摘要
针对目前由人工分析监测数据进行ZPW-2000轨道电路故障判别存在判别效率低、判别周期较长、对数据分析所依赖的人员经验程度高的缺陷,将粗糙集理论和模糊认知图的概念引入ZPW-2000轨道电路的故障判别中,通过主分量启发式算法对原始故障数据进行属性约简,得到故障特征参数,利用模糊认知图构建出基于属性约简和模糊认知图的分类器,并在此过程中利用自适应遗传算法完成FCM权重的学习。通过仿真实验验证,在模糊认知图的基础上利用粗糙集进行特征的提取,再对ZPW-2000轨道电路故障进行诊断,这种方法是有效可行的,并与人工分析故障数据进行诊断的方法进行比较,发现基于属性约减和模糊认知图的分类器有较高的故障识别率以及较短的诊断时间。
In view of the disadvantages of low diagnostic efficiency,long diagnosis cycle and high depen-dence on the experiences of data analyzers in the fault diagnosis of ZPW-2000 track circuit fault by manual analysis of monitoring data,the concepts of fuzzy cognitive map and rough set theory were introduced into fault diagnosis of ZPW-2000 track circuit.Firstly,the attribute reduction of the original fault data was performed by the main component heuristic algorithm,to obtain the fault characteristic parameters.Then the fuzzy cognition map was used to construct the classifier based on reduction attribute and fuzzy cognitive map,to complete FCM weights learning in term of the adaptive genetic algorithm during the process.Computer simulations show that the method of using the rough set to extract the features on the basis of fuzzy cognitive map,and then to diagnose the fault of the ZPW-2000 track circuit is effective and feasible.Compared with the diagnosis method of using manual analysis of monitoring data,the classifier based on attribute reduction and fuzzy cognitive map has high failure recognition rate and short diagnosis time.
作者
董昱
陈星
DONG Yu;CHEN Xing(School of Automation ＆ Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)
出处
《铁道学报》
EI
CAS
CSCD
北大核心
2018年第6期83-89,共7页
Journal of the China Railway Society
基金
国家自然科学基金(61763023)
