Title

Session

Authors

Abstract

Centrifugal fan stages are used in many applications where relatively high pressure rise is required in compact size. The application can vary from household appliances, industrial to airconditioning and data centre cooling applications. In many of these applications the fan stages are required to meet multi-point requirements in terms of both pressure rise and efficiency.
In this paper we present the design of a centrifugal fan stage for multi-point and multi-objective requirements. The paper starts from basic requirements for the design such as pressure rise, flow rate and rpm at 2 operating points. The fan stage needs to meet a maximum torque requirement set by the motor power. An initial flow path for the stage is generated by using a meanline code by using one of these operating points. This meridional geometry is then used in a 3D inverse design method in which the 3D blade geometry of the impeller is generated for a given distribution of loading ( or pressure jump) on the blade. The aim of this initial design is to create a baseline for multi-point optimization. For optimization the meridional geometry of the impeller and inlet nozzle are parametrized. The blade loading is also parametrized on one streamwise location in order to meet the requirements for the blade to be manufactured by a metal pressing process. In total 14 parameters are used for meridional shape and blade shape.

By using a wide range of variation of these 14 design parameters and Design of Experiments method a design matrix is generated for about 120 fan geometries. These geometries are then run in steady 3D RANS code for the two operating points. The CFD set up used tries to represent the measurement set up and hence covers a suitably large inblock and outblock boundary. Important objectives such as efficiency, pressure rise and torque are then extracted from the CFD solution at the two operating points. These results are then used together with the design matrix in a surrogate model based Kriging method. A multi-objective Genetic Algorithm (MOGA) is then run on the surrogate model to find trade offs between the efficiency and pressure rise at the two operating points subject to constraints on maximum torque and also slope of the pressure rise change between the two operating points. The design with the best efficiency at the two operating points is then exported from the surrogate model and run in CFD. The CFD results confirm significant improvement in efficiency over the baseline design and also show that the actual CFD values for efficiency, torque and pressure rise at the two operating points are very close that predicted by surrogate model. Hence confirming the accuracy of the surrogate model used for design optimization. This process can be used to speed up the design optimization of centrifugal fan for multiple operating points under industrial time scales.