Fig. 1: Pipeline for collating data and training models to enable performance prediction and design automation of SE and DE droplet generation.
From: Design automation of microfluidic single and double emulsion droplets with machine learning
a Composition of datasets exploring effects of geometry (orifice width (Wor), continuous inlet width (CIW), dispersed inlet width (DIW), channel height (H), and outlet channel width (OCW)), fluid properties, and flow rates on (i.) SE and (ii.) DE droplet generation collated to yield a final (iii.) comprehensive dataset with 868 entries. The combined dataset includes 8 dispersed fluids, 6 continuous fluids, and 46 devices that yield aqueous-in-oil and oil-in-aqueous droplets with diameters from 15 to 250 μm at rates of 5–12,000 Hz. b Schematic of model training to predict: (1) droplet diameter based on device geometry, fluid properties, and flow rates, and (2) droplet generation rates based on predicted diameters and conservation of mass (see Methods). c Predictive models were integrated with a custom search algorithm to convert user-specified desired droplet characteristics to an optional device design and flow rates. This open-source software tool, DAFD 3.0, is available at: dafdcad.org. Source Data are provided as a Source Data file.
