基于PSO-BP优化算法的围岩参数智能反分析研究

发布时间：2018-05-01 08:03

本文选题：围岩参数 + PSO-BP算法　；参考：《中南大学》2014年硕士论文

【摘要】：：如何通过比较简单的方法获得准确的隧道围岩力学参数,是近代隧道科技工作者和工程技术人员在积极研究的课题。反分析法是确定隧道围岩力学参数比较好的方法之一。本文结合五道梁隧道项目,提出了BP神经网络和PSO优化算法相结合的混合算法,在ASP.NET平台上开发了隧道围岩参数智能反分析系统,并在实际工程中得到了应用。本文的具体工作内容包括以下三个方面： (1)介绍了BP神经网络和粒子群(PSO)优化算法的基本理论,利用PSO算法的全局寻优能力对BP神经网络的权值和阈值进行了优化,并在MATLAB上利用神经网络工具箱函数编程实现,最后利用常用的测试函数对PSO算法的寻优能力进行了测试； (2)在微软的ASP.NET平台上利用.NET与MATLAB混合编程理论开发了基于Web方式的隧道围岩参数智能反分析系统,实现了项目管理、监测数据的录入、回归分析、参数反演等功能,系统界面良好、操作简便,能够对围岩参数的反分析起到推广作用； (3)结合五道梁隧道现场情况,并利用正交试验设计和均匀试验设计构造了神经网络的学习和测试样本,利用隧道围岩参数智能反分析系统对五道梁隧道DK236+650断面进行了参数反演计算,并对计算结果进行了分析。最后,将反演出的围岩参数代入FLAC3D正分析程序进行了隧道开挖的位移计算,并将计算结果与实测结果进行了对比分析。
[Abstract]:How to obtain accurate mechanical parameters of tunnel surrounding rock by relatively simple method is an active research topic for modern tunnel science and technology workers and engineers. Inverse analysis is one of the better methods to determine the mechanical parameters of tunnel surrounding rock. In this paper, a hybrid algorithm based on BP neural network and PSO optimization algorithm is proposed, and an intelligent back analysis system for surrounding rock parameters of tunnel is developed on ASP.NET platform, which is applied in practical engineering. The specific work of this paper includes the following three aspects: In this paper, the basic theory of BP neural network and particle swarm optimization algorithm is introduced. The weight and threshold of BP neural network are optimized by using the global optimization ability of PSO algorithm, and realized by using neural network toolbox function on MATLAB. Finally, the optimization ability of PSO algorithm is tested by using common test functions. Based on the mixed programming theory of .NET and MATLAB, an intelligent back-analysis system of tunnel surrounding rock parameters based on Web is developed on Microsoft's ASP.NET platform. The functions of project management, monitoring data input, regression analysis, parameter inversion and so on are realized, such as project management, monitoring data input, regression analysis, parameter inversion and so on. The system has good interface and easy operation, which can popularize the back analysis of surrounding rock parameters. (3) considering the field situation of Wudaoliang tunnel and using orthogonal experimental design and uniform experimental design, the learning and testing samples of neural network are constructed. The parameter inversion calculation of DK236 650 section of Wudaoliang tunnel is carried out by using the intelligent back analysis system of surrounding rock parameters of tunnel, and the calculation results are analyzed. Finally, the displacement calculation of tunnel excavation is carried out by adding the parameters of surrounding rock back to the FLAC3D program, and the calculated results are compared with the measured results.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U451.2

【参考文献】