A Neural Network Method Application in Cloth Dynamic Simulation

This study proposes an algorithm integrated with a neural network method to simulating 3D cloth dynamics. The model built by an element finite method, is possibly improved and optimized by the neural network method. The finite element mesh generation has to be continuously controlled to meet the continuously changing geometric model in dynamic simulation. The neural network method is applied to dynamically generate the finite element meshes, whose nodal points and number need to be continuously redefined in order to archive a better mesh generation with a perfect computing speed and acceptable calculation error. The experimental results show that the algorithm has high effectiveness and satisfying simulation performance.