基于SK和LSSVM的飞机空气制冷机寿命趋势分析  

Life Trend Analysis of Aircraft Air Refrigerator Based on SK and LSSVM

作　　者：崔建国[1] 滑娇娇 董世良[2] 崔霄 齐义文[1] 蒋丽英[1] 

机构地区：[1]沈阳航空航天大学自动化学院,沈阳110136 [2]沈阳飞机设计研究所,沈阳110035 [3]沈阳空气动力研究院,沈阳110134

出　　处：《振动．测试与诊断》2016年第3期471-477,602,共7页Journal of Vibration,Measurement & Diagnosis

基　　金：航空科学基金资助项目(20153354005);国防基础科研计划资助项目(A0520110023);国防预研资助项目(Z052012B002);辽宁省自然科学基金资助项目(2014024003)

摘　　要：针对工程上所获取的飞机空气制冷机信息本身具有的不确定性等特点,提出了谱峭度(spectral kurtosis,简称SK)和最小二乘支持向量机(least squares support vector machine,简称LSSVM)相结合的寿命趋势分析方法。首先,将寿命表征参量通过经验模态分解(empirical mode decomposition,简称EMD)分解为多个固有模态分量(intrinsic mode function,简称IMF),分别对多个IMF分量进行傅里叶变换,得到多个频谱,计算每根谱线的谱峭度;其次,通过谱峭度寻找到影响飞机空气制冷机失效的主要IMF分量,对其进行信号重构;最后,提取重构后信号的能量特征,创建最小二乘支持向量机寿命趋势模型,对飞机空气制冷机寿命趋势进行预测。通过对实际监测数据的分析和验证结果表明,该方法可以实现飞机空气制冷机寿命趋势分析效能,具有很好的工程应用价值。Considering the uncertainty characteristics of the aircraft air refrigerator′s information acquired from project,a lifetime trend analysis method is proposed based on spectrum kurtosis(SK)and least squares support vector machine(LSSVM).First,the life parameter is decomposed into several intrinsic mode functions(IMFs)by empirical mode decomposition(EMD),taking the IMFs for Fourier transform respectively to get the spectrums and calculate the SK of each spectral line.Then through the SK,the important IMFs can be found out those effect on the air refrigerator failure,to signal synthesis.Finally,the energy features of the synthetic signal are extracted to establish the LSSVM trend analysis model to predict the air refrigerator′s life trend.Experimental results show that the proposed method of aircraft air refrigerator life trend analysis based on SK and LSSVM can effectively realize the air refrigerator life trend analysis,with strong potential for engineering applications.

关 键 词：飞机空气制冷机 谱峭度 最小二乘支持向量机 寿命 趋势分析 

分 类 号：TH133.3[机械工程—机械制造及自动化] TP206.3[自动化与计算机技术—检测技术与自动化装置]