Session: 36-04 UQ & Sensitivity Analysis - Part 1

Submission Number: 178015

Optimization Strategy for Planar Blade Cascade Based on Uncertainty Quantification of Reynolds Averaged Navier Stokes Turbulence Model 

With the development of turbomachinery toward higher efficiency and wider operating ranges, the Reynolds-Averaged Navier-Stokes (RANS) method is widely used in aerodynamic design, but its inherent model-form uncertainty leads to deviations between numerical predictions and actual performance, resulting in "pseudo-optimal solutions" in traditional deterministic optimization. To address this issue, this study establishes a systematic optimization framework for planar blade cascades considering RANS turbulence model uncertainty. Taking the LS89 highly loaded transonic turbine cascade as the research object, the framework integrates the non-uniform perturbation uncertainty quantification (UQ) method, CAD-based parameterization technology, Kriging surrogate modeling, and genetic algorithm (GA). The optimization objective is to minimize the worst-case entropy production while ensuring the flow turning capability. Validation results show that the non-uniform perturbation UQ method effectively constructs physically meaningful uncertainty intervals, reducing the prediction deviation of peak Mach number from 3.5% to 0.6%. Comparative analysis of two optimized profiles (OPT1 from deterministic optimization and OPT2 from robust optimization considering UQ) indicates that although both achieve significant entropy generation reduction compared with the original LS89 cascade (17.7% for OPT1 and 16.4% for OPT2), OPT2 exhibits superior robustness. Under turbulence model perturbations, OPT2 maintains a narrow uncertainty band of isentropic Mach number distribution, suppresses flow separation, and stabilizes pressure recovery, while OPT1 is sensitive to model uncertainties with obvious flow separation risks. Sensitivity analysis identifies key design variables: suction-side throat thickness and trailing-edge thickness dominate core flow loss, and outlet metal angle enhances robustness without compromising baseline performance. This framework enriches the theoretical system of uncertainty-based turbomachinery optimization and provides technical support for high-reliability blade cascade design.

Presenting Author: Tongxi Li Xi’an Jiaotong University

Presenting Author Biography: Li Tongxi is a Ph.D. candidate jointly affiliated with the Department of Fluid Machinery and Engineering at Xi'an Jiaotong University and Taihang National Laboratory, China. He earned his Bachelor of Engineering degree from Xi'an Jiaotong University in 2022 and obtained his Master of Engineering degree in 2025. His current research primarily focuses on the uncertainty quantification of turbulence modeling and its applications in the optimization design of turbomachinery.

Authors:

Tongxi Li Xi’an Jiaotong University
Jiayi Zhao Taihang National Laboratory
Zhu Huang Xi’an Jiaotong University
Zhiheng Wang Xi’an Jiaotong University
Guang Xi Xi’an Jiaotong University