基于RVM-CS的工程区地层渗透系数反演分析

为了准确确定工程区各地层渗透系数,基于相关向量机(RVM)与布谷鸟算法(CS)建立反演分析模型,对研究区域内各地层的渗透系数进行反演分析。首先,基于均匀设计法构造渗透系数组合,利用有限元计算测点水位值并生成RVM学习样本;随后,通过训练RVM构建渗透系数与水头之间的映射关系,使其能够代替有限元方法快速地完成渗流计算;最后,根据工程区钻孔的实测水位值,利用CS算法搜索确定区域内各地层的渗透系数。将建立的渗流反演模型应用于某大型抽水蓄能电站工程区的初始渗流场中,结果表明:所建立的渗流反演模型能够考虑多地层渗透系数和钻孔水头间的非线性,RVM能代替有限元模型,可快速、精确地确定渗透系数;此渗流反演模型在大型抽水蓄能电站工程区反演结果合理,精度满足工程要求。

Inversion Analysis on Permeability Coefficient of Stratum in Engineering Area Based on RVM-CS

An inversion analysis model integrating relevance vector machine (RVM) and cuckoo search (CS) is established to accurately determine the permeability coefficients of strata in engineering area. Firstly, the uniform design method is employed to construct combinations of permeability coefficients, and the finite element method is used to calculate the water head values and generate RVM learning samples. In subsequence, the mapping relation between permeability coefficient and water head is constructed by RVM training which replaces the finite element method in calculating seepage. According to the measured water head values of drilling holes in the project area, the CS algorithm is adopted to search and determine the permeability coefficient of stratum. The seepage inversion model is applied to the inversion of initial seepage field of a large pumped storage power station project. Results demonstrate that the proposed model reflects the nonlinear relation between water head in borehole and permeability coefficient of multiple strata. RVM could replace the finite element method to determine quickly and accurately the permeability coefficient. The inversion results for the large pumped storage power station are reasonable and the accuracy of the proposed model meets engineering requirements.