Abstract: An improved parallel grid assembly method is proposed, aiming at promoting the efficiency of the grid assembly process, which is necessary for the rotor flow fields simulation with embedded grid frame. In this method, structured auxiliary mesh (SAM) with compressed storage is used for parallel hole cutting and interpolate points generation. The coordinate ranges of the grid vertices, which are needed for donor searching, are pre-exchanged among all processors. Another adaptive structured auxiliary mesh (ASAM) is used for the localization of the search start points, which cuts the memory consumption by shrinking spacing range for the grid generation based on the interpolate point message. The neighbor to neighbor and localized brute force donor searching methods are combined with above mentioned module. The geometry intersection judgment process is sidestepped when the searching path encountered with boundary face (wall, far field, grid partition interface) in donor searching procedure, and the complexity for coding is simplified. In essence, this method is suitable for multi-block grid assembly issue with any grid topology of structured, unstructured and Cartesian grid systems. The Navier-Stokes equations are solved with implicit dual-time step scheme, and several status of rotorcraft flow fields are simulated for validation purpose. Obtained results indicate that the donor searching speed with the improved methods is approximately 10 times faster than that with original one when coarsen structured auxiliary grid is used. The time consumption ratio of grid assembly to the flow field iteration can be reduced with low memory application. It is less than 1% when the flow field sub-iteration number is greater than 7. The present methods are useful for the numerical process based on the embedded grid system, especially for the rotor CFD simulation where several revolutions must be executed for periodical solutions.