Parallel Execution in Metaheuristics for the Problem of Solving Parametric Interval Linear Systems

This paper addresses the problem of solving linear algebraic systems whose elements are nonlinear functions of parameters varying within prescribed intervals. Evolutionary algorithm, differential evolution and two variants of simulated annealing are applied to approximate the hull solution of such systems and compared in terms of accuracy and efficiency. As the computation time for larger problems is significant, calculations for optimisation problems in the family of equations are done in parallel. Structural engineering case studies and numerical experiments involving large uncertainties are carried out to demonstrate accuracy of the metaheuristics and the impact of the parallelization.