摘要
针对SOM网络(自组织特征映射神经网络)可视化方法简单、直观的特点,文中将其应用到液压系统的故障分类中。以电流信号的频域能量作为特征参数,用db2共轭正交滤波器组对所获数据进行小波包分解,提取系统在正常及故障运行状态下的特征向量,作为训练样本,然后利用U矩阵图和D矩阵图等可视化工具对分类结果进行仿真与分析,并与一般结果进行比较。结论表明,该方法可行且对故障的判别率高。
SOM network （self-organizing feature map neural network） visual method which is simple, intuitionistic used for hydraulic power system fault classification in this paper. The parameters gained with frequency domain energy of signals , the data obtained is decomposed used wavelet packet （Db2 conjugate orthogonal filter bank） and the feature vector is extracted in normal and fault operating state of system.Then use the U matrix map and D matrix map as visual tools to simulate and analyse the classification results. Finally,it is compared with the general results.The Conclusion is that SOM network visual method is feasible and has high rate to fault discrimination.
出处
《流体传动与控制》
2009年第3期5-9,共5页
Fluid Power Transmission & Control
基金
国家自然科学基金项目(50575168)
陕西省工业攻关项目(2008K05-04)
关键词
液压动力系统
故障分类
SOM神经网络
可视化
self-organizing feature map neural network visualization hydraulic power system fault classification
