Currently, the most widely used droplet collision model is O'Rourke's algorithm. However, it has large mesh dependency in spray simulations. The calculated results have different spray structures in different meshes and the predicted mean drop sizes increase with the mesh resolution. The predicted mean drop size can increase 30 microns from the coarsest mesh to the finest one. There are two reasons for this. Firstly, the O'Rourke model does not consider the collision between parcels that reside at the edges of two adjacent cells. Secondly, it assumes the probability of a pair of parcels colliding is proportional to their relative velocity. The result of this assumption is that parcels furthest apart from each other in a cell are most likely to collide because they have the greatest relative velocity.
In order to reduce mesh dependency, a new model, called CMC (cross mesh collision), is proposed. It is an extension of the O'Rourke droplet collision model. The collision between parcels that reside at the edges of two adjacent cells is considered in the new model and the problem where the parcels have the largest distance are most likely to collide is resolved using the method of collision limit angle.
This model is briefly described, and incorporated into the KIVA code. The predicted results using the CMC model and the O'Rourke model are compared in different meshes, mesh resolutions, and number of parcels. The new CMC model shows less mesh dependency and less sensitivity to the number of parcels. The difference in predicted mean drop size decreases to 5 microns from the coarsest mesh to the finest one. But the improved code performs only slightly better in predicting spray penetration than the original one.