基于聚类和神经网络相结合的设备状态诊断分析

Equipment Status Diagnosis and Analysis Based on Combination of Clustering and Neural Network

根据某电厂一次主轴密封漏水的事件,在没有先验特征的情况下,通过对设备状态数据进行分析,针对设备不是完全健康的情况,提出用聚类分析的方法先对数据进行处理,得到较接近设备健康状态和异常状态的特征数据。在此基础上采用BP神经网络的方法对监控数据进行分析并建立诊断模型,使用得到的诊断模型对每个月的监控数据进行分析,统计异常数据出现率并加以比较,实验结果能较好地反映出设备状态的劣化,为设备状态检修提供了参考依据。

According to a leakage event of a main shaft sealing in a power plant,in the absence of a priori feature,through the analysis of equipment status data,aiming at the situation that the equipment is not completely healthy,the clustering analysis method is proposed to process the data first,and get the characteristic data which are close to the equipment health status and abnormal status.On this basis,the BP neural network method is used to analyze the monitoring data and establish a diagnostic model.The diagnostic model is used to analyze the monitoring data every month.The occurrence rate of abnormal data is counted and compared.The experimental results can better reflect the deterioration of equipment status and provide a reference for condition based maintenance of equipment.