Robust aerodynamic optimization and design exploration of a wide-chord transonic fan under geometric and operational uncertainties

Axial compressors are inevitably affected by various uncertain factors in the process of manufacture and operation. These uncertainties obviously lead to reduced efficiency and large performance dispersion. However, researches on uncertainty quantification and robust design of compressors still faces severe difficulties due to the complexity of compressor structure and internal flow. This paper aims to present an automated and effective framework for uncertainty quantification and aerodynamic robustness optimization of axial compressor. The manufacturing error distribution is derived from the measurement data of machined fan blades, and the sparse grid-based polynomial chaos expansion method is used to propagate the uncertain factors and predict the probability density distribution of the fan performance. A novel surrogate model that combines a self-organizing mapping and a back-propagation neural network is constructed to explore and visualize the correlation between uncertainty parameters and performance responses. Robust aerodynamic design optimization is achieved based on the genetic algorithm. The results indicate that the coupled neural network model exhibits good accuracy for uncertain approximate modeling. Compared with the prototype fan, the optimized fan's mean isentropic efficiency and pressure ratio increase by 0.97% and 0.72%, respectively. The standard deviation of isentropic efficiency, pressure ratio, and mass flow rate decrease by 46.3%, 21.4%, and 15.2%, respectively. The present study provides a reference and exploration for uncertainty quantification and robust optimization of advanced refined multi-stage turbomachinery.