<![CDATA[This manuscript studies the influence of spatial heterogeneity in soil effective cohesion c' on the displacement of an axially loaded pile under static compression. The Monte Carlo method was used to perform statistical evaluation, while the K-Means clustering technique was employed to reduce the necessary number of simulations without compromising the accuracy and reliability of the analysis results. In this study, a two-dimensional correlated random field of c' was generated using the spectral method, assuming a Gaussian distribution with a coefficient of variation of 10%. The spatial correlation lengths of the scaled model were set to 80 mm and 19.4 mm in the horizontal and vertical directions, respectively. The static pile loading process was simulated using the finite element method implemented in the commercial software PLAXIS, with six loading stages ranging from 20 N to 120 N. The influence of soil effective cohesion randomness on pile displacements was evaluated based on statistical data obtained from the numerical simulation of 1000 realizations of random fields. The generated correlated random fields were verified to follow a normal distribution. Meanwhile, the pile displacements at different loading levels tend to follow normal distribution as well, except at the 40 N loading stage, where a deviation from normality was observed. Compared with the homogeneous soil model with a mean value of c', up to 63% of the simulation cases exhibit pile displacement exceeding the allowable limit. Moreover, the effectiveness of the K-means clustering technique in reducing the number of required realizations, and thus the computational workload, was evaluated. The pile displacement results show that when the number of representative clusters reaches 100 or more, the subset data cover more than 98.2% of the full data set, i.e., data from 1000 realizations.]]>
Copyrights © 2026
