Influence of the particle injection rate, droplet size distribution and volume flux angular distribution on the results and computational time of water spray CFD simulations.

The paper presents a detailed sensitivity analysis on the volume flux probability density function (PDF) to represent water spray patterns with computational fluid dynamics (CFD). The effects of the turbulent viscosity model and the cell size are also investigated. The test case considered herein is a 30 ° full cone water mist spray emerging from a nozzle that operates at a pressure of 750 kPa and delivers a water flow rate of 0.084 lpm. The errors solely induced by the limited number of computational droplets per second, N p , are proportional to 1 / N p and could reach up to 35%. The computational time generally increases linearly with N p . The paper illustrates also the better numerical performance of the lognormal-Rosin-Rammler droplet size distribution over the Rosin-Rammler distribution, especially in terms of reaching a converged volume-median diameter with increased N p . Furthermore, a uniform angular distribution is shown to provide results in better agreement with experimental data than a Gaussian-type distribution for the case at hand. For a sufficiently fine grid, the dynamic Smagorinsky and the modified Deardorff models converge to similar radial profiles of the water volume flux at 300 mm from the nozzle, with a deviation of less than 6% from the experiments. The deviations for the volume-median diameter are about 50% in the core region of the spray. [ABSTRACT FROM AUTHOR]