基于贝叶斯压缩感知与SVM算法的智能化勘察研究

Research on Intelligent Reconnaissance Based on Bayesian Compressed Sensing and SVM Algorithm

在当前工程勘察中,现场测得的地层数据较为稀疏,且钻孔与钻孔之间由于间隔较大,土层的性质依靠传统经验难以准确判定。文章研究基于贝叶斯压缩感知原理,对钻孔测得的比贯入阻力进行插值,通过迭代计算得到插值矩阵。预测钻孔之间的比贯入阻力大小,并将计算结果可视化,从而得到整个研究剖面上比贯入阻力的分布图,并对插值结果做准确率分析。同时收集场地周边的地质勘察资料,利用支持向量机(SVM)做出土层分类模型,得出了较为准确的土层分类结果。为工程勘察的数据扩充和地层划分提供一种新技术、新方法。

In the current engineering investigation,the stratum data measured in the field is relatively sparse,and because of the large interval between boreholes and boreholes,it is difficult to accurately determine the properties of the soil layer based on traditional experience.In this study,based on Bayesian compression sensing principle,the specific penetration resistance measured by boreholes is interpolated,and the interpolation matrix is obtained by iterative calculation.The specific penetration resistance between boreholes is predicted,and the calculation results are visualized,thus the distribution map of the specific penetration resistance on the whole research section is obtained,and the accuracy of the interpolation results is analyzed.At the same time,the geological survey data around the site are collected,the soil layer classification model is made by using support vector machine(SVM),and a more accurate soil layer classification result is obtained.It provides a new technology and method for data expansion and stratigraphic division of engineering investigation.