We present an optimization platform for turbomachineries with complex mesh configuration in a parallel computation environment. A continuous adjoint solver for 3-D viscous internal flow is coded under the same parallel framework as the flow solver. To meet the various permitted extents of reshaping on blade surface and cut down the computation cost in grid perturbation, a localized two-level mesh deformation method is developed based on Gaussian radial basis function (RBF). This method works efficiently for both the O mesh surrounding the blade and the O-H mesh blocks inside tip gap. In optimization of the transonic NASA Rotor 67 for higher adiabatic efficiency with a mass flow rate constraint, an adjoint sensitivity analysis is conducted. The relations between the design sensitive regions and physical phenomena in internal flow are discussed. Flow fields before and after adjoint optimization are investigated, including shock system, tip leakage flow and flow separation.
展开▼