基于FSVM改良隶属度的发动机振动故障识别  

Improved FSVM and multi-class fuzzy membership method for aeroengine vibration fault identification

作　　者：白斌[1] 白广忱[1] 林学柱[2] 

机构地区：[1]北京航空航天大学能源与动力工程学院,北京100191 [2]秦皇岛玻璃工业研究设计院,秦皇岛066001

出　　处：《振动与冲击》2013年第20期23-28,共6页Journal of Vibration and Shock

基　　金：国家自然科学基金(51175017;51275024);国家自然科学基金面上项目(51375032)

摘　　要：为了更好的对航空发动机整机振动进行故障诊断和识别,提出了改良的FSVM隶属度和多类隶属度与信息熵的融合定量分析方法,并且和传统的FSVM隶属度分析方法进行比较。对传统的FSVM的模糊隶属度函数改良后建立了多类模糊隶属度计算模型。通过实验实例验证了该多类模糊隶属度与信息熵相结合的技术对于航空发动机整机振动状态评估和故障诊断识别非常有效,计算出振动故障模式与故障原因之间的权值,建立了一个多参数的发动机振动状态分析模型;并对各类振动原因对发动机整体状态的影响进行定量分析,为发动机的振动抑制提供量化参考指标。In order to diagnose and identify effectively faults for an aeroengine＇s whole-body vibration, the improved fuzzy support vector machine （FSVM） and multi-class fuzzy membership method combined with the information entropy technique was proposed here. It was compared with the traditional FSVM membership analysis method. The calculation model of multi-class fuzzy membership was established based on improving the traditional FSVM fuzzy membership function. Tests and examples for aircraft engines＇ overall vibration performance, and fault diagnosis and identification verified that the technology of multi-class fuzzy membership combined with information entropy is very effective. The weighted values between fault modes and fault causes were determined and the multi-parameter vibration performance analysis model was developed, the effects of various vibration causes on the overall state of an aeroengine were analyzed quantitatively, and a quantitative reference index was provided for＇ aeroengine vibration suppressing.

关 键 词：模糊支持向量机 信息熵 多类模糊隶属度 模糊隶属度 故障诊断识别 

分 类 号：V231.92[航空宇航科学与技术—航空宇航推进理论与工程] TH165.3[机械工程—机械制造及自动化]