基于EMD和SVM的传感器故障诊断方法  

Sensor fault diagnosis based on empirical mode decomposition and support vector machines

作　　者：冯志刚[1,2] 王祁[1] 信太克规[1] 

机构地区：[1]哈尔滨工业大学自动化测试与控制系,哈尔滨150001 [2]沈阳航空工业学院自动化学院,沈阳110136

出　　处：《哈尔滨工业大学学报》2009年第5期59-63,共5页Journal of Harbin Institute of Technology

基　　金：国家自然科学基金资助项目(60572010)

摘　　要：为了解决自确认压力传感器的故障诊断问题,提出了一种基于经验模式分解(EMD)和支持向量机(SVM)的传感器故障诊断方法,该方法对传感器输出信号进行经验模态分解,将其分解为若干个固有模态函数(IMF),对每个IMF通过一定的削减算法增强故障特征,然后计算每个IMF和残余项的能量以及整个信号的削减比作为特征向量,以此作为输入来建立支持向量多分类机,判断传感器的故障类型.通过压力传感器的故障诊断结果表明,该方法能有效的应用于传感器的故障诊断中.To solve the fault diagnosis problem of self-validating pressure sensor, a sensor fault diagnosis approach based on empirical mode decomposition （EMD） method and support vector machines （SVM） is proposed. The EMD method was used to decompose the sensor output signal into a number of intrinsic mode function （IMF） components and a residue component. With a certain cutting algorithm, the IMFs with fault character were strengthened. After that, the energy of each IMF and residue component as well as the average cutting ratio of all the IMFs and residue component was calculated, which is regarded as the feature vector. Then, the support vector machines for multi-classification used as fault classifiers were established to identify the condition and fault pattern of the sensor. Practical example of pressure sensor shows that the proposed approach can be applied to the sensor fault diagnosis effectively.

关 键 词：经验模态分解 支持向量机 特征提取 传感器故障诊断 

分 类 号：TH133[机械工程—机械制造及自动化] TP183[自动化与计算机技术—控制理论与控制工程]