作者: Zhao Li , Haixin Chen , Yufei Zhang , Song Fu
DOI: 10.2514/1.C033396
关键词:
摘要: G RID-CONVERGENCE study always plays a fundamental role in verifying the reliability of computational fluid dynamics (CFD) methods predicting aerodynamic performance. Roy [1] reviewed code and solution verification procedures summarized generalized Richardson extrapolation as one error evaluation approaches. Recently, lot such efforts [2–4] have been carried out on accuracy either numerical schemes or physical models past decades. Vassberg Jameson [5] investigated grid-convergence trends two-dimensional Euler solutions flow around NACA 0012 airfoil. Computation was run until residual reached machine-level zero. Three well-known solvers, including FLO82, OVERFLOW, CFL3D, were compared. In their paper,Vassberg listed two problems. First, convergence found to stall at values well above zero when Roe’s flux difference scheme (FDS) [6] adopted for OVERFLOW CFL3D. The authors had switch other instead. Second, so-called hyperconvergence observed drag prediction transonic nonlifting case. Note that all three solvers showed first-order accurate lifting condition.When angle attack (AOA) zero, trend CFL3D raised second order subcritical cases. solver FLO82 even surprisingly gave an high 2.631, hyperconvergence, according authors. Motivated by aforementioned problems, based in-house NSAWET with FDS is presented. purpose check whether will happen CFD codes seek possible explanations. II. Numerical Methods