• 主题
高级搜索  >
历史搜索 清空历史
首页> 美国政府科技报告 >Study on system identification of a distributed parameter system using feed-forward neural networks
【24h】

Study on system identification of a distributed parameter system using feed-forward neural networks

机译：基于前馈神经网络的分布式参数系统辨识系统研究

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Artificial neural networks have been applied to several system identification problems. However, most of those systems exclude dynamics or are lumped parameter systems. System identification of a distributed parameter system is considered in this report. The advantage of using neural networks lie in their ability to learn the process dynamics from the observations of the gross behaviour of the process, without a mathematical model. The process consists of a metal rod of constant cross section and constant (temperature invariant) specific heat, density and thermal conductivity being heated at one end and maintained at a constant temperature at the other end. 4 simulation runs were carried out to generate data on the dynamics of the process. 400 training instances were created from this data. The neural networks output changes in temperatures at sampling points over the sampling interval, based on the knowledge of the state at a given time, and the heat input to the rod. The system was identified well using feed-forward neural networks. Temperature at the heated end of the rod was not predicted as accurately as the other sampling points. When this sampling point was treated separately, using an additional input, the change in temperature at that point was also predicted accurately. Linear activation functions may be used to estimate the change in temperature at the other points

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号