基于高斯过程—差异进化算法的隧道施工多元信息反分析研究

发布时间：2019-03-30 12:31

【摘要】：本文通过对现有国内外隧道工程监测、围岩参数反分析、隧道时间序列预测的研究和总结的基础上,重点研究并完成了以下内容：(1)提出高斯过程-差异进化协同优化算法(GP-DE),开发基于matlab的GP-DE程序,发挥高斯过程对非线性映射关系优良的处理能力,利用差异进化算法优化GP-DE包含的超参数,有效提高非线性映射关系模型的精度,为地下工程岩体参数智能优化反分析及岩体变化时间序列预测提供一种新方法。(2)结合大连地铁实际工程,在工程现场布置多元信息自动化监测系统,以获得更加及时、丰富、准确的围岩变化信息,并对所采集信息进行了综合分析。另设计正交试验方案进行数值计算,对围岩参数进行敏感性分析,分析各围岩参数对控制变量的影响。利用现场监测所得围岩位移应力变化信息作为控制值,进行围岩参数GP-DE位移应力联合反分析,并与围岩参数DE位移反分析作对比,验证方法的优越性。(3)结合陈家店山岭隧道实际工程,考虑加入渗流作用,通过数值计算、围岩参数敏感性分析及围岩参数GP-DE反分析,获得现场围岩参数,并通过数值计算,将模拟结果与实际对比,对反分析结果进行验证。另利用反分析参数进行模拟,对不同工法、有无渗流作用的施工结果作对比。(4)利用GP-DE算法对大连地铁隧道拱顶沉降时间序列进行预测，，通过主成分分析法优化了训练样本,实现了隧道拱顶沉降值和监测断面与掌子面距离的二变量时间序列预测,另外比较不同样本构成方法、单一变量时间序列与多变量时间序列的预测效果。
[Abstract]:Based on the research and summary of existing tunnel engineering monitoring, back analysis of surrounding rock parameters and prediction of tunnel time series at home and abroad, The main contents are as follows: (1) the Gao Si process-differential evolution collaborative optimization algorithm (GP-DE) is proposed, and the GP-DE program based on matlab is developed. The Gao Si process has good processing ability to the nonlinear mapping relationship. In order to improve the precision of nonlinear mapping relation model, differential evolution algorithm is used to optimize the superparameters contained in GP-DE. This paper provides a new method for intelligent optimization of rock mass parameters and prediction of time series of rock mass change in underground engineering. (2) combined with the actual project of Dalian Metro, multi-information automatic monitoring system is arranged in the project site to get more timely. Rich and accurate surrounding rock change information, and comprehensive analysis of the collected information. In addition, the orthogonal test scheme is designed for numerical calculation, the sensitivity analysis of surrounding rock parameters is carried out, and the influence of surrounding rock parameters on the control variables is analyzed. Using the displacement stress change information obtained from field monitoring as the control value, the joint back analysis of surrounding rock parameter GP-DE displacement stress is carried out and compared with the back analysis of surrounding rock parameter DE displacement. The advantages of the method are verified. (3) combined with the actual project of Chenjiadianshan Tunnel, the surrounding rock parameters are obtained by numerical calculation, sensitivity analysis of surrounding rock parameters and GP-DE back analysis of surrounding rock parameters. Through numerical calculation, the simulation results are compared with the actual ones, and the inverse analysis results are verified. In addition, the inverse analysis parameters are used to simulate and compare the construction results of different construction methods with or without seepage. (4) the GP-DE algorithm is used to predict the settlement time series of the arch roof of Dalian metro tunnel. The training samples are optimized by principal component analysis, and the two-variable time series prediction of the settlement value of tunnel arch roof and the distance between the monitoring section and the palm surface is realized. In addition, different sample composition methods are compared. Prediction effect of single variable time series and multivariable time series.
【学位授予单位】：大连海事大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U455

【引证文献】