question the degree of correlation between pyrolysis and flame lengths.