An optimisation framework for minimising power consumption, dead volume, and shear rate of mixing tanks containing shear-thinning fluids utilising CFD

Abstract

A hydrodynamic model of a laminar mixing tank containing a shear-thinning power-law fluid stirred by a helical ribbon impeller is created using the Lattice Boltzmann Method and used to train regression models to predict power number, percentage of dead volume, and shear rate statistics. A Reduced Gradient Algorithm is used to maximise a Composite Desirability Function, composed of the regression models, to optimise impeller geometry for selected flow behaviour indices following a Design of Experiments methodology. The regression models are trained on sixty simulations, and validated on eight separate simulations. The range of deviation between regression models and each output response are as follows: impeller power number = 0.047 % to 18 %, percentage of tank dead volume = 0.34 % to 18 %, average shear rate = 0.56 % to 9.0 %, maximum shear rate = 0.46 % to 11 %, and standard deviation of shear rate = 0.69 % to 5.5 %. This is found to be an efficient methodology for optimising mixing conditions, when only a hydrodynamic model is available.