基于神经网络的关联规则在故障诊断中的应用  

Neural Network-Based Association Rules for Fault Diagnosis

作　　者：刘晶[1,2] 季海鹏[3] 

机构地区：[1]中国传媒大学信息工学院,北京100024 [2]中国传媒大学高性能计算中心,北京100024 [3]燕山大学机械工程学院,河北秦皇岛066004

出　　处：《工业工程》2011年第2期118-121,共4页Industrial Engineering Journal

基　　金：河北省自然科学基金资助项目(G2010001331)

摘　　要：采用加权关联规则算法对设备历史数据库进行挖掘,建立加权关联规则模式库。设备监控数据通过与模式库匹配,实现设备故障诊断。同时,针对钢铁企业中液压设备的特殊性,提出利用自组织竞争神经网络模型确定权值,即将设备故障信息的3个主要属性:重要程度、易损程度、故障等级作为模型的输入,通过训练样本确定设备故障的加权关联规则的权值。实例证明了该方法的有效性。With historical equipment database,weighted association rule algorithm is adopted to conduct data mining.By using weighted association rules,a model base is established.Based on the model base,fault diagnosis can be performed by using equipment monitor data.In the meanwhile,self-organizing competitive neural network model is used to determine the weight of hydraulic equipment in a steel enterprise.Three properties are considered.They are degree of importance,degree of vulnerability,and level of fault.The model takes these three properties of the fault information as inputs and determines the connection weights of equipment fault through sample training.Experiments show that this algorithm can improve the accuracy of fault diagnosis of hydraulic equipment.

关 键 词：数据挖掘 加权关联规则 神经网络 故障诊断 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]