ІНТЕГРАЦІЯ СУРОГАТНОГО МОДЕЛЮВАННЯ ТА ЕВОЛЮЦІЙНИХ АЛГОРИТМІВ У ПРОЄКТУВАННІ ВІДЦЕНТРОВИХ КОМПРЕСОРІВ

The paper presents a modern approach to optimizing the geometry of a high-pressure centrifugal compressor impeller using surrogate modeling and evolutionary algorithms. The main objective is to improve energy efficiency and reduce computational costs during CFD analysis by constructing an analytical model that approximates the relationship between aerodynamic characteristics and geometric parameters. The Kriging method (Gaussian Process Regression) combined with the NSGA-II multi-objective genetic algorithm was implemented in ANSYS Design Exploration. The compressor was parameterized by three key variables: impeller blade outlet angle, radial diffuser inlet blade angle, and splitter blade relative length. A central composite design of experiments was applied to limit the number of CFD simulations to 15 cases. The developed surrogate model demonstrated high approximation accuracy (coefficient of determination close to 1, minimal RMSE), confirmed by cross-validation. Optimization using NSGA-II produced a Pareto front illustrating the trade-off between polytropic efficiency and pressure ratio. Sensitivity analysis revealed that the impeller outlet angle has the greatest impact on efficiency, while splitter length and diffuser inlet angle play a stabilizing role. The results confirm the feasibility of applying surrogate modeling and evolutionary algorithms for turbomachinery optimization, enabling significant computational savings without compromising accuracy and providing flexibility in selecting optimal design solutions.