摘要
针对传统重载铁路电务设备运维能力低,导致运维作业精准度不高的问题,提出基于深度学习的重载铁路电务设备智能运维系统。首先采用监测运维一体化采集电务设备运维数据;然后基于大数据技术,将递归神经网络RvNN、树形长短期记忆网络Tree-LSTM和树形卷积神经网络TBCNN三个树形神经网络进行融合,并与循环神经网络及衔生算法相结合,构建一个基于联锁逻辑时序与深度学习相结合的设备运维模型;最后通过构建模型进行联锁故障诊断和设备状态评估。实验结果表明,在二分类任务中,本模型的故障诊断准确率高达98.54%;在多分类任务中的诊断准确率为89.13%,对比于单一的树形结构模型和BP神经网络模型,多分类任务中的诊断率分别高出了15%和20%。系统应用发现,该系统能够进行重载铁路电务设备运维状态预测和准确评估,实现了电务设备全生命周期管理和自动化运维。
In view of the problem of low operation and maintenance capacity of traditional heavy load railway electrical equipment and low accuracy of operation and maintenance operations, it is proposed to build an intelligent operation and maintenance system of heavy load railway electrical equipment based on deep learning. Then, three tree neural networks, RvNN, tree-long short-term memory network Tree-LSTM and TBCNN, combine with recurrent neural network and generation algorithm to construct a model based on interlocking logic timing and deep learning, and finally. The experimental results show that the diagnostic accuracy is 98.54% and 89.13%. Compared with the single tree classification model and the diagnosis rate is 15% and 20% higher respectively. The system application found that the system can predict and accurately evaluate the operation and maintenance status of heavy-load railway electrical equipment, and realize the full life cycle management and automatic operation and maintenance of electrical equipment.
作者
张斌
谢智多
胡启正
常浩
ZHANG Bin;Xie Zhiduo;HU Qizheng;CHANG Hao(Chn Energy Shuohuang Railway Development Company Ltd,Cangzhou Hebei 062350,China;China Academy of Railway Sciences Corporation Limited Signal and Communication Research Institute,Beijing 100081,China)
出处
《自动化与仪器仪表》
2023年第3期184-189,共6页
Automation & Instrumentation
基金
国能集团科技创新项目《重载铁路基础设施智能运维技术研究与应用》(GJNY-20-231)。
关键词
重载铁路
电务设备运维
树形神经网络
联锁时序逻辑
故障诊断
heavy load railway
electrical equipment operation and maintenance
tree neural network
interlocking timing logic
fault diagnosis
