Performance of different optimization concepts for reactive flow systems based on combined CFD and response surface methods

Philip Roßger, Andreas Richter

CHALLENGE - Optimization of reactive flow systems, e.g. coal and biomass combustion, gasification or partial oxidation, is usually a multi-parameter problem, while geometry configurations (e.g. reactor dimensions, burner dimensions) and operating conditions (e.g. fuel mass flow, reactor temperature, cooling capability) mainly determine the process efficiency and profitability. Indicators are fuel conversion rate and syngas quality, respectively. The complex interaction of turbulent mixing and several homogeneous reactions requires the use of CFD, since simplified 0D and 1D approaches cannot capture the underlying phenomena.

SOLUTION - In this work a metamodel (MM)-based optimization procedure is developed for an accelerated multi-parameter and multi-objective optimization of reactive flow systems.​ The general focus was to provide a reliable and fast procedure​ allowing the optimization of complex reactive systems.​ An existing CFD-based optimization concept was enhanced for reduced time-consumption​. First, two common multi-objective optimization algorithms available in modeFRONTIER, MOGA-II and NSGA-II, were compared, while MOGA-II outperforms NSGA-II and was chosen as a favorable optimization algorithm​. Next, several MM types will be compared to the CFD results. Furthermore, a FAST algorithm was considered for CFD-based optimization, which introduces MMs for virtual exploration and optimization within one algorithm loop.

BENEFITS - The use of the FAST algorithm in conjunction with radial basis functions (trained by 60 CFD results) was the most effective optimization strategy for the reactive flow system considered and will be used in further work for the optimization of more complex systems.