• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Journal of the Institution of Engineers (Inida). Aerospace Engineering Division Board >Neural Partial Differentiation for Aircraft Parameter Estimation Under Turbulent Atmospheric Conditions
【24h】

Neural Partial Differentiation for Aircraft Parameter Estimation Under Turbulent Atmospheric Conditions

机译：湍流大气条件下飞机参数估计的神经偏微分

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

An approach based on neural partial differentiation is suggested for aircraft parameter estimation using the flight data gathered under turbulent atmospheric conditions. The classical methods such as output error and equation error methods suffer from severe convergence issues; resulting in biased, inaccurate, and inconsistent estimates. Though filter error method yields better estimates while dealing with the flight data having process noise, it has few demerits like computational overheads and it allows estimation of a single set of process noise distribution matrix. The proposed neural method does not face any such problem of the classical methods. Moreover, the neural method does not require parameter initialization and a priori knowledge of the model structure. The neural network maps the aircraft state and control variables into the output variables corresponding to aerodynamic forces and moments. The parameter estimation, pertaining to lateral-directional motion, of the research aircraft de Havilland DHC-2 with simulated process noise, is presented. The results obtained using the neural partial differentiation are compared with the nominal values given in literature and with the classical methods. The neural method yields the aerodynamic derivatives very close to the nominal values and having quite low standard deviation. The neural methodology is also validated by comparing actual output variables with the neural predicted and neural reconstructed variables.
机译：建议使用基于神经偏微分的方法，使用在湍流大气条件下收集的飞行数据估算飞机参数。输出误差和方程误差等经典方法存在严重的收敛问题。导致估算结果有偏差，不准确和不一致。尽管滤波器误差法在处理具有过程噪声的飞行数据时会产生更好的估计，但是它具有诸如计算开销之类的缺点，并且允许估计一组过程噪声分布矩阵。所提出的神经方法没有面对传统方法的任何此类问题。而且，神经方法不需要参数初始化和模型结构的先验知识。神经网络将飞机的状态和控制变量映射到与空气动力和力矩相对应的输出变量中。提出了带有模拟过程噪声的德哈维兰德DHC-2型飞机的与横向运动有关的参数估计。将使用神经偏微分法获得的结果与文献中给出的标称值和经典方法进行比较。神经方法产生的空气动力学导数非常接近标称值，并且具有非常低的标准偏差。通过将实际输出变量与神经预测变量和神经重构变量进行比较，也可以验证神经方法。

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