基于时间序列的动态神经网络沉降预测
|
摘要
针对常见的非线性预测模型预测地表沉降点预测精度不高且稳定性较差的问题,通过实测分析和理论建模的方式,建立了基于时间序列的动态神经网络预测模型对地表沉降进行预测。该模型不需考虑变形因子对监测点位沉降的影响所具有的模糊性,且预测过程简单便捷。选取某市轨道交通2号线沉降监测点进行验证,来预测模型的可行性。结果表明,预测模型预测沉降量与实测数据的最大绝对误差为0.684 5 mm,并且相对误差均小于4%,说明该模型预测精度高,稳定性好,能够短期内有效地预测地表的沉降。研究成果对地表沉降的动态预测具有很好的参考价值。
Aiming at the disadvantages of low prediction accuracy and poor stability of the common nonlinear prediction model for predicting the surface settlement,a dynamic neural network prediction model based on time series was used to predict surface settlement by measurement analysis and theoretical modeling. The model does not need to consider the fuzziness of the effect of the deformation factor on the monitoring site settlement,and the prediction process is simple and convenient. The subsidence monitoring points of Metro Line 2 in a city was selected in order to verify the feasibility of the prediction model. The result showed that the maximal absolute error between the prediction model and the measured data was 0.684 5 mm,and the relative error was less than 4%. It showed that the model has high prediction accuracy,good stability,and can effectively predict the surface settlement in a short time. The results have a good application and reference value for the dynamic prediction of surface subsidence.
Aiming at the disadvantages of low prediction accuracy and poor stability of the common nonlinear prediction model for predicting the surface settlement,a dynamic neural network prediction model based on time series was used to predict surface settlement by measurement analysis and theoretical modeling. The model does not need to consider the fuzziness of the effect of the deformation factor on the monitoring site settlement,and the prediction process is simple and convenient. The subsidence monitoring points of Metro Line 2 in a city was selected in order to verify the feasibility of the prediction model. The result showed that the maximal absolute error between the prediction model and the measured data was 0.684 5 mm,and the relative error was less than 4%. It showed that the model has high prediction accuracy,good stability,and can effectively predict the surface settlement in a short time. The results have a good application and reference value for the dynamic prediction of surface subsidence.
引文
[1]张王乐元,张荠丰,孙增林,等.基于灰色理论的公路工程施工造价动态控制研究[J].交通科技与经济,2017,19(1):72-74,80.
[2] Deng Z,Ke Y H,Gong H L.Land subsidence prediction in Beijing based on PS-In SAR technique and improved Grey-Markov mode[J]. Giscience&Remote Sensing,2017,54(6):797-818.
[3]沙爱敏,吕凡任,王晓东.基于非等时距GM(1,1)模型的公路路基沉降预测[J].施工技术,2017,46(19):134-138.
[4]张昆,郭菊彬,王鹰.基于时间序列的沉降预测[J].铁道建筑,2006(4):44-45.
[5]谢文国,高俊强,徐东风.时间序列在地铁隧道变形监测数据处理中的应用[J].交通科技与经济,2015,17(4):114-116.
[6]聂侥,吴建军.在线时间序列预测方法及其应用[J].北京工业大学学报,2017,43(3):386-393.
[7]陈舒阳,徐林荣,曹禄来.基于动态神经网络法考虑区域沉降的高速铁路沉降预测[J].铁道学报,2015,37(5):83-87.
[8]徐兰花,曹福才,高宁波.基于神经网络的高速铁路安全态势预测方法研究[J].交通科技与经济,2016,18(6):35-38.
[9]张慧梅,李云鹏,毛成.人工神经网络在软土地基路基沉降预测中的应用[J].长安大学学报:自然科学版,2002(4):20-22.
[10]刘文豪,黎曦,胡伍生.基于神经网络和双曲线混合模型的高速公路沉降预测[J].东南大学学报:自然科学版,2013,43(S2):380-383.
[11]彭涛,杨岸英,梁杏,等.BP神经网络-灰色系统联合模型预测软基沉降量[J].岩土力学,2005(11):119-123.
[12]李长冬,唐辉明,胡斌,等.小波分析和RBF神经网络在地基沉降预测中的应用研究[J].岩土力学,2008(7):1 917-1 922.
[13]王蕊颖,王清,张颖,等.基于时间序列—动态神经网络吹填土沉降预测研究[J].工程地质学报,2013,21(3):351-356.
[14]禹建丽,黄鸿琦.神经网络在复杂自相关预测过程中的应用及对比研究[J].数学的实践与认识,2016,46(19):212-220.
[2] Deng Z,Ke Y H,Gong H L.Land subsidence prediction in Beijing based on PS-In SAR technique and improved Grey-Markov mode[J]. Giscience&Remote Sensing,2017,54(6):797-818.
[3]沙爱敏,吕凡任,王晓东.基于非等时距GM(1,1)模型的公路路基沉降预测[J].施工技术,2017,46(19):134-138.
[4]张昆,郭菊彬,王鹰.基于时间序列的沉降预测[J].铁道建筑,2006(4):44-45.
[5]谢文国,高俊强,徐东风.时间序列在地铁隧道变形监测数据处理中的应用[J].交通科技与经济,2015,17(4):114-116.
[6]聂侥,吴建军.在线时间序列预测方法及其应用[J].北京工业大学学报,2017,43(3):386-393.
[7]陈舒阳,徐林荣,曹禄来.基于动态神经网络法考虑区域沉降的高速铁路沉降预测[J].铁道学报,2015,37(5):83-87.
[8]徐兰花,曹福才,高宁波.基于神经网络的高速铁路安全态势预测方法研究[J].交通科技与经济,2016,18(6):35-38.
[9]张慧梅,李云鹏,毛成.人工神经网络在软土地基路基沉降预测中的应用[J].长安大学学报:自然科学版,2002(4):20-22.
[10]刘文豪,黎曦,胡伍生.基于神经网络和双曲线混合模型的高速公路沉降预测[J].东南大学学报:自然科学版,2013,43(S2):380-383.
[11]彭涛,杨岸英,梁杏,等.BP神经网络-灰色系统联合模型预测软基沉降量[J].岩土力学,2005(11):119-123.
[12]李长冬,唐辉明,胡斌,等.小波分析和RBF神经网络在地基沉降预测中的应用研究[J].岩土力学,2008(7):1 917-1 922.
[13]王蕊颖,王清,张颖,等.基于时间序列—动态神经网络吹填土沉降预测研究[J].工程地质学报,2013,21(3):351-356.
[14]禹建丽,黄鸿琦.神经网络在复杂自相关预测过程中的应用及对比研究[J].数学的实践与认识,2016,46(19):212-220.