Abstract
Mesh deformation based on radial basis functions (RBF) is widely used in CFD simulations with moving boundaries due to its simplicity, generality, and robustness, among which selecting an optimum reduced set from all boundary nodes is a crucial data reduction step for fewer support points. However, there remains a great need to reduce the computation costs for large-scale mesh. To solve this issue, a parallel point selection strategy and incremental LDLT decomposition method are proposed. With all boundary nodes distributed to all processors on average, the displacements of boundary nodes are calculated at local ones, and the nodes with local maximum error in divergent processors are added to the support point set in parallel. Besides, the matrix decomposition and solution of weight coefficients are conducted with full uses of corresponding results in the last iteration in incremental LDLT decomposition method. The calculation costs of conventional methods and proposed approaches are theoretically analyzed and compared. Three typical test cases of different mesh scales and motion modes are carried out for validation, including the rigid motion of a NACA0012 airfoil, the bending of an ONERA M6 wing, and the pitch of a missile. The results demonstrate that the proposed approaches greatly improve the efficiency of the data reduction procedure compared with the conventional methods, with similar deformation effect represented by support point distribution and deformed mesh quality.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call