Droplet generators are at the heart of many microfluidic devices developed for life science applications but are difficult to tailor to each specific application. The high fabrication costs, complex fluid dynamics, and incomplete understanding of multi-phase flows make engineering droplet-based platforms an iterative and resource-intensive process.
First, we demonstrate the suitability of desktop micromills for low-cost rapid prototyping of thermoplastic microfluidic devices. With this method, microfluidic devices are made in 1 - 2 hours, have a minimum feature size of 75 μm, and cost less than $10. These devices are biocompatible and can accommodate integrated electrodes for sophisticated droplet manipulations, such as droplet sensing, sorting, and merging.
Next, we leverage low-cost rapid prototyping to characterize the performance of microfluidic flow-focusing droplet generators. Specifically, the effect of eight design parameters on droplet diameter, generation rate, generation regime, and polydispersity are quantified. This was achieved through orthogonal design of experiments, a large-scale experimental dataset, and statistical analysis.
Finally, we capitalize on the created dataset and machine learning to achieve accurate performance prediction and design automation of flow-focusing devices. The developed capabilities are captured in a software tool that converts high-level performance specifications to a device that delivers the desired droplet diameter and generation rate. This tool effectively eliminates the need for resource-intensive design iterations to achieve functional droplet generators. We also demonstrate the tool’s generalizability to new fluid combinations with transfer learning.
We expect that our newly established framework on rapid prototyping, performance characterization informed by design of experiments, and machine learning guided design automation to enable extension to other microfluidic components and to facilitate widespread adoption of droplet microfluidics in the life sciences.
| Identifer | oai:union.ndltd.org:bu.edu/oai:open.bu.edu:2144/42599 |
| Date | 15 May 2021 |
| Creators | Lashkaripour, Ali |
| Contributors | Densmore, Douglas M. |
| Source Sets | Boston University |
| Language | en_US |
| Detected Language | English |
| Type | Thesis/Dissertation |
| Rights | Attribution 4.0 International, http://creativecommons.org/licenses/by/4.0/ |
Page generated in 0.0016 seconds