基于鸟击事故征候预测的通用航空安全研究

作为民航运输的两翼之一,通用航空安全水平直接影响民机系统的安全。目前，对通航以及鸟击事故征候进行预测的研究较少,本文根据收集到的美国从事通用航空活动发生鸟击事故征候安全状况数据,采用长短期记忆(long short-term memory, LSTM)神经网络模型对鸟击事故征候数据进行训练和预测。实验结果显示，与传统模型相比, LSTM模型具有更好的预测效果,精确度更高。基于此,提出了预测稳定性更好的LSMT-均方根误差(LSTM-root mean square error, LSTM-R)模型，为通用航空鸟击事故征候预测提供了手段和方法,加强了通用航空安全管理。

Reduced dimension of derivative based updating STAP algorithm for airborne forward looking radar

To deal with the disadvantage of large computational load for derivative based updating(DBU),a method named reduced dimensional DBU algorithm is proposed.The algorithm uses the matrix blocking reverse theorem to get a revised matrix,then uses the reduced dimensional space time adaptive processing(STAP)method of 3DT.This method can decrease the computational load and reduce the number of independent identically distributed(IID) samples,and it approximates to the original optimal DBU method in performance.The results of experiment show that the method can still obtain a good performance near the optimal processing under clutters with some uncertain errors.