In this paper, an aerodynamic optimization of the radial inflow turbine for a 100kW-class micro gas turbine is conducted based on the metamodel-semi-assisted idea. The idea is applied by first using the metamodel as a rapid exploration tool and then switching to the accurate optimization without metamodel for further exploration of the design space [1]. The non-dominated sorting genetic algorithm (NSGA-II) is used to drive the optimization process and the BP neural network is used to construct the metamodel.

The optimization of this radial inflow turbine is divided into two parts, the stator optimization and the rotor optimization. The stator optimization is based on the accurate optimization strategy. The minimum total pressure loss of the stator and the maximum isentropic total-to-static efficiency of the stage are considered as the objective functions with constant mass flow rate as a constraint. The rotor optimization is conducted through the metamodel-semi-assisted idea. The maximum power output and isentropic total-to-static efficiency of the stage are considered as objective functions while keeping the mass flow rate to be constant.

The accurate optimization system is demonstrated to be effective for the stator optimization, and the total pressure loss is reduced by 11.6% while the mass flow rate variation is less than 1%.

The rotor optimization is conducted based on the metamodel-semi-assisted optimization and the results confirm the effectiveness of this new idea. The output power of the rotor increased by 1.5%, the isentropic total-to-static efficiency of the stage increased by 1.19% and the mass flow variation is less than 1%.

This content is only available via PDF.
You do not currently have access to this content.