A novel optimization of blade angle distribution for supercritical CO₂ centrifugal compressor based on the artificial neural network

The centrifugal compressor is an important component of the supercritical carbon dioxide Brayton cycle, and the blade shape of the compressor is crucial for the efficient and stable operation of the system. However, it is still unclear how the continuous parameters that control the geometric shape of the blades affect the specific flow field structure, which poses a challenge for designers to design quickly and efficiently. This study selected the most critical blade angle curve for controlling blade geometry and conducted a quantitative analysis using Bezier curves. This work is based on Latin hypercube sampling using a translation propagation algorithm, and it constructs an artificial neural network surrogate model. The non-dominated sorting genetic algorithm II is used for global optimization, and the results reveal the detailed mechanism of the effect of blade angle curve changes on the flow field structure. After optimization, the isentropic efficiency of the compressor increased by 0.418%, and the total pressure ratio increased by 1.70%. A linear model was proposed to quickly evaluate the degree of separation of tip leakage flow caused by changes in blade angle curve. This indicates that the blade angle curve, as a geometric factor, constantly impacts the separation of leakage flow. This work provides the theoretical support for the design of compressor blade angle curves.