特征信息加权融合的进气道斜板故障诊断  

Fault diagnosis of inlet slant plate based on weighted fusion of feature information

作　　者：王宽[1] 李盘文[1] 祁晓鹏 WANG Kuan;LI Panwen;QI Xiaopeng(Chinese Flight Test Establishment,Xi’an 710089,China)

机构地区：[1]中国飞行试验研究院,陕西西安710089

出　　处：《中国测试》2020年第11期43-47,58,共6页China Measurement & Test

基　　金：军工科研院所稳定支持项目(WD-2019-2-2)。

摘　　要：针对飞机进气道斜板故障定位难度大、故障诊断准确率低的问题,提出一种特征信息加权融合的进气道斜板故障诊断算法。该方法首先对采集信号进行预处理,计算出能够反映故障信息的多个特征指标;然后分别采用BP神经网络、径向基RBF神经网络和SVM向量机3种不同类型的分类模型对特征指标进行分类,实现故障的预诊断;之后采用加权D-S证据理论对预诊断结果进行融合决策,得到诊断结果,提高故障诊断的准确性。最后通过对比试验,表明该方法的故障诊断准确率高达97%以上,同时通过实际试飞数据进一步证明该算法能够对飞机进气道斜板故障进行诊断,具有较高的工程应用价值。Aiming at the problems of aircraft slant plate fault location difficulty and low fault diagnosis accuracy,a weighted fusion of feature information for inlet slant plate fault diagnosis algorithm is proposed.This method first preprocesses the collected signals and calculates multiple characteristic indicators that reflect the fault information.Second use three different types of classification models of BP neural network,radial basis RBF neural network and SVM vector machine to classify the feature indicators to realize the prediagnosis of faults.Then,the weighted D-S evidence theory is used to make fusion decision on the prediagnosis results,and the diagnosis results are obtained,which improves the accuracy of fault diagnosis.Finally,the comparison test shows that the method has a fault diagnosis accuracy rate of more than 97%.At the same time,the actual flight test data further shows that the algorithm can diagnose the aircraft inlet slant plate failure and has high engineering application value.

关 键 词：特征信息 故障诊断 加权D-S 数据融合 神经网络 支持向量机 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]