Adding-point strategy for reduced-order hypersonic aerothermodynamics modeling based on fuzzy clustering

摘要: Reduced order models(ROMs) based on the snapshots CFD high-fidelity simulations have been paid great attention recently due to their capability of capturing features complex geometries and flow configurations. To improve efficiency precision ROMs, it is indispensable add extra sampling points initial snapshots, since number achieve an adequately accurate ROM generally unknown in prior, but a large reduces parsimony ROMs. A fuzzy-clustering-based adding-point strategy proposed fuzzy clustering acts indicator region which ROMs relatively low. The method applied construct for benchmark mathematical examples numerical example hypersonic aerothermodynamics prediction typical control surface. can 34.5% improvement than estimated mean squared error algorithm shows same-level accuracy.