A critical assessment of methods of correcting for drift from the yield surface in elasto‐plastic finite element analysis

The analysis of elasto-plastic boundary value problems using the finite element method involves many discretizations. These lead to the problem of yield surface drift in which the stress state predicted at the end of an elasto-plastic increment of loading does not lie on the current yield surface. As such discrepancies are comulative it is important to ensure that the stresses are corrected back to the yield surface during each increment of loading. In this paper five methods of accounting for this drift are examined. These involve correcting the stresses by projecting back along the plastic flow, the total strain increment and the accumulated effective stress direction. In addition a ‘correct’, method which accounts for the changes in elastic strains which accompany any stress correction is considered. This method is theoretically more sound than the other approximate approaches. All five methods have been used in finite element analyses of the stress changes that occur adjacent to a single pile installed in a uniform deposit of soil on pile loading. The soil was assumed to be normally consolidated and was modelled using a form of modified Cam Clay. Comparison of these results with an analysis, in which yield surface drift was negligible indicated that only the ‘correct’ method and the method involving projecting back along the plastic flow direction give accurate predictions. Substantiai errors occur if the other methods of correcting for yield surface drift are employed. It is recommended that the ‘correct’ method be adopted for finite element calculations.