Spray characterization

It used to be a single parameter, the surface-to-volume diameter or Sauter Mean Diameter which was of primary interest of sprays in combustion applications a few decades ago. Nowadays, the tightening pollutant emission standards and the demand for highly fuel efficient applications made a detailed analysis of spray a hot topic that involves detailed size characterization and its modeling.

Airblast atomization, which is extensively investigated by the group, is an excellent platform to analyze the atomization process since the characteristic dimensions of the nozzle have a low influence on the final spray quality. Besides combustion, the range of applications is rather wide which makes it an excellent topic for base research that has several uses.

Demonstration of spray visualization by two laser sheets
High-speed camera image of diesel oil (top left), light heating oil (top right), crude rapeseed oil (bottom left), and water (bottom right) atomization in the vicinity of the nozzle

Statistical analysis

The first problems of droplet size measurement are that the droplet may travel in a dense spray region, its shape is not perfectly spherical, and the size and velocity are measured in parallel. Therefore, the uncertainty of this measurement is presently unknown. However, it is claimed that it provides information with at least 5% accuracy which seems a rather coarse estimation.

The second problem is the sample size. How many droplets are required to characterize the spray under steady-state conditions? Usually, statistics struggle with low number of samples where enough is about ten. But tens of thousands rise several interesting questions since the power of the statistical tests will be excessively high, making general characterization impossible.

The spray characteristics can be captured through the use of probability density functions. The variation of the parameters shows the evolution of the spray. Nevertheless, the periphery should be evaluated with care since the smooth inner characteristics fail there. How can we develop an appropriate characterizing model then?