基于LVQ神经网络的飞行作动器故障诊断

为了对飞行作动器的故障进行有效辨识,使飞行员能够在更短时间内对故障进行处理,提出了基于自组织映射神经网络的学习向量量化算法;使用此方法在大步长采样下对飞行作动器的卡死和损伤故障进行训练和辨识,并尝试运用小波包技术分解小步长采样数据,结合自组织映射网络对分解后的数据进行分析;检验结果表明,大步长采样下,检测和分类效果令人满意,且具有良好的网络的泛化能力,而在小步长采样下,自组织映射网络不能有效区分故障类型,识别失败。

Fault-Diagnosis System of Flight Actuator Based on LVQ Neural Networks

In order to distinguish the faults of flight actuator efficiently and make the pilot be able to solve these faults in time,a method based on Learning Vector Quantization algorithm of self-organizing feature map is proposed.With this method the data for the fault of block and impairment are trained and distinguished in large-step and small-step sampling.And wavelet packet technique is utilized to decompose the small-step sample data tentatively.The results show that this method serves well for diagnosis and classification given large-step sample.But for small-step sampling,this method can not distinguish the different fault types efficiently.