Inverse Design of Pseudo Periodic Lattice Structures by Point Cloud Neural Network
Lattice materials have received significant attention due to their lightweight,high strength,and potential for mul-tifunctional design,and have been applied in fields such as aerospace and shipborne equipment.The structural efficiency of uni-formly distributed lattice structures is not optimal.Researchers have explored methods for designing non-uniform lattice structures for specific applications.However,the existing algorithms have limitations,such as difficulty in adapting to three-dimensional lat-tice structures,lack of an integrated framework,and insufficient design parameters.Taking advantage of the characteristic that fi-nite element analysis can be transformed into a matrix form combining position and stress,this study proposes a method for non-uniform lattice design based on point cloud neural network.The method takes the node stress matrix in finite element analysis as input,and the non-uniform lattice design parameter matrix as output,realizing the inverse operation of finite element analy-sis.The method utilizes edge convolution and max-pooling modules to address the interaction and disorder problems in represen-ting stress information in a matrix form,which can effectively solve the non-uniform design problem of lattice structures.
point cloud neural networkmaterial designinverse designfinite element
万方数据
维普
