Geometry-free prediction of inertial lift forces in microfluidic devices using deep learning

Short summary

Researchers developed a neural network model that predicts particle lift forces in microfluidic devices without requiring explicit geometric parameters, addressing the burden of training separate models for each channel type. The model generalizes to unseen channel geometries far better than existing approaches while maintaining performance on trained geometries. The technique successfully transfers to particle tracing software and enables accurate migration prediction across diverse channel designs.

• •Geometry-agnostic neural network replaces need for geometry-specific models
• •Achieves superior generalization to unseen channel designs
• •Readily integrates with existing particle simulation software

Generated with AI, which can make mistakes.