Evaluation of Soil Properties at Pile Location Based on CPT Data Using Spectral Clustering and Hidden Markov Chain

Cone penetration test (CPT) is an in-situ experiment measurement with low disturbance to soil, widely used in geological exploration. Using the CPT data to evaluate the soil distribution and property at pile location is an important goal for the design of pile structure. This paper aims to evaluate the needed soil properties for design at pile location with the limited CPT data at several boreholes. In this paper, a new soil classification method using spectral clustering (SC) is applied to mine the CPT data, tip resistance (qc) and side friction (fs), of 6 boreholes. Then a hidden Markov chain method (HMCM) is used to describe the distribution of soil type and qc, fs values of the site based on the results of soil classification and CPT data. Using the hidden Markov chain models, the qc and fs values at pile location can be evaluated only based on the qc and fs values of 6 boreholes. The result shows that this SC-HMCM method can evaluate the soil property at pile location accurately with limited geological exploration data and provide reference for design of pile.