| 摘要: |
| 利用人工神经网络的高度并行和非线性映射的功能,提出用多层BP网络描述任一弹性结构的应力、位移等量和结构设计变量之间的映射关系,建立了基于人工神经网络的结构全局近似分析方法。利用复合形优化思想对BP网络进行设计和学习,建立结构近似分析的神经网络模型。桁架算例分析表明,利用神经网络进行近似分析的结构优化设计具有较高的设计精度和适应性。 |
| 关键词: 神经网络 结构近似分析 复合形法 结构优化 |
| DOI： |
| 投稿时间：2000-08-03 |
| 基金项目:电科院预研基金资助项目(编号:DJ7.3.9.2)。 |
|
| Structural Optimization Using Approximation Analysis by Neural Networks |
|
Li Wenyong, Li quanyong
|
| (Dept. of Electronic Machinery & Traffic Engineering, Guilin Institute of Electronic Technology, Jingjilu, Guilin, Guangxi, 541004, China) |
| Abstract: |
| The mapping between the elastic structure's stress and displacement and it's design variables with multi-layer BP networks is described using the high parallelity and nonlinear mapping of the Artificial Neural Networks (ANNs)'. An approach to the global Structural Approximation Analysis (SAA) is developed by ANNs.The idea of multiplex shape optimization method is introduced to design and train the neural networks, and then establish the ANNs' model for SAA. The example of truss shows the structural optimization using approximation analysis by ANNs has high precision and adaptability. |
| Key words: neural networks structural approximation analysis multiplex shape method structural optimization |
