Large-scale parallel simulation with implicit conservative overlapped grid technique for compressor flow

Although significant advances have been made in full-annulus simulation of rotating stall in rotors, little work has been devoted to multi-row compressors in this field. Conducting efficient conservative data exchange at rotor/stator interface (R/S interface) in the full-annulus simulation of flow in multi-row compressors in the massively parallel environment has been a challenge. This paper aims to develop a large-scale parallel simulation method with a conservative overlapped grid technique, which is used to simulate unsteady flow including rotating stall flow in multi-row compressors. A general parallel implicit conservative interpolation algorithm for data exchange at the R/S interface is proposed, which has few restrictions on the grid configuration. The parallel strategy is optimized, the space complexity for processing data exchange at the R/S interface in the massively parallel environment is greatly reduced. The development of rotating stall in NASA Stage 35 is investigated, it is found that the parameter conservation at the R/S interface calculated by conservative interpolation is better than that calculated by distance based interpolation, no matter whether the parameter is mass flow, momentum or energy. What is more, the conservative interpolation performs better when simulating the propagation of stall cells across R/S interfaces compared with distance based interpolation. The proposed algorithms are applied to the simulation of unsteady flow in a 9-stage compressor with 1.3 billion grid cells using 4096 processes. The space complexity for overlapped grid assembly is reduced from O(1.2e7) to O(3.0e4), the proportion of time for processing R/S interface in the whole solving process is 7.51%, the difference between inlet mass flow and outlet mass flow is 0.08%, the interaction between blade rows is correctly captured.