Improving E-jet Process Capabilities with Black Box Machine Learning

Electrohydrodynamic printing or simply E-jet printing is a relatively new way of creating products in the additive manufacturing domain. Like any other manufacturing process, Improving the process's throughput is one of the essential conditions for it to be a recognized free-form manufacturing method. Tuning operating conditions by optimization route proved to be an effective approach to enhance E-jet performance capability. In this paper, an artificial intelligence technique namely random forest regression algorithm is implemented to search for a favorable manufacturing environment of E-jet. Nozzle to substrate gap, applied potential difference, and supply rate of the material are chosen as operating control parameters in the experimental investigation. Deposited feature diameter is taken as the performance measure of the E-jet. Despite having limited training data, the Random Forest (RF) regression model outperformed (in terms of prediction accuracy) several other machine-learning models used in the recent past. It is found that random forest black box model performs admirably for predicting the feature dimension with a root mean square deviation of 6.67% and R2 value of 94.07%. The current study can be used as a template plan to maneuver other operating conditions of E-jet to get better additive manufacturing capability.