基于多源异构数据下的地铁隧道掘进机预测模型研究  

Metro tunnel boring machine prediction model based on multi-source heterogeneous data

作　　者：赵雷 ZHAO Lei(China Railway 19th Bureau Group Rail Transit Engineering Co.,Ltd.,Beijing 101300,China)

机构地区：[1]中铁十九局集团有限公司,北京101300

出　　处：《现代城市轨道交通》2024年第12期103-109,共7页Modern Urban Transit

摘　　要：文章以某地铁区间现场数据为依托,建立隧道掘进机掘进速度LSTM预测模型,在此基础上,对比分析地质参数对该预测模型的精确度影响,为挖掘出更有效的信息,将各影响指标先进行动态因子模型降维后,再进行数据训练。结论如下:①LSTM时间序列模型可以很好的预测隧道掘进机掘进速度,预测精确率高达99.02%,且F1值均在95%以上;②考虑地质参数对隧道掘进机掘进速度的影响后,模型预测精确率由96.91%提高至99.02%;③通过动态因子模型数据降维后再进行数据训练,模型预测精确率从98.96%提高至99.02%。相关研究可为隧道掘进机重大掘进装备的多源异构混合数据建模提供参考和借鉴。Based on the on-site data of a certain metro section,this article establishes an LSTM prediction model for tunnel boring machine excavation speed.On this basis,the influence of geological parameters on the accuracy of the prediction model is compared and analyzed.In order to excavate more effective information,the various influencing indicators are subjected to dynamic factor model dimensionality reduction in the first place before data training.The conclusion is as follows:①The LSTM time series model can effectively predict the excavation speed of tunnel boring machines,with a prediction accuracy of up to 99.02%and F1 values all above 95%;②After considering the influence of geological parameters on the excavation speed of tunnel boring machines,the model prediction accuracy increased from 96.91%to 99.02%;③By reducing the dimensionality of the dynamic factor model data and then carry out data training,the model’s prediction accuracy increased from 98.96%to 99.02%.The relevant research can provide reference and inspiration for multi-source heterogeneous mixed data modeling of large boring equipment for tunnel boring machines.

关 键 词：地铁 隧道掘进机 多源异构 地质参数 动态因子模型 LSTM预测模型 

分 类 号：U231.3[交通运输工程—道路与铁道工程]