Abstract
Proposed in this paper is a boundary-type approach as a way to increase efficiency of reverse engineering simulation of complex freeform digital models acquired from scanning. The developed method utilizes surface discretization nature of the boundary element method (BEM), and eliminates time-consuming NURBS surface/solid generation and volume discretization needed in domain-type simulation. Unstructured as-scanned image data is imported and regularized into a well-behaved BEM mesh in the computation. The developed BEM achieves a near O(N) computational complexity, by coupling the conventional BEM (of O(N3)) with the fast multipole acceleration. Numerical examples are presented to demonstrate its effectiveness and efficiency. Preliminary results show that the developed BEM can be a promising tool for many reverse engineering simulation applications involving highly complex organic shapes and large-scale models, e.g., the ones acquired from medical imaging.