A Fitness Granulation Approach for Large-Scale Structural Design Optimization

The complexity of large-scale mechanical optimization problems is partially due to the presence of high-dimensional design variables, the nature of the design variables, and the high computational cost of the finite element simulations needed to evaluate the fitness of candidate solutions. Evolutionary algorithms are ruled by competitive games of survival and not merely by absolute measures of fitness. They can also exploit the robustness of evolution against uncertainties in the fitness function evaluations. This chapter takes up the complexity challenge of mechanical optimization problems by proposing a new fitness granulation approach that attempts to cope with several difficulties of fitness approximation methods that have been reported in the specialized literature. The approach is based on adaptive fuzzy fitness granulation having as its main aim to strike a balance between the accuracy and the utility of the computations. The adaptation algorithm adjusts the number and size of the granules according to the perceived performance and level of convergence attained. Experimental results show that the proposed method accelerates the convergence towards solutions when compared to the performance of other, more popular approaches. This suggests its applicability to other complex finite element-based engineering design problems.