Block Lower-Upper Symmetric Gauss-Seidel Scheme for Adjoint Solvers

摘要

The block lower-upper symmetric Gauss-Seidel (BLU-SGS) scheme is adopted to develop an implicit discrete steady adjoint solver. In the BLU-SGS approach, the inviscid flux is approximated by first order AUSM+-up scheme, and Jacobian matrix is evaluated by auto-differentiation(AD) method. The BLU-SGS and original LU-SGS are compared with ONERA M6 wing case. Computational results indicate that the BLU-SGS converges 9 times faster than original LU-SGS scheme with extra memory to store the Jacobian matrix. At current stage, the BLU-SGS solver is implemented for structured grid and validated only for inviscid case, but the solver shows great potential for solving adjoint equations.