盾构管片结构应力混沌时间序列预测模型

发布时间：2018-06-22 09:27

本文选题：地铁工程 + 盾构隧道施工　；参考：《华中科技大学》2015年硕士论文

【摘要】：盾构在始发、穿越风井、接收、联络通道等重要区段的施工会造成管片受力的变化,进而引起管片的变形,严重时会出现裂缝,导致隧道内涌水涌砂现象,发生灾难性的后果。目前管片结构应力监测技术日趋成熟,施工安全在一定程度上得到了保障,但在风险应对机制还不是十分完善的情况下,利用预测模型来争取到更多的风险应对时间,可以更好地保证施工安全,减少事故造成的损失。本文以武汉地铁四号线盾构隧道工程实测数据为基础,通过数据的初步处理,指出了管片结构应力时序变化的一般规律。对管片结构应力的影响因素进行分析,从定性的角度说明了在盾构掘进过程中,管片结构应力变化具有混沌性。结合混沌理论与BP神经网络,构建了混沌时间序列预测模型,将工程实测数据代入模型,分析了数据的混沌性,并进行管片结构应力预测分析。随机选取一个2000组的管片结构应力时序段作为混沌时序预测的样本,基于MATLAB软件,完成管片结构应力混沌时序重要参数的计算。其中C-C法算得的嵌入维数m?5和延迟时间??3,将算得参数结合坐标延迟法,实现相空间的重构。Wolf法算得的混沌时序的最大Lyapunov指数λ=0.64870,验证了时间序列的混沌性。重构后的相空间为BP神经网络提供了输入层、输出层数据和输入层节点数。根据嵌入维数初始化BP神经网络结构。进行神经网络训练,得出预测结果,利用MSE(均方误差)、MPE(平均百分比误差)、最大最小绝对误差、最大最小误差百分比来衡量预测结果的准确度,其中MSE=0.03130.05,MPE=0.042%0.1%,衡量结果表明混沌时间序列预测模型对于管片结构应力的预测能保证较好的精度。
[Abstract]:The construction of shield machine in the beginning, passing through the air well, receiving, connecting channel and other important sections will cause the change of the stress of the segment, and then cause the deformation of the segment, and when it is serious, the cracks will appear, which will lead to the phenomenon of water gushing and sand gushing in the tunnel, which will have disastrous consequences. At present, the stress monitoring technology of segment structure is maturing day by day, and the construction safety is guaranteed to a certain extent, but under the condition that the risk response mechanism is not very perfect, the prediction model is used to gain more risk response time. Can better guarantee the construction safety, reduce the loss caused by the accident. Based on the measured data of shield tunneling project of Wuhan Metro Line 4, this paper points out the general law of stress time series change of segment structure through the preliminary processing of the data. The influencing factors of segment structure stress are analyzed and the chaos of segment structure stress variation in shield tunneling process is explained qualitatively. Based on the chaos theory and BP neural network, the chaotic time series prediction model is constructed, the engineering measured data is substituted into the model, the chaos of the data is analyzed, and the stress prediction of the segment structure is analyzed. A random 2000 group of segment structural stress time series was selected as the sample of chaotic time series prediction. Based on MATLAB software, the important parameters of segment structure stress chaotic time series were calculated. The embedding dimension m5 and the delay time are calculated by C-C method. The maximum Lyapunov exponent 位 _ (0.64870) of the chaotic time series calculated by the reconstruction of phase space. Wolf method is realized by combining the parameters with the coordinate delay method, which verifies the chaos of the time series. The reconstructed phase space provides input layer, output layer data and input layer node points for BP neural network. The BP neural network structure is initialized according to the embedding dimension. Neural network training is carried out to obtain the prediction results. The accuracy of the prediction results is measured by MSE (mean square error) MPE (mean percentage error), maximum and minimum absolute error, maximum and minimum error percentage. Among them, MSE 0.03130.05 MPEG 0.0420.1, the measured results show that the prediction model of chaotic time series can ensure a good accuracy for the prediction of segment structure stress.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U455.43

【参考文献】