Interpreting random field parameters of spatially variable soil using ball penetrometer

Natural soil exhibits inherent spatial variability, which is crucial for site characterization and the reliability design of foundations. However, the interpretation accuracy of random field parameters for spatially variable soil, based on the undrained shear strength (s_u) profile obtained by the penetrometer, is hampered by the spatial averaging effect and weakest path effect. These effects become more pronounced as the observation scale (i.e., the ball diameter in this study) and the coefficient of variation of soil strength increase, respectively. Interpreting the spatial variability of actual soil strength based on s_u profile obtained by penetrometer remains challenging. In the present study, the random field parameters of the s_u profile obtained by the ball penetrometer are estimated using large deformation finite element analysis (LDFEA) and Markov chain Monte Carlo method. A filtering procedure, based on fast Fourier transform, is proposed to filter out the numerical oscillation of LDFEA on the s_u profile. Then the effect of the observation scale on the estimated random field parameters is investigated and it is proved that the estimated random field parameters are significantly influenced by the observation scale. The smaller the penetrometer dimension is, the closer the estimated random field parameters of the s_u profile approach those of actual soil strength. Therefore, a correction procedure is proposed and reliability analysis of a caisson and a circular foundation show that directly employing the estimated random field parameters in foundation design may lead to unsafe outcomes. Therefore, the estimated random field parameters should be corrected by the proposed interpretation procedure.