在岩土工程实践中对于诸如各向异性岩体的应力一渗流耦合分析等复杂问题,建立用于同时确定多类参数的反分析方法具有重要意义。以位移量测信息为基础信息,对基于复变量微分法的优化算法(Levenberg-Marquit法)引入异步长因了矩阵Y[k],建立同时反演确定初始地应力、围岩性态参数和渗透系数的协同优化变步长反演算法。对不同的参数借助搜索迭代过程,确定不同的步长因子系数。异步长因子矩阵可使在峰值附近不收敛的参数重新选择搜索方向,使其收敛至精确值。算例表明,异步长因子矩阵可成功地用于协调各类参数敏感性之间的差异,从而有效改进多参数耦合反演的收敛性和收敛速度,使算法具有更好的适应性和有效性,为岩土工程的复杂问题的求解提供一种方法。
A new method for determining of parameters simultaneously m the coupled seepage-stress analysis m anmotroplc rock is essential for large-scale cavern shape selection, stability analysis, and rock support system design. Based on the Levenberg-Marquit optimization method, a modified method has been put forward by employing a new step factor matrix y(k), in order to calculate both of the parameters of rock mass modulus (E), the in situ stress (P) and the permeability coefficient (K) in the back analysis. Compared with traditional step factor in optimization algorithm, the new step factor matrix y(k) can change the step factors respectively according to the different sensitivities of parameters, with which a new algorithm is established on MATLAB platform. Two case studies show the method's feasibility and validity in the coupled seepage-stress analysis in a anisotropic rock tunnel engineering; and the results demonstrate that the modified method with new step factor can improve convergence of iterative algorithm obviously which cannot be solved with traditional step factor.