数据驱动的动车组滚动轴承故障预测

Antifriction Bearing Failure Prediction of EMU Based on Data Driven Approach

为了有效提高动车组滚动轴承故障的发现率,减少故障监控系统的误报现象,基于Apache Hadoop大数据平台对经典Apriori算法进行改进,并将其应用于动车组滚动轴承故障的预测研究工作中。首先,针对经典Apriori算法的不足,在MapReduce框架下提出以业务经验为约束的改进的Apriori算法。其次,基于文中提出的改进的Apriori算法对某铁路局的动车组状态、故障预警、维修历史等信息进行深度数据挖掘,并通过得出的关联规则进行动车组滚动轴承故障的预测。实验结果表明,文中提出的算法准确率达72%,减少了80%以上的误报报警信息,在实验环境中运算效率较传统的Apriori算法提高了50%。

In order to effectively increase the antifriction bearing failure discovery rate in EMU and reduce the failure misinformation, the classical Apriori algorithm was improved based on the Apache Hadoop big data platform, and the improved algorithm was applied into the research work of EMU antifriction bearing failure prediction. To overcome the limitations of classical Apriori algorithm, an improved Apriori algorithm under the constraints of professional experience was proposed in the MapReduce framework. Then the depth data such as the status, failure warning and maintenance history of a certain railway bureau＇s EMU were mined based on the improved Apriori algorithm and the prediction of EMU antifriction bearing failures was obtained based on some association rules. The results show that with an accuracy rate of 72 %, the failure misinformation of the proposed algorithm is decreased by 80%, and compared with the classical Apriori algorithm, the computing efficiency of the proposed algorithm is increased by 50%.