Evaluation of Soil-Geogrid Pullout Models Using a Statistical Approach

Abstract

Abstract The accuracy of the current Federal Highway Administration (FHWA) in-soil geogrid pullout model was examined using a statistical approach applied to a large database of pullout test results. The accuracy of data interpretation and model type was quantified using the mean and coefficient of variation (COV) of model bias values and possible hidden dependencies identified using the Spearman rank correlation coefficient. Model bias values were computed as the ratio of measured to predicted pullout capacity. When project-specific pullout test data were used to fit a linear approximation to dimensionless interaction coefficients, the result was judged to be an acceptably accurate model (mean bias value of one and a small spread in bias values, i.e., COV=0.13). However, in many cases project-specific pullout data are not available. If the current FHWA model with default values is used, the prediction accuracy is very poor based on the same quantitative measures (mean of bias=2.23 and COV=0.55). Two new models were examined to overcome this deficiency. One model is bi-linear and the other is non-linear. The non-linear model was shown to be more accurate with a mean bias value close to one and COV=0.36. The non-linear model also has the advantage of being smoothly continuous with practically no detectable hidden dependencies. Finally, the large number of test results in the database allows recommendations to be made on how to select reinforcement lengths during the experimental design to increase the likelihood of a pullout mode of failure in the laboratory.