| 引用本文: | 邓宗全,田大可,刘荣强,郭宏伟.构架式空间可展开天线结构优化参数预测模型[J].哈尔滨工业大学学报,2011,43(11):39.DOI:10.11918/j.issn.0367-6234.2011.11.009 |
| DENG Zong-quan,TIAN Da-ke,LIU Rong-qiang,GUO Hong-wei.Prediction model of optimal parameters for space deployable truss antenna[J].Journal of Harbin Institute of Technology,2011,43(11):39.DOI:10.11918/j.issn.0367-6234.2011.11.009 |
|
| 摘要: |
| 为了得到构架式空间可展开天线结构优化中目标函数与设计变量的解析表达式,基于BP神经网络建立一种天线结构优化参数的预测模型.根据天线背架的结构及神经网络的训练原理,构建对优化参数进行预测的网络模型;应用有限元软件ANSYS对优化参数进行数值计算,通过正交试验设计,得到BP神经网络的训练样本;调整传递函数、隐层节点数及训练算法,建立满足误差要求的优化参数的预测模型,利用检验样本对预测模型进行泛化能力检验.结果表明:网络预测值与有限元计算结果吻合较好,整体预测误差≤10%;并且模型运行时间短,仅需0.13 s.该模型能够较准确地预测结构的优化参数,为结构的优化设计提供了理论参考. |
| 关键词: 可展开天线 背架 BP神经网络 预测模型 正交试验设计 |
| DOI：10.11918/j.issn.0367-6234.2011.11.009 |
| 分类号:V443.4 |
| 基金项目:国家自然科学基金资助项目(50935002,11002039);高等学校学科创新引智计划资助项目(B07018);机器人技术与系统国家重点实验室(哈尔滨工业大学)自主研究课题(SKLRS200802C) |
|
| Prediction model of optimal parameters for space deployable truss antenna |
|
DENG Zong-quan1,2, TIAN Da-ke1,2, LIU Rong-qiang1,2, GUO Hong-wei1,2
|
|
1.State Key Laboratory for Robotics and System,Harbin Institute of Technology,150001 Harbin,China;2.School of Mechanical and Electrical Engineering,Harbin Institute of Technology,150001 Harbin,China)
|
| Abstract: |
| In order to obtain the analytical expression between objective functions and design variables for space deployable truss antenna during the structure optimization,prediction model of optimal parameters was built based on BP neural network.According to the truss structure and training principle of neural network,network model was established.Optimal parameters were calculated by using the finite element software ANSYS,and the training samples were obtained by orthogonal design.Adjusting the transfer function,the number of hidden nodes and training algorithm,prediction model was constructed,which satisfied the error requirement.Generalization capacity of neural network was tested by means of testing samples.The results show that predictive values of neural network agree well with finite element calculation results,which the relative error is within 10%.Moreover,the run time of model is shorter than that of ANSYS,which only needs 0.13 s.This model can predict optimal parameters more accurately and provide a theoretical reference for the structure optimization. |
| Key words: deployable antenna truss structure BP neural network prediction model orthogonal design |
