大数据环境下离散制造车间异常事件发现方法  

ABNORMAL EVENT DISCOVERY METHOD OF DISCRETE MANUFACTURE WORKSHOP IN BIG DATA

作　　者：马超[1] 徐迭石 张淑丽[1] 刘胜辉[1] 

机构地区：[1]哈尔滨理工大学软件学院,黑龙江哈尔滨150080

出　　处：《计算机应用与软件》2017年第9期288-293,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(51375128);黑龙江省教育厅科学技术研究项目(12541159)

摘　　要：针对大数据环境下离散制造企业车间生产过程中生产异常难以有效管控的问题,先从理论上研究建立车间异常事件预警模型的合理性和实用性。然后从技术实现角度给出异常触发事件的数据来源及其计算方法。接着综合时间序列和因果关系两个维度,建立基于时序序列上多决策树的车间异常事件预警模型,保证了预测结果的准确性和可靠性。最后采用某型号燃气轮机转子的生产过程数据验证模型的有效性。To deal with the problems of effectively controlling abnormal events happened during the production process of the discrete manufacture enterprise in big data,this paper firstly studied the rationality and utility of building the early warning model of abnormal events in workshop in theory. Then the paper gave the data source and its calculation method of the abnormal triggering event from the technical realization aspect, combined the time series and the causal relationship,and established the early warning model of the workshop anomaly based on the multi-decision tree on the time series,which ensures the accuracy and reliability of the forecast. Finally,the validity of the model was verified by the production process data of a gas turbine rotor.

关 键 词：异常事件发现 时间序列 决策树 预警 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]