An accelerated augmented Lagrangian algorithm with adaptive orthogonalization strategy for bound and equality constrained quadratic programming and its application to large-scale contact problems of elasticity

Abstract

Augmented Lagrangian method is a well established tool for the solution of optimization problems with equality constraints. If combined with effective algorithms for the solution of bound constrained quadratic programming problems, it can solve efficiently very large problems with bound and linear equality constraints. The point of this paper is to show that the performance of the algorithm can be essentially improved by enhancing the information on the free set of current iterates into the reorthogonalization of equality constraints. The improvement is demonstrated on the numerical solution of a large problem arising from the application of domain decomposition methods to the solution of discretized elliptic variational inequality describing a variant of Hertz's two-body contact problem.