SQP methods for incremental plasticity
Prof. Dr. Christian WienersDec. 5, 2006, 3:30 p.m. HF 136
The standard procedure in computational plasticity reformulates the incremental step into a minimization problem or an equivalent nonlinear variational problem, where the nonlinearity results from the projection onto the set of admissible stresses. Numerically, the incremental problem is solved by a semismooth Newton, where the consistent tangent is chosen from the multi-valued derivative of the projection.
This standard procedure is compared with an realization of the SQP method, where the Newton method is replace by a sequence of quadratic minimization problems (which are solved approximately by a small number of semi-smooth Newton steps). We show that is optimization approach is more robust and more efficient in difficult cases, e. g., near to the limit load in perfect plasticity.