基于模态分解的运载火箭惯性器件故障诊断方法  

Research on Fault Diagnosis Method for Heavy Launch Vehicle Based on Mode Decomposition Method

作　　者：周虎[1,2] 段然 屈辰[1] 李刚 王苑瑾 Zhou Hu;Duan Ran;Qu Chen;Li Gang;Wang Yuan-jin(Beijing Aerospace Automatic Control Institute,Beijing,100854;National Key Laboratory of Science and Technology on Aerospace Intelligence Control,Beijing,100854)

机构地区：[1]北京航天自动控制研究所,北京100854 [2]宇航智能控制技术国家级重点实验室,北京100854

出　　处：《导弹与航天运载技术》2020年第4期85-90,共6页Missiles and Space Vehicles

摘　　要：针对运载火箭电气系统部分缓变输出故障识别困难的现状,以捷联惯组输出为例,提出了一种基于数据驱动的故障实时诊断方法。该方法利用模态识别算法将指定的惯组输出数据分解为多个模态分量,然后通过频率判别法实现模态分量的自适应选择,并将模态分量映射为故障特征向量,最后采用概率神经网络定义故障分类器,从而形成对故障的可信诊断能力。仿真结果表明,该方法能够在不涉及设备工作原理前提下实现对多种故障类型的诊断,平均正确率在85%以上,可作为常规诊断方法的有益补充。A real-time fault diagnosis strategy based on data-driven method is introduced for electronic system of launch vehicles. The part of on-line faults of slow-varying signals can be recognized by the technique. Taking strapdown inertial measurement unit as an instance, at first, the output signals in different channels of the device are decomposed as several modal components by modal recognition approach;subsequently adequate components are selected adaptively according to their frequency parameters. Then the fault eigenvector is obtained by components mapping. At last the classifier is built with probabilistic neural network arithmetic in order to detect and diagnose the potential faults of the device. The validity of the method is showed by fault diagnosis simulation process, and the average fault diagnosis ratio can reach to 85%, which means the fault diagnosis method can be extended to different types of launch vehicles with general fault diagnosis approach together.

关 键 词：故障诊断 数据驱动 捷联惯组 自适应 运载火箭 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]