Assessment of differential evolution for multi-objective optimization in a natural convection problem solved by a local meshless method

Abstract

ABSTRACTThe performance of Differential Evolution for Multi-objective Optimization (DEMO) in a nonlinear coupled transport problem, solved by a Meshless Local Strong form Method (MLSM), is assessed from different points of view. First, the behaviour of the optimization algorithm is tested for different scenarios, ranging from optimization of trivial diffusive transport to more complex nonlinear natural convection problems. Second, a hybrid parallel implementation of both the optimization and simulation codes, is introduced to optimize execution time, since such simulation-based optimization might require a vast amount of computational power. The goal of optimization is partially to cover the differentially heated cavity with non-permeable obstacles so as maximally to obstruct the flow with minimal possible coverage. Different scenarios are taken into account to analyse the optimization performance. The results are presented in terms of temperature contour plots, velocity profiles, analysis of heat losses, Pareto fronts of optimal solutions, convergence of optimal solutions, and sensitivity analysis of the optimizer and parallel execution performance.