Uncertainty quantification of geometric deviation effects on compressor performance based on an optimized support vector regression

The influence of leading-edge radius deviation on the aerodynamic performance and dissipation loss of compressor rotor blades is investigated in this study. A surrogate model is constructed based on the Genetic Algorithm Assisted Heterogeneous Improved Dynamic Multi-Swarm Particle Swarm Optimization) algorithm optimized support vector regression method to conduct an uncertainty quantification analysis. Additionally, dissipation source region segmentation is employed to examine the effect of deviation variation on dissipation in different regions. The results indicate that, under various conditions, the probability of performance degradation caused by leading-edge radius deviation exceeds 90%. Under the peak efficiency condition, the performance parameters in the upper half of the blade span (above 50% blade height) are more sensitive to leading-edge radius deviation, particularly in the blade tip region. In this region, the deviation primarily intensifies dissipation loss by enhancing leakage flow strength and altering the shock-leakage vortex interaction location. In near-stall conditions, the sensitivity of performance parameters to leading-edge radius deviation is lower across the entire blade span compared to peak efficiency conditions. In practical manufacturing, efforts should be made to minimize a reduction in leading-edge radius at the blade tip and to avoid excessive leading-edge radius at other blade spans to reduce performance losses.