时间序列在故障诊断中的应用  

Fault diagnosis based on time series

作　　者：方泽南[1,2,3] 傅尚新 张勇[1,2,3] 

机构地区：[1]清华大学精密仪器与机械学系 [2]清华大学热能工程系 [3]深圳桑达电信技术有限公司

出　　处：《清华大学学报（自然科学版）》1998年第9期123-126,共4页Journal of Tsinghua University(Science and Technology)

摘　　要：提出了一种用时间序列理论进行故障诊断的方法，可以在缺乏对实际故障机理了解的情况下从机组自身的运行过程中动态获取故障的统计特征信息，识别机组运行的状态。这种方法先对实际机组的振动信号建立时间序列模型，然后用模型参数来训练一个模式识别神经网络作为故障诊断的工具。实验结果表明，这种方法有很好的实用性。A fault diagnostics method based on the time series is proposed. Even though there was little knowledge about the characteristics and features of faults, the fault pattern can still be recognized in a statistic way. After modeling the actual machine running process of vibration signal by means of time series theory, the set of parameters of the corresponding model is taken as the condition characteristics according to which the fault pattern is identified. This method was proved to be practical and efficient by a series of experiments.

关 键 词：时间序列 故障诊断 神经网络 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置]