基于LSTM+FP-Growth算法的印刷设备故障预警及诊断  

Fault warning and diagnosis of printing equipment based on LSTM+FP growth algorithm

作　　者：江朋 陆远[1] 胡莹[1] JIANG Peng;LU Yuan;HU Ying(School of Mechatronics Engineering,Nanchang University,Nanchang 330031,China)

机构地区：[1]南昌大学机电工程学院,江西南昌330031

出　　处：《南昌大学学报（工科版）》2021年第3期276-284,共9页Journal of Nanchang University(Engineering & Technology)

摘　　要：印刷凹印设备对连续运行要求较高,且设备传动系统检修困难,这些问题都将导致企业无法完成对设备的短周期维护及设备故障的实时预警。针对现状,提出一种基于长短期记忆(LSTM)网络和FP-Growth算法相结合的预警模型。采用FP-Growth关联规则挖掘算法实现对设备历史运行数据及故障记录的特征提取,构建频繁模式树,获取与故障相关的诊断规则,完成故障诊断专家知识库的建立。运用LSTM网络训练模型得到设备运行参数的时空特性,实现对下一时段参数特征变化的预测。分别使用现场数据与预测数据,借助专家知识库完成对设备的故障诊断及故障预测。通过测试分析,表明此模型的应用能有效提升印刷设备故障诊断的准确率,并能对设备起到提前预警的效果。The gravure printing equipment has high requirements for continuous operation,and the transmission system of the equipment is difficult to be repaired.These problems will lead to the enterprise unable to complete the short-term maintenance of equipment and real-time warning of equipment failure.In view of the current situation,an early warning model based on the combination of long short-term memory network and frequent pattern-growth algorithm was proposed.The FP-Growth association rule mining algorithm was used to extract the features of the equipment's historical operation data and fault records,construct frequent pattern-tree,obtain the fault related diagnostic rules,and complete the construction of fault diagnosis expert knowledge base.Then,the LSTM network training model was used to receive the spatiotemporal characteristics of the operating parameters,so as to realize the prediction of the parameter characteristics changes in the next period.With the help of expert knowledge base,the fault diagnosis and prediction of the equipment were completed by using the field data and prediction data,respectively.Through the test and analysis,it showed that the application of this model could effectively improve the accuracy of printing equipment fault diagnosis,and could play an early warning effect on the equipment.

关 键 词：长短期记忆网络 频繁模式树 专家知识库 故障诊断 故障预测 

分 类 号：TH17[机械工程—机械制造及自动化] TP277[自动化与计算机技术—检测技术与自动化装置]