基于LSTM/NN的道岔故障特征提取与识别研究
|
摘要
道岔转辙机是保证列车安全运营的重要室外设备之一。分析道岔动作电流曲线,可以及时判断道岔的实时工作状态。采用基于长短期记忆(LSTM)模型结合传统神经网络的识别算法,从原始电流序列中自动提取特征,再根据特征利用神经网络分类器(NNC)来对道岔动作电流曲线进行智能故障识别。实验结果表明,所提算法不会丢失电流曲线的有效信息,并且提高了准确率,训练集上的准确率为100%,在测试集上准确率达到了99. 7%。算法能够满足铁路现场实际应用需要,对保障道岔的正常运行具有十分重要的现实意义。
The switch machine is one of the most important outdoor devices to ensure the safe operation of the train.The real-time working state of the switch machine can be judged in time by analyzing the action current curve. In this paper,based on LSTM model,we adopted the recognition algorithm which combined traditional neural network. Features were extracted automatically from the original current sequence. According to the characteristics,the neural network classifier( NNC) was used to intelligently recognize the fault of action current curve. The experimental results show that the algorithm does not lose the effective information of the action current curve. It improves the accuracy. The accuracy rate on the training set is 100%. The accuracy rate reaches 99. 7% on the test set. It can meet the practical needs of the railway field. It is of great practical significance for ensuring the normal operation of the switch.
The switch machine is one of the most important outdoor devices to ensure the safe operation of the train.The real-time working state of the switch machine can be judged in time by analyzing the action current curve. In this paper,based on LSTM model,we adopted the recognition algorithm which combined traditional neural network. Features were extracted automatically from the original current sequence. According to the characteristics,the neural network classifier( NNC) was used to intelligently recognize the fault of action current curve. The experimental results show that the algorithm does not lose the effective information of the action current curve. It improves the accuracy. The accuracy rate on the training set is 100%. The accuracy rate reaches 99. 7% on the test set. It can meet the practical needs of the railway field. It is of great practical significance for ensuring the normal operation of the switch.
引文
[1]韩煜霖,杨静,邢宗义.基于定性趋势分析的道岔故障诊断方法研究[J].铁道标准设计,2017,61(12):120-125,130.
[2]梅欢,马艳东,单九思,等.基于样条插值与RBF网络的道岔故障诊断系统[J].计算机技术与发展,2017,27(5):160-163,169.
[3]余飞.信号专家诊断系统及其在设备故障预防中的应用[J].神华科技,2016,14(1):87-90.
[4]张凯.基于神经网络的铁路道岔故障智能诊断研究[D].西安:长安大学,2016.
[5]张星,魏文军.基于支持向量机的道岔故障动作电流识别[J].电脑知识与技术,2016,12(7):264-265.
[6]程宇佳.基于核方法的高速铁路道岔故障诊断[D].北京:北京交通大学,2016.
[7]唐德权,史伟奇,张波云.基于多模态信息特征融合的犯罪预测算法研究[J].计算机应用与软件,2018,35(7):221-225,262.
[8]冷建伟,李鹏.基于最优特征更新分类器的压缩跟踪算法[J].计算机应用与软件,2018,35(2):206-211.
[9] Hochreiter S,Sehmidhuber J. Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.
[2]梅欢,马艳东,单九思,等.基于样条插值与RBF网络的道岔故障诊断系统[J].计算机技术与发展,2017,27(5):160-163,169.
[3]余飞.信号专家诊断系统及其在设备故障预防中的应用[J].神华科技,2016,14(1):87-90.
[4]张凯.基于神经网络的铁路道岔故障智能诊断研究[D].西安:长安大学,2016.
[5]张星,魏文军.基于支持向量机的道岔故障动作电流识别[J].电脑知识与技术,2016,12(7):264-265.
[6]程宇佳.基于核方法的高速铁路道岔故障诊断[D].北京:北京交通大学,2016.
[7]唐德权,史伟奇,张波云.基于多模态信息特征融合的犯罪预测算法研究[J].计算机应用与软件,2018,35(7):221-225,262.
[8]冷建伟,李鹏.基于最优特征更新分类器的压缩跟踪算法[J].计算机应用与软件,2018,35(2):206-211.
[9] Hochreiter S,Sehmidhuber J. Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.