An Iterative Method for Blade Profile Loss Model Adaptation Using Streamline Curvature

Abstract

The various incidence, deviation, and loss models used in through-flow analysis methods, such as streamline curvature, are nothing more than statistical curve fits. A closer look at public domain data reveals that these statistical correlations and curve fits are usually based on experimental cascade data that actually display a fairly large scatter, resulting in a relatively high degree of uncertainty. This usually leads to substantial differences between the calculated and actual performances of a given gas turbine engine component. Typically, matching calculated results from a through-flow analysis against experimental data requires the combination of various correlations available in the public domain, through a very tedious, complex, and time consuming “trial and error” process. This particular study supports the view that it might actually be much more time effective to “adopt” a given loss model against experimental data through an iterative, physics-based approach, rather than try to identify the best combination of available correlations. For example, the well-established “Swan’s model” for calculating the blade profile loss factor in subsonic and transonic axial flow compressors depends strongly on approximate correlations for calculating the blade wake momentum thickness, and therefore represents such a case. This study demonstrates this by looking into an iterative approach to blade profile loss model adaptation that can provide a relatively simple and quick, but also physics-based way of “calibrating” profile loss models against available experimental data for subsonic applications. This paper presents in detail all the analysis necessary to support the above concept and discusses Swan’s model in particular as an example. Finally, the paper discusses the performance comparison of a two-dimensional, streamline curvature compressor model against experimental data before and after the adaptation of that particular loss model. This analysis proves the potential of the simulation strategy presented in this paper to “adopt” a given loss model against experimental data through an iterative, physics-based approach.