电子设备BIT状态的神经网络预测

首先对基于ＬＭ 算法的神经网络预测性能进行研究。相对快速ＢＰ网络而言，其预测精度高、收敛速度快。然后提出将该方法应用于电子设备ＢＩＴ输出及相关量的状态预测，对存在于航空电子设备中的压力、温度等环境应力的典型变化曲线进行了预测，并在环境应力影响下的ＢＩＴ状态综合预测中得到验证。结果表明，利用时空两方面信息进行状态预测和综合分析是一条提高ＢＩＴ诊断能力、降低虚警的重要思路。

Electronic Equipment BIT Condition Predicted by Neural Network

The predictive property by LM neural network is studied first. Compared with the fast BP network, this method is more accurate and the constringed speed is more quick. It is applied to predicting the BIT outputs and relevant variables such as typical pressure, temperature and other environmental stress curves existing in avionics systems. The conclusion is proved by the synthetic condition prediction of BIT influenced by environmental stresses. The results show that the condition prediction and synthetic analysis making use of time/spatial information is one of the important methods of improving BIT diagnostic ability and reducing false alarm.