A least-squares-based method for a class of nonsmooth minimization problems with applications in plasticity (Q811227)

Problems in mechanics such as limit analysis of beams and plates are studied. The above limit analysis problems are embedded in a wider class of nonsmooth functionals (involving a linear weighted least-squares problem), which contains also some location problems (e.g. the Fermat- Weber problem). An algorithm is proved for minimizing the functionals in the above class, which is based on a combination of smoothing and successive approximation (CSSA). This CSSA algorithm is a modification of that by \textit{W. H. Yang} [Comput. Methods Appl. Mech. Eng. 33, 575-582 (1982; Zbl 0478.73022)]. A convergence analysis of the CSSA algorithm is presented including an excellent performance of the algorithm in the first step, regardless of the starting point. A dual problem to the smoothed primal is derived. It is shown how to obtain a stopping criterion for the algorithm and an estimation of the solution error. Computational results are presented for two limit analyses: a small-scale three bar truss and a large-scale simply supported square plate.