A Parallel Flow Solver for Unsteady Multiple Blade Row Turbomachinery Simulations

Abstract

A parallel flow solver has been developed to provide a turbomachinery flow simulation tool that extends the capabilities of a previous single–processor production code (TURBO) for unsteady turbomachinery flow analysis. The code solves the unsteady Reynolds-averaged Navier-Stokes equations with a k–ε turbulence model. The parallel code now includes most features of the serial production code, but is implemented in a portable, scalable form for distributed–memory parallel computers using MPI message passing. The parallel implementation employs domain decomposition and supports general multiblock grids with arbitrary grid–block connectivity. The solution algorithm is an iterative implicit time–accurate scheme with characteristics–based finite–volume spatial discretization. The Newton subiterations are solved using a concurrent block–Jacobi symmetric Gauss–Seidel (BJ–SGS) relaxation scheme. Unsteady blade–row interaction is treated either by simulating full or periodic sectors of blade–rows, or by solving within a single passage for each row using phase–lag and wake–blade interaction approximations at boundaries. A scalable dynamic sliding–interface algorithm is developed here, with an efficient parallel data communication between blade rows in relative motion. Parallel computations are given here for flat plate, single blade row (Rotor 67) and single stage (Stage 37) test cases, and these results are validated by comparison with corresponding results from the previously validated serial production code. Good speedup performance is demonstrated for the single–stage case with a relatively small grid of 600,000 points.