The loss of soil strength in undrained behavior is one of the main causes of failure in geotechnical engineering projects. The most widely used constitutive model worldwide is the linear-elastic-perfectly plastic Mohr-Coulomb model, which is unable to adequately represent undrained behavior, leading to non-conservative designs or parameter configurations not recommended for simulating strength loss.
To address this problem, the parameter updating methodology proposed by Corral (2013), based on the genetic algorithm, is used. Three constitutive models are compared: Mohr-Coulomb, Hardening Soil small-strain stiffness, and NorSand, available in the commercial software PLAXIS. The methodology for parameter optimization using undrained triaxial compression tests and considering the maximum allowable measurement errors according to ASTM D4767. The advantage of this methodology over traditional parameter adjustment is demonstrated.
Additionally, the different constitutive models are compared, and the superiority of advanced models over the linear-elastic-perfectly plastic model is evidenced, achieving improvements ranging from 50 % to 98 % in some undrained triaxial tests. It is observed how, through advanced models and this methodology, it is possible to represent the stress path and simulate the strength loss that can cause brittle failures in geotechnical projects.
