Generation method of three-dimensional joint solid models based on generative adversarial networks

XIA Zhuang,DU Wenfeng,HAN Leyu,GAO Boqing,DONG Shilin

DOI: https://doi.org/10.14006/j.jzjgxb.2022.0517

2023-01-01

Abstract:The intelligent generation of 3D solid models is the core aspect of achieving structural intelligent design.In order to overcome this key technology and realize the structural intelligent design,a method called 3D-JointGAN was proposed,which combines 3D generative adversarial networks(GAN)and reverse engineering technology.Firstly,the basic idea,relevant theories and specific implementation process of this approach were introduced.Next,it was applied to the intelligent generation of a practical three-branch cast-steel joint,and the mechanical properties of the representative joints selected through evaluation were compared with those of the hollow spherical joint,bionic joint and topology-optimized joint.Finally,reduced-scale models of the representative joints were manufactured using 3D printing technology to verify the manufacturability of intelligently generated results.The research results demonstrate that the joint model generation method based on GAN exhibits strong generation and optimization ability,enabling it to generate a variety of innovative and highly realistic 3D solid models.Furthermore,the representative joints selected after evaluation demonstrate improved mechanical properties.Compared to the initial joint,the masses of representative joints a,b and c are reduced by 22.42％,26.82％and 22.62％,respectively.The maximum displacement is reduced by 43.71％,38.34％and 42.49％,and the maximum equivalent stress is reduced by 8.29％,5.28％and 6.94％,respectively.