Data-driven approach for design and optimization of rotor-stator mixers for miscible fluids with different viscosities

Rotor-stator mixers (RSMs) are essential equipment used in the food, pharmaceutical, and cosmetic industries. The performance of RSMs is conventionally analyzed using three approaches based on empirical correlations and experimental and numerical analyses. This paper introduces a new approach based on a multifidelity data-driven framework for designing and optimizing RSMs that systematically combines elements of experimental and highresolution computational fluid dynamics (CFD) models to train a machine learning model to predict RSM performance. It was subsequently coupled to an optimization solver based on a genetic algorithm. The developed framework was applied to a small-scale RSM with variable rotor blade radius, rotor speed, stator mesh width, and mass flow rate that is operated to mix two miscible fluids with different viscosities. The RSM was optimized at two regimes of low and high viscosities to achieve the best performance in terms of shaft power and mixing indices. The results demonstrate the successful application of the developed frameworks. The optimization of the mixing of high-viscosity fluids yielded a higher improvement than that for low-viscosity fluids. A decrease in shaft power of approximately 50% was obtained for the high-viscosity fluids, whereas the mixing indices increased up to 20% by the optimization of the geometrical input design parameters. The proposed framework can be advantageous for designers and operators at the industrial scale.