Computational fluid dynamics and machine learning as tools for optimization of micromixers geometry

Microfluidic devices have become a new trend in different fields and attracted attention due to their compact size capability deal with small amount of fluid. Micromixing is an efficient way mix miscible fluids at this microfluidic level. This work explores approach for optimization microfluidics, using CFD (Computational Fluid Dynamics) ML (Machine Learning) techniques. The objective combination enable global lower computational costs. A Y-type micromixer inspires the initial geometry cylindrical grooves on surface main channel obstructions inside it. Simulations circular were carried out OpenFOAM software observe influence these obstacles. effects obstruction diameter its offset percentage mixing, pressure drop energy cost investigated. Numerical experiments analyzed machine learning. neural network was used train dataset composed as inputs mixing outputs. genetic algorithm find that offers maximum value minimum value. optimal obtained 131 mm 10 mm, which corresponds medium close wall. It worth mentioning each simulation takes around 4h, total time guarantee would be about 25.000 days. With methodology, production, training 40 procedure tremendous advantage optimization.