Health State Evaluation and Prediction of Switch Equipment Using SOM and LSTM
-
摘要: 针对道岔转辙设备故障频发,且工作人员无法准确评估及预测其健康状态的问题,进行结合SOM-LSTM混合神经网络的道岔转辙设备健康状态评估及预测方法研究。首先,依据道岔动作功率曲线特点分三段提取其时域特征参数,利用自组织映射神经网络(Self organizing map,SOM)中最小量化误差求解道岔转辙设备健康因子(Health index,HI);其次,运用长短时记忆神经网络(Long short term memory,LSTM)算法预测道岔转辙设备后续健康因子曲线;最后,利用现场采集数据,对算法的有效性进行验证。实验结果表明:SOM方法可有效追踪道岔转辙设备健康状态变化规律,实现对健康因子的快速准确计算;相较于误差反向传播神经网络(Back propagation,BP神经网络)和循环神经网络(Recurrent neural network,RNN神经网络),LSTM算法预测效果较好,准确度提升,对道岔转辙设备的健康管理具有一定的指导意义。Abstract: Aiming at the problem that the fault of switch equipment occurs frequently and the staff can not accurately evaluate and predict its health state, the health state evaluation and prediction method of switch equipment combined with SOM-LSTM hybrid neural network is studied. Firstly, according to the characteristics of switch action power curve, the time-domain characteristic parameters are extracted in three segments, and the Health Index (HI) of switch equipment is solved by using the minimum quantization error in Self Organizing Map (SOM); Then the Long Short-term Memory Networks (LSTM) is used to predict the subsequent Health Index curve of switch equipment; Finally, the effectiveness of the algorithm is verified by using the field data. The experimental results show that SOM algorithm can effectively track the change law of health state of switch equipment and realize the rapid and accurate calculation of Health Index; Compared with Back Propagation (BP) neural network and Recurrent Neural Network (RNN) neural network, LSTM algorithm has better prediction effect and improved accuracy, which has certain guiding significance for the health management of switch equipment.
-
Key words:
- switch equipment /
- SOM /
- LSTM /
- health state /
- prediction
-
表 1 8种时域特征计算公式表
Table 1. Calculation formulas for 8 types of domain features
项目 公式 标准差(STD) $\sqrt{\dfrac{1}{C}{\displaystyle {\sum }_{c=1}^{C}{\left(F(c)-\frac{1}{C}{\displaystyle {\sum }_{c=1}^{C}F(c)}\right)^2}} }$ 峰度(KT) $\dfrac{ {\dfrac{1}{C}\displaystyle\sum\nolimits_{i = 1}^C {F{ {(i)}^4} } } }{ { {\sigma ^4} } }$ 峰峰值(P2P) $\max (F) - {\text{min} }(F)$ 偏斜度(SK) $\dfrac{ {\dfrac{1}{C}\displaystyle\sum\nolimits_{i = 1}^C {F{ {(i)}^2} } } }{ { {\sigma ^2} } }$ 均方根(RMS) $\sqrt {\dfrac{1}{C}\displaystyle\sum\nolimits_{c = 1}^C {F(c)} {}^2}$ 波峰因子(CF) $\dfrac{ {{\rm{MAX}}(F)} }{ {{\rm{RMS}}} }$ 均值(Mean) $\dfrac{1}{C}\displaystyle\sum\nolimits_{c = 1}^C {F(c)}$ 方差(Var) $\dfrac{1}{C}\displaystyle\sum\nolimits_{c = 1}^C { { {\left(F(c) - \frac{1}{C}\displaystyle\sum\nolimits_{c = 1}^C {F(c)} \right)}^2} }$ 表 2 不同预测算法实验结果对比
Table 2. Comparison of experimental results of differentprediction algorithms
算法平均相对误差
MRE/%均方根误差
RMSE/%决定系数R2 LSTM 0.473 5.048 0.703 BP 1.071 5.822 0.614 RNN 0.583 5.631 0.587 -
[1] 钟志旺, 陈建译, 唐涛, 等. 基于SVDD的道岔故障检测和健康评估方法[J]. 西南交通大学学报, 2018, 53(4): 842-849. doi: 10.3969/j.issn.0258-2724.2018.04.024ZHONG Z W, CHEN J Y, TANG T, et al. SVDD-Based research on railway-turnout fault detection and health assessment[J]. Journal of Southwest Jiaotong University, 2018, 53(4): 842-849. (in Chinese) doi: 10.3969/j.issn.0258-2724.2018.04.024 [2] 黄世泽, 陈威, 张帆, 等. 基于弗雷歇距离的道岔故障诊断方法[J]. 同济大学学报(自然科学版), 2018, 46(12): 1690-1695. doi: 10.11908/j.issn.0253-374x.2018.12.011HUANG S Z, CHEN W, ZHANG F, et al. Method of turnout fault diagnosis based on fréchet distance[J]. Journal of Tongji University (Natural Science), 2018, 46(12): 1690-1695. (in Chinese) doi: 10.11908/j.issn.0253-374x.2018.12.011 [3] 钟志旺. 铁路道岔健康状态评估与预测方法研究[D]. 北京: 北京交通大学, 2019.ZHONG Z W. Research on methods of health condition assessment and prediction of railway turnout[D]. Beijing: Beijing Jiaotong University, 2019. (in Chinese) [4] 杨菊花, 于苡健, 陈光武, 等. 基于CNN-GRU模型的道岔故障诊断算法研究[J]. 铁道学报, 2020, 42(7): 102-109.YANG J H, YU Y J, CHEN G W, et al. Research on turnout fault diagnosis algorithms based on CNN-GRU model[J]. Journal of the China Railway Society, 2020, 42(7): 102-109. (in Chinese) [5] 魏文军, 刘新发. 基于EEMD多尺度样本熵的S700K转辙机故障诊断[J]. 中南大学学报(自然科学版), 2019, 50(11): 2763-2772.WEI W J, LIU X F. Fault diagnosis of S700K switch machine based on EEMD multiscale sample entropy[J]. Journal of Central South University (Science and Technology), 2019, 50(11): 2763-2772. (in Chinese) [6] 高利民, 许庆阳, 李锋, 等. 基于SOM-BP混合神经网络的道岔设备退化状态研究[J]. 中国铁道科学, 2020, 41(3): 50-58. doi: 10.3969/j.issn.1001-4632.2020.03.06GAO L M, XU Q Y, LI F, et al. Research on degradation state of turnout equipment based on SOM-BP hybrid neural network[J]. China Railway Science, 2020, 41(3): 50-58. (in Chinese) doi: 10.3969/j.issn.1001-4632.2020.03.06 [7] YILBOGA H, EKER Ö F, GÜÇLÜ A, et al. Failure prediction on railway turnouts using time delay neural networks[C]//Proceedings of 2010 IEEE International Conference on Computational Intelligence for Measurement Systems and Applications. Taranto: IEEE, 2010: 134-137. [8] YANG J H, LI X T, XING D F, et al. Turnout fault diagnosis based on DBSCAN/PSO-SOM[J]. Journal of Measurement Science and Instrumentation, 2022, 13(3): 371-378. [9] ZHANG K. The railway turnout fault diagnosis algorithm based on BP neural network[C]//Proceedings of 2014 IEEE International Conference on Control Science and Systems Engineering. Yantai: IEEE, 2014: 135-138. [10] 李园祥. 基于神经网络的道岔转辙机退化模型的研究[D]. 北京: 北京交通大学, 2020.LI Y X. Research on degradation model of turnout switch based on neural network[D]. Beijing: Beijing Jiaotong University, 2020. (in Chinese) [11] HE Y M, ZHAO H B, TIAN J, et al. Railway turnout fault diagnosis based on support vector machine[J]. Applied Mechanics and Materials, 2014, 556-562): 2663-2667. [12] 钟健康, 陈元华, 张瑞宾. SOM与EWMA在滚动直线导轨故障预测中的应用[J]. 机械科学与技术, 2022, 41(2): 278-287.ZHONG J K, CHEN Y H, ZHANG R B. Application of SOM and EWMA in fault prediction of rolling linear guideway[J]. Mechanical Science and Technology for Aerospace Engineering, 2022, 41(2): 278-287. (in Chinese) [13] 王建利. 滚动轴承性能退化评价与趋势预测研究[D]. 大连: 大连理工大学, 2013.WANG J L. Research on performance degradation assessment and trend prediction of rolling bearing[D]. Dalian: Dalian University of Technology, 2013. (in Chinese) [14] 王磊, 刘永强. 基于不平衡样本的CF-SOM-MQE感应电机状态分析[J]. 华南理工大学学报(自然科学版), 2019, 47(3): 30-36.WANG L, LIU Y Q. State analysis of induction motor based on CF-SOM-MQE under unbalanced sample condition[J]. Journal of South China University of Technology (Natural Science Edition), 2019, 47(3): 30-36. (in Chinese) [15] KODIKARA G R L, MCHENRY L. Self-organizing Maps for identification of zeolitic diagenesis patterns in closed hydrologic systems on the Earth and its implications for Mars[J]. International Journal of Sediment Research, 2021, 36(5): 567-576. doi: 10.1016/j.ijsrc.2021.04.003 [16] 佘道明, 贾民平, 张菀. 一种新型深度自编码网络的滚动轴承健康评估方法[J]. 东南大学学报(自然科学版), 2018, 48(5): 801-806. doi: 10.3969/j.issn.1001-0505.2018.05.004SHE D M, JIA M P, ZHANG W. Deep auto-encoder network method for health assessment of rolling bearings[J]. Journal of Southeast University (Natural Science Edition), 2018, 48(5): 801-806. (in Chinese) doi: 10.3969/j.issn.1001-0505.2018.05.004 [17] 李婉婉. 基于LSTM的提速道岔故障预测研究[D]. 兰州: 兰州交通大学, 2020.LI W W. Study on fault prediction of speed-up switch based on LSTM[D]. Lanzhou: Lanzhou Jiaotong University, 2020. (in Chinese) [18] YUAN M, WU Y T, LIN L. Fault diagnosis and remaining useful life estimation of aero engine using LSTM neural network[C]//Proceedings of 2016 IEEE International Conference on Aircraft Utility Systems. Beijing: IEEE, 2016: 135-140. [19] 孙洁娣, 毛新茹, 温江涛, 等. 深度卷积长短期记忆网络的轴承故障诊断[J]. 机械科学与技术, 2021, 40(7): 1091-1099. doi: 10.13433/j.cnki.1003-8728.20200170SUN J D, MAO X R, WEN J T, et al. Bearing fault diagnosis using deep CNN and LSTM[J]. Mechanical Science and Technology for Aerospace Engineering, 2021, 40(7): 1091-1099. (in Chinese) doi: 10.13433/j.cnki.1003-8728.20200170