María José Toledo Arcic, Jens Engel

Abstract: This study presents a predictive framework for estimating peak shear strength and the corresponding shear displacement in direct shear tests, with a particular focus on applications to multistage testing. A normalized hyperbolic function, originally developed for triaxial tests, is adapted to represent the shear stress–displacement curve up to failure. Based on a dataset of 484 direct shear tests performed on 175 different soils, the parameters of the model were derived through regression and empirically linked to the normalized secant elastic modulus. In multistage direct shear tests, early termination of the initial shearing phases often prevents the direct measurement of peak values. To address this, a prediction algorithm was developed that estimates the unknown peak shear strength and displacement based on the initial portion of the shear curve. This algorithm combines empirical relationships with a stochastic search method based on differential evolution to minimize the prediction error. The model was validated across the full dataset, and simulations showed that peak values could be predicted with high accuracy even when only 60% of the displacement at failure was used as input. The results highlight the potential of this approach to improve the reliability and efficiency of multistage shear testing in fine-grained, coarse-grained, and mixed soils.

Keywords: Direct shear tests, shear strength prediction, multistage tests, Kondner model, stochastic optimization.

Date Published: April 30, 2025
DOI: 10.11159/ijci.2025.004

View Article