Surrogate-assisted CFD optimization of a radial turbine rotor
The design and optimization of radial turbine rotors play a critical role in improving the performance and efficiency of turbomachinery systems. Traditional optimization approaches often require extensive computational resources due to the need for high-fidelity CFD simulations of numerous design configurations. In this study, an integrated workflow combining automated 3D geometry generation, high-fidelity CFD evaluation, and machine-learning-assisted surrogate modeling was developed to efficiently explore the rotor design space. K-means clustering was employed to select representative training cases, and Gaussian Process Regression (GPR) was used to predict performance across untested configurations, guiding the search for optimal designs. The optimization successfully identified rotor geometries that increased isentropic efficiency by 2% relative to the baseline design. Flow visualizations revealed the aerodynamic mechanisms underlying the performance improvement, including improved flow guidance and reduced secondary losses. The study demonstrates that combining automated geometry, CFD simulations, and machine learning provides a powerful and computationally efficient approach for radial turbine rotor optimization, offering a practical pathway for achieving significant performance gains in turbomachinery applications.