No label defined (Q4763007)

Many problems of mathematical physics are formulated as minimization problems for an integral functional \(J\) defined on a set \(K\) of a Hilbert space \(H\) of a function of the real argument \(x\in \Omega\), \(\Omega \subset R^n\): \[ \min_{v\in K}J(v), J(v) = \int_\Omega \Phi (v,x)dx. \] It is supposed that the functional \(J(v)\) is continuous and strongly convex in \(H\); \(\Omega\) is a many-sided domain in \(R^n\) with a boundary; \(K\) is a bounded convex set in \(H\). The possibility to combine the penalty function method and the finite-element method for the numerical optimization of the functional \(J(v)\) is considered. An algorithm suggested is based on an iterative procedure for enlargement of the number of nodes of finite elements and of penalty parameters. The conditions obtained for combining the mentioned methods not only guarantees the convergence of calculations proposed but also may be used in practice if these calculations are combined in addition with some concrete methods of the unconstrained optimization.