Abstract
Large Eddy Simulation (LES) of turbomachinery stages has been recently brought to attention due to its potential increased prediction fidelity and its reduced dependency to modeling. Such simulations are however often very CPU intensive, with potentially long return times and only possible for reduced periodic sectors. For real applications, such limitations are prohibitive for a daily use in a design phase. Indeed, most industrial turbomachinery applications rely on designs where at least one of the blade rows has a prime number of blades. Full 360° simulations are in such a case required for appropriate flow dynamics predictions, which implies prohibitive computational costs although recent demonstrations prove these feasible. To make LES affordable in an industrial context, it is clearly necessary to find ways to reduce its cost and return time, one approach being the reduction of the computational domain size.
The Profile Transformation Approach (PTA) is one of such specific methods that allows to simulate down to a single blade passage per blade row, thus decreasing the domain size of the problem and its CPU cost. PTA has been devised and validated in a URANS context and its limits are well known in this specific context. In terms of development and implementation in a code, PTA essentially consists in re-scaling the flow field at the rotor/stator interface to comply with the geometrical constraints on both sides of the interface since these often have different angular extents. Thanks to this flow re-scaling, periodic flow conditions can be applied on the azimuthal limits of both domains while retaining only one passage per row. In the following, the method is assessed in the context of fully unsteady LES simulations in an attempt to identify generated approximations and errors. This LES approach is then used to address a set of cases of increasing complexity ranging from the academic problem focusing first on the convection of a vortex across an interface and finishing with simulations of industrial relevance.
