摘要
针对铁路信号联锁故障诊断准确率、查准率、查全率低的问题,结合铁路信号设备故障特征属性及铁路站场树形拓扑结构,提出一种基于树形卷积神经网络(tree-based convolutional neural network,TBCNN)的铁路信号联锁故障诊断模型,并在实验室平台上进行仿真。结果表明,所提的铁路信号联锁故障诊断模型可有效诊断单一铁路信号联锁故障和复合铁路信号联锁故障,相较于BP神经网络与RvNN网络等传统故障诊断模型,在准确率、查准率、查全率和F1值各项评价指标上具有更优异的表现,故障诊断准确率分别可达到86.95%和76.25%,查准率分别可达到85.40%和68.94%,查全率分别可达到94.94%和71.19%,F1值分别为89.83%和70.05%,具有一定的有效性和优越性。
Aiming at the low accuracy,precision and recall of railway signal interlocking fault diagnosis,combined with the fault characteristic attributes of railway signal equipment and the tree topology of railway station,a railway signal interlocking fault diagnosis model based on tree based convolutional neural network(tbcnn) is proposed and simulated on the laboratory platform.The results show that the proposed railway signal interlocking fault diagnosis model can effectively diagnose single railway signal interlocking fault and composite railway signal interlocking fault.Compared with traditional fault diagnosis models such as BP neural network and rvnn network,it has better performance in accuracy,precision,recall and F1 value.The fault diagnosis accuracy can reach 86.95% and 76.25% respectively,and the precision can reach 85.40% and 68.94% respectively,The recall rate can reach 94.94% and 71.19% respectively,and the F1 value is 89.83% and 70.05% respectively,which has certain effectiveness and superiority.
作者
杜巧玲
罗永
DU Qiaoling;LUO Yong(Xi'an Traffic Engineering Institute,Xi'an,Shaanxi 710300,China;China Railway First Group Electric Engineering Co.,Ltd.,Xi'an,Shaanxi 710300,China)
出处
《自动化与仪器仪表》
2022年第4期38-43,共6页
Automation & Instrumentation
基金
陕西省教育厅科研计划项目资助(21JK0745)。