Drop collision analysis by using many-body dissipative particle dynamics and machine learning

Abstract

Droplet collisions are widely observed in daily life and industries. The study of droplet collision dynamics can guide engineering applications in, for examples, inkjet printing, fan cooling, and engine spraying. In this Letter, a numerical simulation method of droplet collision is proposed on the basis of the many-body dissipative particle dynamics. For the collision of two droplets of the same size, the post-collision morphology is analyzed in terms of two key factors: the initial eccentricity parameter and the Weber number. Then, the collision morphology is learned and classified in conjunction with a multilayer perceptron in order to quickly predict the collision morphology from the initial conditions. A machine learning model linking the initial conditions of collision with the post-collision droplet morphology is developed based on three typical morphologies generated by the collision of identical volume droplets. This study provides more insights into droplet dynamics and may benefit related engineering applications.