Rotor Stator Aeroelastic Analysis in 3D & Parallel Time Decomposition

    Work at CDI on unsteady aerodynamics of transonic cascades and 3D rotor/stator interaction during the 1990s led to the development of the RSA3D code. RSA3D is a comprehensive 3D viscous turbomachinery analysis capable of accommodating multiple blades per passage and multiple rotor and stator rows. The flow solver is implemented on an unstructured, deforming mesh with careful attention directed to the treatment of the moving fluid-structural interface to ensure adequate accuracy and enforcement of the governing conservation laws. A node-based discretization is adopted since this minimizes memory requirements and CPU time for a given mesh and also is amenable to both multigrid and implicit time marching acceleration methods. Structural models available for the aeroelastic analysis include general 3D isoparametric continuum finite elements suited for thick blades possibly containing cooling passages and shell elements intended for modeling thin compressor blades.

Contour plots of Mach number for supersonic rotor/stator interaction.

    The development effort that led to RSA3D involved two other critical developments supporting accurate and efficient modeling of turbomachinery flutter and forced response. The first entailed development of highly accurate interface routines for the mutual interaction of fluid dynamic models on the one hand and finite element structural models on the other. Once appropriate structural and fluid dynamics analyses have been developed for turbomachinery applications, extreme care must be taken with the numerical treatment of the interface joining the structure and the fluid since this is only path of communication between the two domains. An inconsistent coupling that fails to take into account the unique modeling aspects of the fluid and structural domains introduces spatial and temporal discretization errors that are incommensurate with the accuracy of the selected flow and structural models or time integration scheme. A second key step was development of novel Parallel Time Decomposition (PTD) routines that greatly accelerated the computation of periodic dynamic systems on parallel platforms. The PTD achieves nearly ideal parallel efficiency on systems of N processors by decomposing analysis of a periodic system in time rather than in space, minimizing latency and greatly simplifying parallel implementation of complex codes.

R37 Compressor Blade