Through-flow solvers have historically played a very prominent role in the design and analysis of axial turbomachinery. While three-dimensional, Full Navier-Stokes (FNS) CFD is taking an increasing larger role, quasi-3D through-flow methods are still widely used. Automated optimization techniques that search over a wide design space, involving many possible variables, are particularly suitable for the computationally efficient through-flow solver.
Pressure-based methods derived from CFD solution techniques have gradually replaced older streamline curvature methods, due to their ability to capture flow across a wide range of Mach numbers, particularly the transonic and supersonic regimes. The through-flow approach allows for the solution of the three-dimensional problem with the computational efficiency of a two-dimensional solution. Since the losses are explicitly calculated through empirically based models, the need for detailed grid resolution to capture tiny flow entities (such as wakes and boundary layers) is also greatly reduced. The combined savings can result in computational costs as much as two orders of magnitude lower than full 3D CFD methods.
A state-of-the-art through-flow solver has several features that are crucial in the design process. One of these is the ability to run in both a design and an analysis mode. Also important, is the ability to generate solutions where critical components are solved using 3D FNS, while others are run using a through-flow method. Other desirable features in a through-flow solver are: an advanced equation of state, injection and extraction ability, the handling of arbitrary (non-axial) shapes, and a link to a capable geometry generation engine.
Through-flow solvers represent a unique mix of higher order numerical methods (increasingly CFD-based) coupled with empirically derived models (generally meanline based). The combination of these two methods in one solver creates a particularly challenging programming problem. This paper details the techniques required to effectively generate through-flow solutions. Special attention is given to an improved off-design loss model for compressors, as well as a transonic loss model needed for high-speed compressor and turbine flows. Validation with recognized test data along with corresponding 3D FNS CFD results are presented.