Computational fluid dynamics has enormous potential to influence the design and optimization of engineering systems; however, the error due to the computational mesh (discretization error) is often the largest source of numerical error. Automatic mesh adaptation can be used to generate an optimal mesh given a smooth indicator of error. Truncation error is the local source of discretization error and has been shown to be a good adaptation driver for structured grids [Roy, 20091]; however, the truncation error for general unstructured meshes is too noisy. A new method was developed that removes the excessive noise by interpolating the numerical solution to a smooth mesh matching only the control volume physical location and size [Phillips and Ollivier-Gooch, 20162]. The resulting truncation error estimate was used to drive an isotropic mesh adaptation procedure. In the current work, the truncation error estimation method is extended to include the effects of aspect ratio to create an anisotropic mesh metric. The new mesh adaptation metric is tested on an aniostropic Poisson solution comparing to a Hessian-based metric and a reconstruction error-based metric where the new truncation error-based mesh metric showed the best reduction in discretization error.
展开▼
