自组织特征映射在柴油机振动信号模式分类中的应用  

Application Research of SOFM in Vibration Signal-based Pattern Recognition for Military Diesel Engine

作　　者：赵寅卯[1] 冯辅周[2] 甘露[2] 杨磊[3] 

机构地区：[1]装甲兵工程学院训练部 [2]装甲兵工程学院机械工程系,北京100072 [3]装甲兵工程学院装备指挥与管理系,北京100072

出　　处：《装甲兵工程学院学报》2007年第2期33-36,共4页Journal of Academy of Armored Force Engineering

基　　金：国家自然科学基金(50405022);中韩国际合作课题

摘　　要：通过分析柴油机在磨合期、不同摩托小时和拉缸等典型状态下的振动信号样本，计算出各类样本在幅域、时域和频域的特征参量，按照类别可分离性判据进行特征选择，寻找出能够代表发动机不同状态的有效特征参数，同时降低特征向量空间的雏数，最后利用自组织特征映射神经网络（Self-OrganizingFeatureMap）进行发动机不同状态的分类。分析结果表明，SOFM能够对各类模式进行有效的分类，准确率达到92％以上。In this paper, vibration signals from diesel engine bench test are captured so as to get various sample data under conditions of running - in period, different motor hour and scuffing of cylinder bore. Then characteristic parameters are extracted in amplitude domain, time domain and frequency domain, and then class separability criterion is employed to find effective characteristic parameters which can represent engine condition, at the same time, to reduce spatial dimension. Finally, SOFM is used to complete classification of different engine pattern. The research results show that SOFM is an effective pattern recognition method for the diesel engine considered in this paper, the classification accuracy is over 92%.

关 键 词：SOFM 特征选择 故障诊断 

分 类 号：TP206[自动化与计算机技术—检测技术与自动化装置]