摘要
为了更好地挖掘高速铁路在地震时的响应信息,提高光纤光栅监测的效率及预测精度,该文针对地震响应数据的时序性及非线性的特点,提出卷积神经网络(CNN)和长短期记忆(LSTM)网络的混合神经网络模型预测方法.通过在高速铁路简支梁桥上布设准分布式光纤光栅采集地震时轨道板、钢轨、底座板、箱梁的响应数据,在每根光纤上布置7个光栅,利用两边光栅的响应数据预测中间点的光栅响应,将采集位置、历史数据及地震波形等信息作为特征图输入.利用CNN提取特征,再将提前提取出来的特征数据以时序方式作为LSTM网络的输入数据,最后LSTM网络进行地震应变响应预测.实验结果表明,LSTM网络在3层时效果最好,CNN-LSTM方法具有较高的预测精度,根均平方误差(R_(RMSE))、平均绝对误差(R_(MAE))、决定系数(R^(2))分别达到了0.3753、0.2968、0.9371.
This paper proposes a hybrid model prediction method based on convolutional neural network(CNN)and long short-term memory network(LSTM)for the characteristics of time-series and nonlinearity of earthquake response data to better exploit the response information of high-speed railroads during the earthquake and improve the efficiency and prediction accuracy of fiber grating monitoring.Seven gratings are arranged on the fiber by laying quasi-distributed fiber optic gratings on the shaking table of the high-speed railroad simple beam to collect the response data of the track plate,rail,base plate,and beam during the earthquake,and the response data of the two side gratings are used to predict the grating response of the middle point,and the continuous feature map is constructed as the input by the time-sliding window of the acquisition location,historical data,and seismic.The CNN is utilized first to extract the feature vector,which is then produced and used as the input data of the LSTM network in a time-series sequence,and ultimately,the LSTM network is used for prediction.The experimental findings reveal that the LSTM network performs best at three layers,and the CNN-LSTM approach has a good prediction accuracy with R_(RMSE),R_(MAE),R^(2)at 0.3753,0.2968,and 0.9371,respectively.
作者
张学兵
谢啸楠
王礼
吴晗
ZHANG Xuebing;XIE Xiaonan;WANG Li;WU Han(College of Civil Engineering,Xiangtan University,Xiangtan 411105,China)
出处
《湘潭大学学报(自然科学版)》
CAS
2024年第1期1-13,共13页
Journal of Xiangtan University(Natural Science Edition)
基金
湖南省教育厅重点项目(20K126)
高铁联合基金项目(U1934207)。