Cone penetration data classification by Bayesian inversion with a Hidden Markov model

A Krogstad,I Depina,H Omre

doi:10.1088/1742-6596/1104/1/012015

Abstract

This study examines the application of the Hidden Markov model (HMM) to the soil classification based on Cone Penetration Test (CPT) measurements. The HMM is formulated in the Bayesian framework and composed of a Markov chain prior and a Gaussian likelihood model. The application of the Bayesian framework is considered as suitable because it allows for the integration of different sources of information commonly available in a CPT-based soil classification. The occurrence of different soil classes along a CPT profile is modeled with the Markov chain, while the Gaussian likelihood model establishes a relation between the different soil classes and CPT measurements. Preliminary performance of the HMM is examined on the classification of CPT measurements from the Sheringham Shoal Offshore Wind Farm.

Highlights

The Cone Penetration Test (CPT) is an in-situ test that is frequently applied to estimate subsurface stratigraphy, soil parameters, and parameters for a direct geotechnical design [1]
This study examines the application of the Hidden Markov model (HMM) to the soil classification based on Cone Penetration Test (CPT) measurements
This study examined the application of the Hidden Markov Model to the soil classification based on CPT measurements

Summary

Introduction

The Cone Penetration Test (CPT) is an in-situ test that is frequently applied to estimate subsurface stratigraphy, soil parameters, and parameters for a direct geotechnical design [1]. This study follows the development path of Bayesian approaches and investigates the application of the Hidden Markov Model (HMM) to the CPT classification problem. The HMM is considered to be applicable to the CPT classification problem because it provides a joint model for the spatial distribution of soil classes and the relation between CPT measurements and different soil classes. The solution to the CPT classification problem with the HMM is found in the Bayesian framework, which allows the incorporation of additional information on soil classes. The likelihood model depends on the data acquisition procedure, with the parameters of the likelihood model assessed by using the actual soil class vector of the training well, κ, along with the CPT data With these two distributions in place the full posterior is defined. The evaluation of the normalizing constant is usually unfeasible and avoided in the majority of the implementations

Likelihood model

Posterior model inference

Conclusions