摘要
现有的粒子滤波故障预报方法主要是通过粒子滤波算法得到对应时刻的预测值,然后比较其与实际值的差值来对故障进行预报.从分析设备正常工作的时间序列数据与潜在故障引起的异常数据之间的相似性的新角度,设计了系统正常度和系统异常度来判别设备是否正常运行,进而对潜在的故障进行预测.实验结果验证了该方法的可行性,并能及时准确地预报出系统故障.
The existing particle filter fault prediction method is the corresponding time＇s predictive value obtained from particle filter algorithm. The particle filter algorithm predicts the system fault by comparison of the predict value and actual value. This paper designs a new method identifying the function of system equipment. By analysis of the normal working equipment＇s time-series data and abnormal data, it can predict the potential system fault. Experimental results demonstrate the feasibility of this method and the accuracy of predicting system fault.
出处
《计算机系统应用》
2015年第1期98-103,共6页
Computer Systems & Applications
基金
国家自然科学基金项目(61175123)
福建省自然科学基金(2013J01223)
福建省高校服务海西建设重点项目
关键词
相似性度量
正常度
异常度
粒子滤波
故障预报
similarity measure
normal degree
abnormal degree
particle filter
fault prediction
