摘要
应用BP神经网络开展高超声速飞行器嵌入式大气数据系统(FADS)算法研究。采用自主研发CACFD软件平台求解欧拉方程,计算获得飞行器头部的压力分布作为神经网络样本训练的输入,对应的来流状态,如静压、马赫数、迎角和侧滑角作为样本的目标训练神经网络,建立基于BP神经网络FADS求解算法,并进行测试研究。研究表明,基于神经网络技术的FADS算法具有较好的鲁棒性和求解精度,实时性强,是一种非常有效的求解算法。研究结果得出,一定样本数范围内,FADS的求解精度随着样本数增加而提高;算法的平均误差随着测压点的增加而减小;包含大锥角位置测压点的布点组合,明显比只有小锥角测压点布点组合的求解结果平均误差要小;去掉顶点测压点,对算法的求解结果影响不大;1%压力测量误差时,神经网络泛化性能表现非常稳定。
In this article,through the engineering method and Computational fluid dynamics,the aerodynamic configurations for the solar powered buoyancy-lifting vehicle in the near-space were investigated. An aerodynamic configuration with high lift to drag that was obtained by optimization design. The differences between results from the two methods were analyzed. Experiment study on the aerodynamics configuration was done by cable mounting wind tunnel testing technology,and testing methods in wind tune experiment of high aspect ratio with two fuselages configuration were building.Aerodynamic performance of the test model in pitching and lateral were analyzed in different Reynolds number. Conclusion of this investigation showed that wind tunnel experiment results and computational results were according well and the configuration design was validated,experiment results could give technology support to solar powered buoyancy-lifting vehicle system engineering as well.
出处
《飞机设计》
2015年第6期1-7,共7页
Aircraft Design
基金
国家自然科学基金重点项目(90816026)
