Hybrid MPI-OpenMP parallel optimization for thermo-fluid problems

This article deals with the minimization of the energy transport through the isolated enclosure by means of obstructing the fluid flow. The differentially heated cavity is partially covered with non-permeable obstacles that affect the natural convection flow. We seek the optimal position and shape of the obstacles to minimize the heat flux through the domain with the minimal cumulative size of the obstacles, i.e. we maximize insulation with minimal material needed. The task is performed by evolutionary multi-objective optimization of a local thermo-fluid solver, which comprises energy transport and incompressible fluid dynamics. An effective parallel algorithm is presented, in which the physical model is solved by a local meshless numerical method on shared memory architecture, coupled by optimization algorithm that exploits the distributed computer parallelism.