基于多变量的集成预测模型在隧道拱顶沉降变形预测中的应用

发布时间：2018-03-31 01:32

本文选题：隧道工程　切入点：多变量模型　出处：《公路交通科技》2017年12期

【摘要】：隧道拱顶沉降是多种因素共同作用的一个十分复杂的过程,很难用数学模型进行精确计算。实际施工过程中,围岩情况经常发生变化,为满足设计要求,必须严格控制隧道拱顶沉降。传统的预测模型都只是利用拱顶沉降监测数据建立单变量模型进行拟合并预测。隧道开挖过程中,拱顶变形所受影响因素较多,导致监测数据序列中常常出现离散型较大的数据,单一变量模型预测精度受这些离散数据的影响较大,而且筛除离散型较大的数据会直接影响模型预测精度,因此单一模型只能对拱顶沉降量做粗略的估计。针对这一问题,根据隧道变形的同时性和内在相关性,利用拱顶变形监测数据和同期两侧收敛变形监测数据构建带输入变量的时序模型、GM(1,2)模型和BP模型分别对拱顶沉降变形进行预测,并运用实例验证了所建模型的有效性。通过对两种模型的预测精度进行对比可知,单一变量的时序模型只能对变形的趋势作出预测,预测精度较低,难以对拱顶沉降进行有效预测,而单一变量的GM(1,1)预测模型则完全失效。为了避免单一模型自身的缺陷导致预测精度降低,同时使不同模型间优势互补,本研究建立了基于以上3种带输入变量模型的集成预测模型,其加权系数采用熵值法确定;最后将该模型运用于宝汉高速白庙子隧道中进行检验,结果表明该集成模型更有效,预测精度更高。
[Abstract]:The settlement of tunnel vault is a very complicated process which is influenced by many factors, so it is difficult to calculate accurately by mathematical model.In the actual construction process, the surrounding rock situation often changes, in order to meet the design requirements, the tunnel vault settlement must be strictly controlled.The traditional prediction models only use the dome settlement monitoring data to establish a univariate model for pseudo-combined prediction.In the course of tunnel excavation, the deformation of arch roof is affected by many factors, which leads to the occurrence of large discrete data in monitoring data sequence, and the prediction accuracy of single variable model is greatly affected by these discrete data.Moreover, screening large discrete data will directly affect the prediction accuracy of the model, so a single model can only make a rough estimate of the settlement of the vault.In response to this problem, according to the simultaneous nature and intrinsic correlation of tunnel deformation,Based on the data of roof deformation monitoring and convergent deformation monitoring on both sides of the same period, a time series model with input variables is constructed, and the BP model and the time series model are used to predict the settlement deformation of the vault respectively. The validity of the model is verified by an example.By comparing the prediction accuracy of the two models, it can be seen that the single variable time series model can only predict the trend of deformation, and the prediction accuracy is low, so it is difficult to predict the settlement of arch roof effectively.The single variable GM-1) prediction model is completely ineffective.In order to avoid the defect of single model and make the different models complement each other, an integrated prediction model with input variables is established, and its weighting coefficient is determined by entropy method.Finally, the model is tested in Bao-Han high-speed Baimaozi tunnel. The results show that the integrated model is more effective and the prediction accuracy is higher.
【作者单位】：武汉理工大学土木工程与建筑学院;
【分类号】：U456.3

【相似文献】