Determination of Optimal CPT Locations for Characterizing Nonstationary Spatial Variability of Geotechnical Properties Using Efficient Bayesian Compressive Sensing

Spatial variability of geotechnical properties within multiple soil layers plays an essential role in geotechnical design or analysis, especially in probability-based analysis for geo-structures (e.g., slopes, tunnels, piles). This is often determined via laboratory methods using samples from drilled boreholes, alternatively in-situ testing methods, such as cone penetration (CPT). Note that CPT has been widely used in recent decades, because it is fast, inexpensive, repeatable, and can be obtain almost continuous soil response data when its cone is pushed into the ground. Because of time and/or technical constraints, the number of CPT in a specific site is often small. Besides, note that subsurface conditions are often inhomogeneous, and CPT at different locations may reveal different spatial variability of geotechnical properties in terms of accuracy. In this case, it is of great interest, but of great difficulty, to determine the optional locations for CPT soundings such that as accurate as possible information on multi-layer geotechnical properties can be obtained. This is often encountered during the multi-stage geotechnical site characterization, and additional CPT locations are often needed in later site characterization. This paper presents an efficient Bayesian compressive sensing method for addressing this issue, which consists of two components: 1) information entropy for determination of optimal CPT locations, and 2) kronecker product to improve its computational efficiency given almost continuous CPT data. The method is demonstrated using numerical datasets. The results indicate that the locations determined by the presented method are effective and can properly characterize the spatial variability of multiple soil layers. ©2022 ISGSR Organizers. Published by Research Publishing, Singapore.