FTA与BP神经网络结合的地平仪故障诊断方法研究  

Research on Fault Diagnosis Method Combining FTA and BP Neural Network of Horizon

作　　者：李锋 陈振 王腾飞 李晨旭 刘麦良 LI Feng;CHEN Zhen;WANG Tengfei;LI Chenxu;LIU Mailiang(College of Aircraft,Xi’an Aviation University,Xi’an 710089,China;Chinese Flight Test Establishment,Xi’an 710021,China)

机构地区：[1]西安航空学院飞行器学院,陕西西安710089 [2]中国飞行试验研究院,陕西西安710021

出　　处：《自动化仪表》2023年第4期39-42,共4页Process Automation Instrumentation

摘　　要：为了提高某型地平仪常见故障的诊断效率与准确率,研究了故障树分析(FTA)与反向传播(BP)神经网络相结合的地平仪故障诊断方法。根据地平仪的结构原理,首先采用FTA法得到了该型地平仪的所有故障模式及最小割集,建立了故障树的结构函数。然后按照最小割集重要度,筛选出BP神经网络训练样本的主要故障模式。最后以某单位该型地平仪的故障统计数据为基础,运用BP神经网络的方法建立了地平仪的故障诊断模型,并对模型进行了验证。验证结果表明,采用FTA法与BP神经网络相结合的故障诊断方法,弥补了2种方法单独诊断时的固有缺陷,提高了故障诊断的准确性和效率。To improve the diagnosis efficiency and accuracy of common faults of a type of horizon,the fault tree analysis(FTA)combined with back propagation(BP)neural network is studied for the fault diagnosis method of the horizon.According to the structure principle of the horizon,firstly,all the fault modes and the minimum cut set of this type of horizon are obtained by FTA method,and the structure function of the fault tree is established.Then,the main fault modes of the BP neural network training samples are filtered according to the importance of the minimum cut set.Finally,based on the fault statistics of this type of horizon in a unit,the fault diagnosis model of the horizon is established by using the BP neural network method,and the model is verified.The validation results show that the fault diagnosis method combining FTA method and BP neural network make up for the inherent defects of the two methods when they are diagnosed separately,and improves the accuracy and efficiency of fault diagnosis.

关 键 词：故障树分析法 反向传播神经网络 地平仪 故障诊断 故障模式 

分 类 号：TH707[机械工程—仪器科学与技术]