Efficient optimization identification method for soil stratification based on cone penetration test and joint posterior distribution of variable dimensionality

The mechanical stratification of soils based on the cone penetration test data (such as soil classification index I_c) is widely applied. However, the soil stratification based on I_c depends on engineering experience, and the subjective uncertainty is prominent. The mechanical stratification of soils is not necessarily consistent with the stratification based on borehole sampling. An efficient optimization identification method based on I_c data and the joint probability density function of mechanical profile parameters of soils is proposed under the framework of the Bayesian learning. The proposed method utilizes a full Gaussian probabilistic model to derive the likelihood function and then optimizes the maxima of the joint probability density function of mechanical profile parameters of soils using the simulated annealing algorithm. Subsequently, the number of soil layers and the associated soil thicknesses or boundaries are obtained by comparing the maxima of the joint probability density function with respect to the different numbers of soil layers. Finally, the rationality and validity of the proposed method are illustrated by a set of CPT data obtained from a subway section in Hangzhou and the simulated data, and the stratification principle and characteristics of the proposed method are illustrated with the identification results of soil profiles. The results show that the calculation efficiency of the proposed method for identifying the mechanical stratification of soils based on the I_c data is significantly improved, and it is suitable for analyzing the CPT data with different sounding depths. The calculation procedure of the proposed approach is relatively simple and is convenient for engineering applications.