Optimization of Compressor Variable Geometry Settings Using Multi-Fidelity Simulation

Abstract

Variable geometry blade rows are a common instrument to avoid compressor instabilities which occur especially at low- and full-speed operation of gas turbines. The operating settings of variable stator vanes (VSVs) are typically obtained from expensive and time consuming performance rig tests and are not known during the early design phase of a gas turbine. During preliminary design of the overall engine it is common practice to use default component characteristics based on considerable engineering experience. These can deviate substantially at off-design and often do not properly account for the impact of changes in component geometry. As a solution, multi-fidelity simulation often referred to as zooming or variable complexity analysis is applied. This proceeding facilitates a transfer of single component performance characteristics obtained in mid- or high-fidelity analysis to a full gas turbine system analysis based on lower resolution level. The purpose of this study is to present a multidisciplinary numerical optimization methodology to define ideal blade row staggering of variable compressor stator vanes during the early preliminary design phase using multi-fidelity simulation. The objective of the resultant multi-dimensional constraint optimization is to find the best solution for the entire gas turbine system for a set of discrete operating points. For the assessment a generic turbofan engine model is designed by taking into account top level engine requirements from an assumed airframe and flight mission scenario. Based on the performance calculation a full 3-D axial multistage high pressure compressor (HPC) is designed. The assumed design considerations are summarized and the modelling techniques are presented. The optimization of VSV staggering mentioned above is carried out by re-staggering the variable geometry blade rows of the high-fidelity HPC and run a full 2-dimensional through-flow calculation. Results are then automatically transferred to the 0-dimensional engine model to calculate the engine overall performance. A Pareto optimized blade row staggering is found by taking into account the surge margin and the specific fuel consumption of the entire engine system as objective functions of the optimization process. Simultaneously several constraints such as DeHaller numbers and diffusion factors are considered. The optimization process chain and the tool coupling are summarized and described in detail. The resulting VSV staggering for a set of discrete operating points is shown.