Machine Learning and CFD: a new approach to Simulate and Optimize Micromixers in Different Geometries

This work seeks to explore a new approach to optimization in the microfluidics field, using Computational Fluid Dynamics (CFD) and Artificial Intelligence techniques. The objective of this combination is to enable the realization of a global optimization with a lower computational cost. This optimization occurs with the possibility of building a simulation dataset in a shorter time, using a dense neural network trained with the data obtained using CFD. In this work, it was possible to test, for Y-geometry micromixers, varying parameters such as obstruction diameter (OD) and obstruction offset (OF), allowing to obtain a neural network and perform an optimization through a genetic algorithm.