|
| The generation method of innovative structures based on topology optimization and deep learning |
| Received:March 19, 2021 Revised:June 08, 2021 |
| View Full Text View/Add Comment Download reader |
| DOI：10.7511/jslx20210319001 |
| KeyWord:innovative structures topology optimization deep learning 3D reconstruction additive manufacturing |
| Author | Institution |
| 杜文风 |
河南大学 土木建筑学院, 开封 |
| 王英奇 |
河南大学 土木建筑学院, 开封 |
| 王辉 |
河南大学 土木建筑学院, 开封 |
| 赵艳男 |
河南大学 土木建筑学院, 开封 |
| 叶俊 |
浙江大学 建筑工程学院, 杭州 |
| 高博青 |
浙江大学 建筑工程学院, 杭州 |
|
| Hits: 790 |
| Download times: 289 |
| Abstract: |
| Computer-aided Design has been widely used in the calculation and analysis of structures,but there are still challenges in intelligently generating optimized innovative structures automatically.Aiming at solving this issue,a new method was proposed for the intelligent generation of innovative structures based on topology optimization and deep learning.Firstly,the topology optimization results under different optimization parameters were extracted using topology optimization analysis to produce the training picture set,and the training set labels were defined as the corresponding load cases.Then,the Least Squares Generative Adversarial Networks (LSGAN) deep learning algorithm was used to train for generating numerous innovative structures.Finally,the generated designs were evaluated and compared by building evaluation indexes and an evaluation system,and the optimal structure was selected according to the design requirements.Combined with an engineering case of designing a baseplate in a cast-steel support joint,the application process of the above method was described in detail.The 3D reconstruction and additive manufacturing techniques were subsequently applied to manufacture the structural models.The results show that the proposed intelligent structure generation method based on topology optimization and deep learning not only can generate innovative structures intelligently but also can optimize structural material consumption and mechanical performance further. |
|
|
|