Publication Date
7-13-2026
Document Type
Article
Publication Title
IEEE Access
Volume
14
DOI
10.1109/ACCESS.2026.3713003
First Page
111879
Last Page
111888
Abstract
Saliva-based point-of-care diagnostic devices provide accessible, non-invasive health monitoring but face engineering challenges in ensuring reliable, cost-effective performance outside laboratory settings. Traditional design approaches rely heavily on repeated cycles of simulation, design, and prototyping, creating computational bottlenecks that slow innovation. This work introduces an automated surrogate-assisted optimization framework that integrates multiphysics simulation data, surrogate modeling, and multi-objective optimization to accelerate microchannel design. Synthetic datasets generated through multiphysics simulations are used to train a neural network surrogate model capable of predicting flow performance metrics, including the Dean number and pressure drop. This trained model achieved a mean squared error of 0.000012 and an R2 of 0.999645. By coupling the surrogate model with the NSGA-II genetic optimization algorithm, the framework efficiently identifies Pareto-optimal geometries that balance mixing strength and flow resistance, reducing reliance on repeated, resource-intensive simulations. Validation against the multiphysics solver yields mean percent errors of 3.15% for the Dean number and 2.03% for pressure drop. The mixing performance of the prototypes generated by our proposed automated framework was also validated, showing that the microfluidic mixers can deliver complete mixing regardless of geometric constraints. This automated design framework significantly front-loads computational effort while enabling rapid, interpretable, and scalable optimization for microfluidic channel design in point-of-care diagnostics.
Keywords
medical device, Microfluidic, micromixer, multi-objective optimization, rapid prototyping, surrogate model
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Department
Mechanical Engineering; Biomedical Engineering
Recommended Citation
Ronald Jose, Yunjian Qiu, Lin Jiang, and Yun Wang. "An Automated Surrogate-Assisted Optimization Framework for Rapid Microfluidic Channel Design of Point-of-Care Diagnostic Devices" IEEE Access (2026): 111879-111888. https://doi.org/10.1109/ACCESS.2026.3713003
