Implementation and testing of parallel derivative-free global algorithms with applications to robust optimization problems for ship design

Our work aims to the implementation and numerical evaluation of multithread methods within the context of global optimization algorithms for complex problems. In particular we develop asynchronous parallel derivative-free algorithms using both MPI and OpenMP paradigms. Evaluation and calibration of these algorithms is performed by using the HPC cluster Matrix of the "Consorzio interuniversitario per le Applicazioni di Supercalcolo Per Universit`a e Ricerca" (CASPUR). Numerical results are presented for applications in robust design optimization of high speed catamarans