Mapping the structure of a liquid spray by means of neural networks

This paper presents the results of a study aimed at mapping the structure of a liquid spray system along the spray cone. Experimental results obtained by phase-Doppler anemometry (PDA) consisted of the number distributions of droplet size, velocity, and interparticle arrival time at different locations within the spray cone. The experimental data were analyzed by means of multivariate statistical techniques, in order to identify different regimes in the spray. Neural network models (NN) were fitted to the experimental data, resulting in good agreement between experimental and calculated results for most locations within the spray cone. However, in some parts of the cone the agreement was poor, and the general trends could not be well predicted by the NN models. The mismatch is due to unsteady spray conditions or incomplete atomization (e.g. existence of non-spherical particles). This fact was adopted as a criterion to identify the regions where the spray is fully established, corresponding to the regions of the spray where PDA measurements can be successfully performed. This criterion has been applied along the spray cone for different operating conditions, and can be used as a tool to map the fluid dynamic characteristics in liquid sprays.