Highly Parallel Surface Crack Simulation

We show a parallel surface crack generation algorithm based on physical method that runs mostly on the GPU. According to existing physically correct surface crack generation, our algorithm firstly proposes a new data structure from the given set of points and triangles, and then computes where the crack should show up, and finally defines how the mesh´s stress tensor updates to generate surface cracks. A key insight of our algorithm is constructing a novel data structure on the GPU, which makes it faster to obtain the neighborhood information of each vertex and triangle. Moreover, a novel method that updates the stress tensor and computes Separation Tensor in real-time on the GPU has been presented. And we got that our algorithm runs almost two orders of magnitude faster than previous CPU algorithms via a large number of testing. In order to support this, images of crack teapot and apple generated from our algorithm were shown.