MULTI-OBJECTIVE OPTIMISATION OF AN INDUSTRIAL MIXING DEVICE USING LES

In this study, the geometry of an industrial mixing device has been optimised such that unmixedness and pressure drop are minimised simultaneously. Such scenarios are baseline problem in many industrial scectors. \noindent A computational work-flow has been set-up to carry out the analysis of the physical problem using LES. It involves the execution of a series of open source software packages where \textbf{OpenSCAD} \cite{OpenSCAD_c} is used for geometry generation and \textbf{OpenFOAM} \cite{OpenFOAM_c} is used for meshing, CFD analysis and post-processing. \noindent For automatic optimisation, the computational work-flow is controlled by the optimisation software suite \textbf{Dakota} \cite{Dakota_c}. This component implements iterative optimisation algorithms which require the evaluation of the objective functions (unmixedness, pressure drop) as functions of the input parameters (geometric design parameters) as implemented in the above work-flows. The system runs \textit{fully automatic}, \textit{in parallel} and \textit{without any user interaction}. Due to large number of variants and computaional expense of the LES runs, the workflow is executed on the JURECA supercomputer (45,216 CPU cores) with good turn-around. \noindent The result of the the multi-objective problem has been solved using a genetic algorithm. The result is the Pareto-front -- the line of optimal solutions where each point represents a configuration with an optimal compromise between pressure drop and temperature increase. The Pare-to front is extremely useful in the engineering practise when the acceptable trade-off has not been defined.