Characterization of model accuracy in prediction of soil nail loads using simplified incremental calculation method

Although the Rankine-based simplified incremental calculation method(SICM) has been widely applied in practice for prediction of nail loads, its model accuracy has not yet fully assessed. In this paper, a large database containing a total of 466 measured nail loads from 178 nails reported in the literature was established. After removing questionable data, the remaining load data from 143 nails were used to evaluate the accuracy of the Rankine SICM, which is characterized by the model bias defined as the ratio of measured to predicted nail loads, and the influence of 3 working conditions on the accuracy of the SICM was also investigated. The results show that, under typical cases, the current Rankine SICM underestimates averagely the maximum nail load by about 10% to 17% with a high prediction dispersion over 90% while overestimates the mean nail load by about 15% with a moderate prediction dispersion of about 70%. In addition, the accuracy is statistically correlated to the magnitude of the predicted value, which is undesirable. It was proposed to estimate the nail load based on the Coulomb active earth pressure theory instead of the Rankine theory for accuracy improvement. A calibration constant of about 0.75 is introduced to calibrate the computed nail load. The Coulomb SICM was demonstrated to be unbiased on average with a medium prediction dispersion from 44% to 66%. It is also shown that the accuracies of both Rankine and Coulomb SICMs are significantly influenced by soil types whereas the influence of load allocation schemes, external loading conditions and wall types is marginal. The model biases of Rankine and Coulomb SICMs respectively follow lognormal and Weibull distributions. Finally, the accuracies of the two types of SICM were discussed complementally from another four angles. © 2021, Science Press. All right reserved.