In springback analysis, I just changed the constrained nodes a little bit away from the original node. The convergence then became very different and the predicted result seems not to be reasonable. What can be the problem?

Usually, when change the constrained node properly, the springback prediction should be unaffected, or the effect should be very small.

However, when the newly constrained node is the slave node from element adaptivity, then the constraint cannot be taken into account properly. Accordingly, the convergence become worse and the springback result, if obtained, will become very bad.

To avoid this problem, it is always necessary to avoid constraining the slave nodes in springback analysis.