基于支持向量机的隧道围岩位移变形预测分析研究

发布时间：2018-02-04 06:47

本文关键词： 围岩位移时间序列预测支持向量机小波核函数　出处：《重庆交通大学》2015年硕士论文　论文类型：学位论文

【摘要】：新奥法是隧道工程建设中广泛使用的方法。监控量测是新奥法中的重要组成部分,量测的数据能够体现隧道围岩的变形状态,依托监测数据对隧道围岩的变形进行预测能够很好的为隧道的设计和施工提供参考价值,现今的位移反演分析能够对隧道进行一定范围内的预测,但反演分析主要依托在沉降稳定值的反推与邻近界面的材料特性相似性上,有一定的滞后性不能满足工程建设中时间投资上的要求。利用量测数据的自规律性对位移变形进行预测的研究还较少。针对位移变化的时间序列预测上,本文依托统计学习理论中的支持向量机开展研究,主要工作有:①介绍了支持向量机的理论基础与具体的推导过程,利用阳宗隧道的位移监测数据实例分析了不同超参数组合下的支持向量机性能,提出了参数选择的方法与依据。对比了多项式回归预测与递归最小二乘法预测的差异,并用MATLAB编程依托推导的原理进行了计算分析,结果表明支持向量机具有较好的稳定性。②以京珠高速公路温泉隧道位移非线性时间序列预测为例子,对比了不同特征表现形式下的RBF核支持向量机预测性能。分析结果表明,采用累加和的位移表现形式能有效提升RBF核函数下的支持向量机预测性能,对比了传统归一化方法与对数空间映射的差异性,分析结果表明对于同一个映射空间下的子集选择RBF核函数不敏感。③针对RBF核函数优良的特性,通过小波理论与核函数的特性构造出了Morlet小波核函数,并经过溪洛渡电站隧道的监测数据的分析验证,Morlet小波核函数在监测频率不同的位移变形预测分析上,有较好的预测性能。
[Abstract]:New Olympic method is widely used in tunnel construction. Monitoring and measurement is an important part of the method. The measured data can reflect the deformation state of tunnel surrounding rock. Relying on monitoring data to predict the deformation of tunnel surrounding rock can provide a good reference value for the design and construction of the tunnel. Now the displacement inversion analysis can predict the tunnel in a certain range. But the inversion analysis is mainly based on the similarity of the material properties between the stability value of settlement and the material characteristics of the adjacent interface. Some lag can not meet the requirements of time investment in engineering construction. There is little research on the prediction of displacement deformation by using the self-regularity of measurement data. Based on the support vector machine (SVM) in statistical learning theory, the main work of this paper is to introduce the theoretical basis and the specific derivation process of support vector machine (SVM). Based on the displacement monitoring data of Yangzong tunnel, the performance of support vector machine (SVM) under different super-parameter combinations is analyzed. The method and basis of parameter selection are put forward, the difference between polynomial regression prediction and recursive least square prediction is compared, and the calculation and analysis are carried out based on the principle of MATLAB programming. The results show that the support vector machine has good stability. 2. The prediction of nonlinear time series of displacement in hot spring tunnel of Jingzhou-Zhuhai Expressway is taken as an example. The prediction performance of RBF kernel support vector machine with different feature forms is compared. The analysis results show that the prediction performance of support vector machine under RBF kernel function can be improved effectively by using the displacement representation of cumulative sum. The difference between the traditional normalization method and the logarithmic space mapping is compared. The results show that the RBF kernel function is insensitive to the RBF kernel function for the subset of the same mapping space. The Morlet wavelet kernel function is constructed by wavelet theory and kernel function, and verified by the monitoring data of Xiluodu Hydropower Station. Morlet wavelet kernel function has good prediction performance in monitoring displacement and deformation prediction with different frequencies.
【学位授予单位】：重庆交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U456.31

【参考文献】