Autor: |
Shakir, Ressol R., Wang, Hui |
Předmět: |
|
Zdroj: |
Arabian Journal of Geosciences; Mar2023, Vol. 16 Issue 3, p1-22, 22p |
Abstrakt: |
This paper develops a Gaussian Mixture Model (GMM) method to produce the soil profile based on the cone penetration test results (CPT). The theoretical probabilistic model utilizes the Bayesian information criterion (BIC) and Akaike information criteria (AIC) to estimate the number of layers. The research analyzed the impact of four types of covariance matrices on soil classification. The proposed approach was applied to the real-life CPT data for the National Geotechnical Experimentation Site (NGES) at Texas, which was widely used for analysis and comparison. Evaluation of the proposed model with many previous classification systems indicated that the GMM could detect soil boundaries and types by clustering the data and employing Robertson chart. In addition to that, the clustering decision depended on the posterior probability of every soil unit to the corresponding clusters. The thickness of the thin layers and the location of boundaries rely on the type of covariance matrix. The study revealed that the four types of covariance give a range of layers from 5 to 7, corresponding to the minimum values of both BIC and AIC. The nonshared full covariance matrix reflected more layer boundaries and thin layers than the shared and diagonal covariance matrix. It is concluded that the proposed method did not require an experience-based decision to classify the soil and indicate the layer boundaries. Further, the proposed technique directly classified the soil in a fast process, expressed the results visually, and was more practical and familiar to geotechnical engineers. It was sensitive to detecting thin layers, which gives a chance to understand and interpret the CPT measurements. [ABSTRACT FROM AUTHOR] |
Databáze: |
Complementary Index |
Externí odkaz: |
|