Calibration of PET strap pullout models using a statistical approach

Abstract

Polyester (PET) straps are being used more frequently in mechanically stabilized earth (MSE) walls. At present, there is no consensus on a model or models that are suitable for design codes to calculate the pullout capacity of these materials. A database of 296 pullout tests from 81 test series with single and closely-spaced parallel double PET strap configurations was collected by the writers. The data were taken from laboratory and in situ pullout tests. Existing linear and bi-linear pullout models for geosynthetic and steel strip reinforcement were investigated as candidate models and then empirically modified to improve pullout capacity predictions for both single and double PET strap arrangements. A non-linear exponential model with the same number of empirical coefficients was also investigated. The accuracy of each model was assessed quantitatively using bias analysis where bias is the ratio of measured to predicted capacity. The paper shows that non-linear models performed best based on mean and coefficient of variation (COV) of bias values and bias dependency with predicted pullout capacity and magnitude of vertical stress acting on the pullout length. The models investigated in this study are useful for pullout limit state calculations using both deterministic and reliability-based analysis and design approaches.