Promoteur : Prof. G. Coussement

Résumé :

The flow performances of axial compressors are subject to several geometrical or operational uncertainties which can be correlated. These uncertainties can be re-sponsible for important variations in the performances of the designed product. There-fore, the conception of a design insensitive to the tolerances and the variation of the operational condition is a major concern in industry. This thesis proposes several im-provements of an existing methodology to allow its industrial use, which includes tak-ing into account the nature of real industrial uncertainties and allowing design optimi-zation robust to these uncertainties.

This thesis improves an existing uncertainty propagation methodology in order to cope with several simultaneous uncertainties: First the method was coupled with a full CFD chain, including the mesh generation, in order to allow the simultaneous treatment of operational and geometrical uncertainties. Then, the Non-Intrusive Prob-abilistic Collocation Method was coupled with a sparse grid technique, which allows the treatment of several simultaneous uncertainties. A Principal Component Analysis has been coupled to this uncertainty quantification tool to cope with the problematic of correlated uncertainties. All these developments lead to a sufficient tool maturity for an industrial use.

Then a pragmatic robust design methodology was developed. This methodology allowed to determine a robust optimum for an axial compressor. It was highlighted that this optimum is different from the deterministic one, meaning that considering the uncertainties during the design phase may have an impact on the resulting geometry. Moreover, a methodology for the selection of an optimal tolerancing of an existing design has been successfully used.

The developed tool for uncertainty quantification and robust design has been ful-ly implemented and tested in FINE/Design3D software edited by NUMECA Interna-tional. Therefore, it offers all the advantages of a top of the art CFD software tool, leading to a facilitated use in industry.