无线传感网络多分类支持向量机设备状态识别

Equipment condition recognition with multi-category support vector machine in wireless sensor networks

无线传感网络由大量传感节点组成,以自组织方式构成网络,通过协作测量获取监测对象的详细信息。基于无线传感网络的生产设备状态识别方法具有可靠性高和柔性好的特点。针对工业生产特点,建立分簇结构的能量异质性无线传感网络模型,并应用于旋转机械设备故障状态的监测识别。采用小波包分解提取故障特征,通过多分类支持向量机对设备状态进行识别。实验表明上述方法能有效地对旋转机械设备故障状态进行在线监测识别,并能适应监测环境的动态变化,是一种可靠性高、柔性好的监测识别方法。

Wireless sensor networks consist of numerous sensor nodes that autonomously form networks and collaboratively perform measurement. Equipment monitoring with wireless sensor networks is highly reliable and flexible. An energy heterogeneous and clustered network model is proposed. It is used for rotary mechanical equipment condition recognition. Wavelet packet is employed to extract features. Multi-category support vector machine is used for condition recognition. Experiments show the method is reliable and flexible. It efficiently performs online recognition and dynamically adapts to environmental changes.