An unstructured dynamic mesh adaptation and load balancing algorithm is developed for the efficient computation of three-dimensional unsteady inviscid flow fields on parallel distributed memory machines. Time integration is achieved by using a second-order accurate point Gauss-Seidel relaxation method with dual time stepping. A cell-centered finite-volume discretization is used in conjunction with the Roe`s flux-difference splitting. The flow solver is parallelized by using the MPI library.
For the present cell-based solver, a new face data structure is suggested for fast and continuous multi-level data transfer such that additional post-processing and re-ordering are not required after mesh refinement and coarsening. The parallel performance and the mesh adaptation algorithm were tested for steady transonic flow around an ONERA M6 wing. The unsteady dynamic mesh adaptation and load balancing were validated for a shock tube and oscillating NACA0012 and F-5 wings.
Applications were also made for the simulation of helicopter rotor blades in hover and in forward flight. To calculate the unsteady rotor wake more efficiently, the flow field was divided into a moving zone rotating with the rotor blades and a stationary zone containing the wake. A contact boundary is constructed between the two zones, and a sliding mesh algorithm is developed for the proper convection of the flow variables through the boundary. A `quasi-unsteady` mesh adaptation algorithm was adopted to enhance the spatial accuracy of the solution. In order to handle the blade motion due to rotor trim in forward flight, a deforming mesh algorithm and spring analogy were adopted in conjunction with edge collapsing to remove highly skewed cells at large blade deflection.
Validation was made for hovering Caradonna and Tung rotor, and the results are compared with the experiment. In forward flight, comparison was made with the AH-1G rotor tested by NASA. The present method was also applied to rotor-fuselage interaction problem by solving the flow around the Georgia Tech experimental configuration. It was found that the present method is efficient and robust for simulating rotor blades in forward flight and rotor-fuselage interactions.