Reducing Grid Dependency in Droplet Collision Modeling

Abstract

Droplet collision models have been criticized for creating large mesh dependency in spray calculations. These numerical errors are very troublesome; they behave erratically and interfere with the predictive ability of physical models. The collision method used in KIVA can cause mesh dependent changes in average drop size of over 40 microns. In order to reduce mesh dependency, a new method has been developed for calculating the incidence of collision. The solution is to create a special collision mesh that is optimized for accuracy. The mesh is created automatically during the spray calculation. Additionally, a different stochastic collision sampling technique is also used. The new method, called the NTC algorithm, was incorporated into KIVA and found to be much faster than older algorithms. Calculations with 60,000 parcels required only a few CPU minutes. With the new NTC method and collision mesh, the mesh dependence of the drop size is only nine microns. This remaining mesh dependency is found to be due to the drag calculations and is not the fault of the collision algorithm.