TEST REPEATABILITY AND REPRODUCIBILITY OF SHEAR STRENGTH PARAMETERS
Volume Title: ICASGE2025
Paper ID : 1145-ICASGE-FULL
Authors
Abstract
Test repeatability and data reproducibility are critical concerns in geotechnical laboratory testing due to inherent soil variability, inhomogeneous sample preparation, and measurement inaccuracies. Triaxial test results on comparable specimens can vary significantly, influencing the derived shear strength parameters. This study investigates the repeatability of consolidated undrained triaxial tests and the reproducibility of effective shear strength parameters for both untreated and cement-treated soil specimens. Thirty remoulded test specimens were prepared with five different soil-cement ratios, with six specimens prepared for each ratio. The tests involved three levels of consolidation pressure, with two tests conducted per consolidation pressure on specimens prepared in the same way. A key focus was on achieving same density during specimen preparation for each pair. Cement-treated specimens were cured for 28 days, during which their densities changed. Shearing was performed at a deformation rate of 0.4 mm/min after sample saturation and consolidation. Effective peak and residual shear strength parameters were estimated through regression analysis of 21 different combinations of failure stresses from the six tests conducted at each soil-cement ratio. These combinations included three, four, five, and six failure stresses selected from each test series. The scattering of shear strength parameters was analyzed using statistical procedures, including measures of central tendency, dispersion, and confidence limits. Additionally, unsupervised clustering and anomaly detection were conducted using the K-Means machine learning algorithm to assess the overlap of shear strength parameters across the different soil-cement ratios. The effective shear strength parameters derived from regression analysis of different failure stress combinations exhibit variability.
Keywords
Subjects