Performance improvement of centrifugal compressors for automotive turbochargers using adjoint method

Abstract

Abstract A sensitivity calculation using adjoint method has a feature of low calculation cost because sensitivity information of all computational grid points can be calculated by only one numerical simulation. From this feature, it has been utilized as a calculation method of sensitivity information for a gradient-based optimization and this optimization procedure with adjoint method can be one of suitable ways for centrifugal compressor impeller which has a lot of design parameters such as meridional shape, blade angle and thickness. During the gradient-based optimization, mesh morphing, which moves the computational grid directly based on sensitivity information, is generally used for shape deformation. However, the mesh morphing is difficult to implement due to shape constraints which is particularly required for rotating blades, such as minimum blade thickness and rotationally axisymmetric shape. In this study, in order to avoid these problems, sensitivity information of each grid point was converted into that of conventional design parameters, and optimization was carried out. In order to verify the effectiveness of this method, two different optimizations were carried out for centrifugal compressor impeller. In the single-point optimization which focuses on efficiency improvement, maximum efficiency was increased by 3.5% compared to a baseline impeller. While, multi-points optimization which considers balance between flow capacity and efficiency showed that the maximum efficiency was improved by 0.6% compared to the baseline impeller while maintaining the same flow capacity.