ABSTRACT
The numerical performance of various load stepping schemes is compared in terms of accuracy, robustness and efficiency. The schemes studied include classical incremental and iterative methods as well as advanced automatic methods developed more recently. In all iterative methods, two different norms for the unbalanced forces are used to compare their effects on the convergence rate. Analyses are carried out for both Mohr-Coulomb and critical state models. The numerical results indicate that the accuracy of incremental schemes and the efficiency and robustness of iterative schemes are strongly influenced by the load increment size. The automatic schemes that choose the size of load steps according to local truncation errors in the computed displacements are found to be the most accurate, robust and efficient. The Euclidean norm appears to be the better unbalanced force norm for terminating iterative schemes.
