论文部分内容阅读
获得较高精度的软夹层横波速度和厚度是瑞雷波频散曲线反演的难点之一,尤其对一些低敏感性的软夹层而言,单纯依靠传统的算法改进以及多模式反演,反演效果往往不是非常显著。首次尝试采用算法改进、多模式及非线性贝叶斯定理相结合反演低敏感性软夹层。算法改进体现在,将阻尼惯性权和混沌思想融入到粒子群算法中,但改进算法并未解决软夹层模型低敏感性的困扰;为从反演解的角度分析评价影响反演精度因素,采用无偏Metropolis-Hastings sampling(MHS)方法对后验概率进行数值积分,并通过参数旋转提高采用效率,积分得到的1D和混合边缘概率分布以及参数相关系数矩阵等参数反应了反演解的不确定性和参数间相关性等信息。为解决低敏感性反演精度低问题,尝试采用贝叶斯信息准则(BIC),判断出最佳参数化模型,而此准则得到的最佳模型与理论模型更为吻合。应用非线性贝叶斯方法和BIC准则反演实测防渗墙数据,得到的反演剖面也与已知防渗墙结构较好吻合。
It is one of the difficulties to obtain the Rayleigh wave dispersion curve to obtain high-precision soft-mesa shear wave velocity and thickness. Especially for some low-sensitivity soft mezzanine, relying solely on the traditional algorithm to improve and multi-mode inversion, The effect is often not very significant. The first attempt to use algorithm to improve, multi-mode and non-linear Bayesian theorem inversion combined low sensitivity soft interlayer. The improvement of the algorithm is reflected in that damping inertia and chaos are integrated into PSO. However, the improved algorithm does not solve the problem of low sensitivity of the soft interlayer model. In order to analyze and evaluate the factors affecting the inversion accuracy from the perspective of inversion solution, The unbiased Metropolis-Hastings sampling (MHS) method numerically integrates the posterior probability, and improves the adoption efficiency by parameter rotation. The integral of the 1D and the mixed edge probability distribution as well as the parameter correlation coefficient matrix and other parameters reflect the uncertainty of the inversion solution Sex and the correlation between parameters and other information. In order to solve the problem of low sensitivity of low sensitivity inversion, try to use Bayesian Information Criterion (BIC) to determine the best parameterized model, and the best model obtained by this criterion is more consistent with the theoretical model. The non-linear Bayesian method and BIC criterion are used to inverse the data of the measured seepage-proof wall, and the obtained inversion section is in good agreement with the structure of the known seepage-proof wall.