Study on failure time series forecasting of metro automatic gate machine’s flap mechanism
-
摘要: 基于CNN+ LSTM混合神经网络构建故障时间序列预测模型，利用某型号地铁闸机扇门机构的故障数据进行实例分析，并与ARIMA、CNN和LSTM 3种单一预测模型对比。结果表明：CNN+LSTM混合神经网络模型的预测准确性较高，具有良好应用前景，研究成果可用于支持地铁闸机维修计划的制定和优化。Abstract: Based on CNN+LSTM hybrid neural network, a fault time series forecasting model was established for the flap mechanism of automatic gate machine for metro. And then a case analysis was presented based on this model by using the fault data of the flap mechanism of a certain type of automatic gate machine. By comparing this hybrid neural network model with other three single forecasting models that are ARIMA, CNN and LSTM, the results show that the accuracy of forecasting by this CNN+LSTM hybrid neural network model is higher and it has a good application prospect. The research results can be used to support the formulation and optimization of the maintenance plan for the flap mechanism of automatic gate machine.
-
Keywords:
- metro automatic gate machine /
- failure /
- neural networks /
- time series forecasting
-
-
表 1 故障率数据分组
组号 输入样本 期望映射 X1 0.004 311,…,0.034 483 0.004 213 X2 0.004 213,…,0.025 862 0.017 241 $ \vdots $ $ \vdots $ $ \vdots $ X61 0.107 758,…,0.129 311 0.116 379 
下载: 导出CSV
表 2 实验环境配置
实验环境 配置 操作系统 Windows 10 CPU I7-9750H GPU RTX 2060 运行内存 16 GB 程序语言 Python 3.7.4 程序框架 Tensorflow 2.1.0
下载: 导出CSV
表 3 CNN+LSTM故障时间序列预测模型的参数设置
模型层 参数 Conv1D （32，5，1） Dropout 0.01 Batch Normalization - Activation relu Conv1D （32，5，1） Activation relu Maxpooling1D 1 LSTM 64 LSTM 64 Activation sigmod Flatten - Dense 32 Dense 1
下载: 导出CSV
表 4 4种模型预测结果准确性对比
模型名称 RMSE MAE R2 CNN+LSTM 0.024 971 0.141 886 0.73 ARIMA 0.051 942 0.164 640 0.37 CNN 0.026 694 0.145 570 0.67 LSTM 0.026 880 0.144 942 0.63
下载: 导出CSV
-
[1] 李立辉,田 翔,杨海东,等. 基于SVR的金融时间序列预测 [J]. 计算机工程与应用,2005(30):221-224. DOI: 10.3321/j.issn:1002-8331.2005.30.070 [2] 王 鑫,吴 际,刘 超,等. 基于LSTM循环神经网络的故障时间序列预测 [J]. 北京航空航天大学学报,2018,44(4):772-784. [3] 李勃旭. 基于ARIMA模型的地铁车门传动系统故障预测[D]. 兰州: 兰州理工大学, 2019. [4] Krishnannair S, Aldrich C. Process Monitoring and Fault Detection using Empirical Mode Decomposition and Singular Spectrum Analysis [J]. IFAC PapersOnLine, 2019, 52(14): 219-224. DOI: 10.1016/j.ifacol.2019.09.190
[5] 连光耀,吕晓明,黄考利,等. 基于最小二乘支持向量机的复杂装备故障预测模型研究 [J]. 计算机测量与控制,2011,19(5):1030-1032. [6] Chulkov A O, Nesteruk D A, Vavilov V P, et al. Optimizing input data for training an artificial neural network used for evaluating defect depth in infrared thermographic nondestructive testing [J]. Infrared Physics and Technology, 2019, 102. DOI: 10.1016/j.infrared.2019.103047
[7] 茅 飞,孔慧慧,李宏胜,等. 轨道交通车站闸机智能识别研究 [J]. 城市轨道交通研究,2017,20(12):123-126. [8] 徐文文,彭建平,邱春蓉. 基于支持向量回归的地铁受电弓滑板磨耗趋势预测模型研究 [J]. 铁路计算机应用,2020,29(1):77-81. DOI: 10.3969/j.issn.1005-8451.2020.01.015 [9] 李建伟,程晓卿,秦 勇,等. 基于BP神经网络的城市轨道交通车辆可靠性预测 [J]. 中南大学学报(自然科学版),2013,44(S1):42-46. [10] Hochreiter S, Schmidhuber J. Long short-term memory [J]. Neural computation, 1997, 9(8): 1735-1780. DOI: 10.1162/neco.1997.9.8.1735
[11] Zeiler M D, Fergus R. Visualizing and understanding convolutional networks[C]//European Conference on Computer Vision(ECCV). Cham, Springer, 2014: 818-833.
[12] Hooshmand A, Sharma R. Energy predictive models with limited data using transfer learning[C]//Proceedings of the Tenth ACM International Conference on Future Energy Systems(ICFES). NewYork, ACM, 2019: 12-16.
[13] Wang Q, Zheng S, Farahat A, et al. Remaining useful life estimation using functional data analysis[C]//2019 IEEE International Conference on Prognostics and Health Management (ICPHM). California, IEEE, 2019: 1-8.
-
期刊类型引用(4)
1. 陈耀标，王晓敏，王亚祝. 厢式车侧帘式货物运输防护装置研制及应用. 电工技术. 2022(08): 108-109+113 . 
百度学术
2. 丁旸. 强化货物装载加固确保铁路运输安全的探讨. 新型工业化. 2022(10): 192-195 . 百度学术
3. 成海涛. 船上重件货物固定过程效率控制方法优化. 舰船科学技术. 2021(20): 19-21 . 百度学术
4. 杜庆，刘英华，黄小维. 加强兴隆场站铁路预应力梁装车安全防控对策探讨. 铁道货运. 2020(08): 50-55 . 百度学术
其他类型引用(0)
计量
- 文章访问数: 127
- HTML全文浏览量: 141
- PDF下载量: 17
- 被引次数: 4
