On the Development of an Efficient Parallel Hybrid Solver with Application to Acoustically Treated Aero-Engine Nacelles

NOISE-CON 2006 Abstract The ability to predict fan noise within complex 3-D aircraft geometries containing nacelle liners is critical for design of quiet aircraft. To this end, several 3-D finite element codes have been developed both in the United States and abroad. These codes are computationally too expensive for application to realistic 3-D geometries and will require supercomputing power as well as more efficient parallel algorithms to obtain reliable results within a reasonable wall clock time. Most of the computational time is spent solving the system of complex algebraic equations associated with the finite element model. Although iterative solvers are faster than direct solvers for problems in 3-D, iterative solvers generally lack robustness when used with nacelle liners. Consequently, the equation solving techniques have been based on direct solvers. In this study, a hybrid solve strategy that is based upon a combination of iterative and a direct, sparse, solve is implemented and linked to a 3-D finite element model that contains nacelle liners. The parallel performance, efficiency, and accuracy of this new method are tested on a supercomputer and the methodology is shown to give super-linear speedup over nearly 400 CPUs and upward of 25 million complex equations are solved.