Statistical prediction of deformations of soil nail walls.

Statistical characterization of model uncertainty has been exclusively focused on reliability-based design of soil nail walls against ultimate limit states; little has been accomplished for predictions of horizontal and vertical displacements at the top of the wall for analysis of serviceability limit states. To fill the gap, this study first develops a database containing 461 measured wall horizontal and vertical displacements from 75 fully instrumented soil nail walls under working conditions reported in the literature. Then, the database is used to evaluate the accuracies of simplified wall displacement models currently adopted in four national soil nail wall design manuals. The results show that on average the current models tend to over-predict displacements for walls in sandy soils, but under-predict displacements for walls in clayey soils. Predictions using the current models are moderately to highly dispersive. Simple models are developed and calibrated for prediction of displacements based on the established database. Predictions using the developed models are demonstrated to be accurate on average and significantly less dispersive. In addition, a simple model for prediction of horizontal wall displacement along depth is developed and optimized. The present work is practically valuable for reliability-based design of soil nail walls against serviceability limit states. [ABSTRACT FROM AUTHOR]