UNCERTAINTY QUANTIFICATION AND ROBUST OPTIMIZATION FOR THROUGHFLOW AXIAL COMPRESSOR DESIGN

Throughflow axial compressor design relies on empirical correlations to estimate the total pressure losses and deviation generated by a blade row. Those correlation data have been obtained with given measurement uncertainties that are usually not taken into account by meridional design tools or throughflow solvers. Uncertainty quantification techniques such as Monte-Carlo sampling or stochastic collocation can help to introduce uncertainties within those numerical tools and quantify their effects on the compressor performance at cheap computational costs. The throughflow solver used for this investigation is ACPreDesign. The code has been recently implemented at VKI and solves the non-isentropic radial equilibrium equations for axial compressors. Realistic uncertainties are introduced from standard loss correlations. The implementation of non-intrusive uncertainty quantification methods is presented and validated on a single stage low pressure compressor geometry studied at VKI. The results are analyzed in terms of efficiency probability density functions which represent the resulting uncertainties on efficiency due to non-deterministic loss correlations. Finally, ACPreDesign is coupled to the VKI optimizer CADO in order to approach the problem of design under uncertainty also referred to as robust design. To address that issue the paper focuses on a robust solidity optimization of a single rotor configuration.