Application of Improved Genetic Algorithms in Structural Optimization Design

Abstract

Abstract. The genetic algorithms (GAs) are broadly applicable stochastic search and optimization techniques. However, there exists premature convergence phenomenon in some GAs. To overcome the deficiency, two improved genetic algorithms are proposed in this study. The first one is a hybrid algorithm of the genetic algorithm and downhill simplex method, while the second one is the combination of genetic algorithm and conjugate gradient method. Then, the mathematical optimization model of the 10 bar truss is built and both of the improved algorithms are identified by the numerical example and compared with the simple genetic algorithm. The simulation results indicate that the two purposed techniques show stronger robustness in finding feasible optimum designs than the simple genetic algorithm.

Inverse design of metasurfaces with non-local interactions

“…2e. In this workflow, we first randomly generate a population N p (~tens to hundreds 33 ) of metasurface structures. For each member in the population, the target properties (e.g., phase profile, transmission, and focusing intensity) are evaluated by solving Maxwell's equations (FDTD simulations) for the entire device structure.…”