The existing particle filter fault prediction methods give the predictive value of the corresponding time by the particle filter algorithm, and then compare the distance between forecasting sequence and observation sequence to predict the fault. However, this fault prediction method can not handle the condition that the length of forecasting sequence is different from that of observation sequence. Dynamic Time Warping is a pattern matching algorithm based on nonlinear, which is suitable for the time sequence of different lengths. This paper is from the new perspective of using the Dynamic Time Warping algorithm to measure the similarity between normal working equipment's time sequence and abnormal sequence caused by potential faults, and design the system normal degree and abnormal degree to distinguish whether the device is operating properly or not, thus predict potential faults. Experimental results demonstrate the feasibility of this method, which can predict the system faults timely and accurately.%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,然后比较预测序列与观测序列的距离来对故障进行预报,但这种基于相同长度时间序列的故障预报方法不能处理预测序列与观测序列长度不同的情况.本文借助适用于不同长度时间序列的动态时间弯曲技术,对故障相关的时间序列进行分析,从动态时间弯曲算法度量设备正常工作的时间序列与潜在故障引起的异常序列之间相似度的角度,设计了系统正常度及反常度来判别设备是否正常运行,进而对潜在故障进行预报.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.
展开▼
