Establishing probabilistic transformation models for geotechnical design parameters using multivariate Gaussian Copula

In geotechnical practice, it is common to transform the measured parameters to the design ones. It is also known that the probabilistic transformation models provide an effective tool for predicting the actual range of the design parameters. The commonly-used methods for establishing probabilistic transformation models based on multivariate normal distribution may induce large errors and have a limitation of incorporating various types of marginal distribution for soil parameters. In this study, a new method for establishing the probabilistic transformation models for geotechnical design parameters based on the multivariate Gaussian Copula is proposed. The global clay database CLAY/6/535 compiled in the literature is employed to verify the effectiveness of the proposed method. The probabilistic transformation models from CPTU indices to undrained shear strength and OCR are then derived. The results indicate that by modeling the marginal distribution and dependence structure individually, the proposed method removes the limitation of incorporating various types of marginal distribution and avoids error propagation from marginal distribution to dependence structure. For the proposed probabilistic transformation models, the uncertainty and correlation of design parameters are inversely proportional to the number of the measured parameters and the correlation between the measured and design parameters.