Using a reduced number of Lagrange multipliers for assembling parallel incomplete field finite element approximations (Q1205648)

The paper deals with the domain decomposition technique. In the standard approach, broadly described in the first section, the spatial domain is partitioned into a set of totally disconnected subdomains. Then the incomplete solution is computed in each subdomain. Finally Lagrange multipliers equal to the number of degrees of freedom located in the binding interface are introduced to enforce compatibility constraints. The authors propose to use a much lower number of Lagrange multipliers. Lagrange multipliers physically represent the interaction forces between subdomains. The authors propose to replace them with a new set of multipliers, not connected to a specific node of interface. They do not enforce the continuity constraint equation at each node of the interface. The average value of the solutions obtained for both subdomains is assumed as the interface restriction. Then real solutions can be computed for a whole discrete structure. Low order polynomial approximation of the interface traction is discussed in the next section. It is also shown that the solution can be improved by the iterative refinement procedure. A wide range of numerical examples proves the advantage of the method with reduced number of Lagrange multipliers. Particularly it can be used for parallel processing. Moreover, the technique can also be interesting for parallel or sequential solution in the case when each subdomain is solved by different solution methods. The paper is written clearly, with all the details required for implementation. The method can be adopted to various problems treated by numerical methods. The paper is directed to engineers, practitioner and research students.