Machine learning electrospray plume dynamics

Machine learning models are applied to simulated electrospray particle data to investigate plume dynamics from emission to final particle properties. A limited set of final particle properties are successfully regressed exclusively from emission property inputs. Random Forest model feature rankings for final plume angle reveal that particle charge has dominant influence when emission velocity is strictly axial, while lateral emission velocity has dominant influence when particles are emitted with an off-axis velocity component. In addition to providing correlations between initial and final particle properties, the machine learning models also identify correlations between different final particle properties. These correlations reveal opportunities for experimental approaches and diagnostic design by determining experimental measurements that offer insight into desired final particle properties.