Research Advances of Chaos Optimization Algorithms for Engineering Global Optimization
LIU Zhen-jun,YANG Di-xiong
DOI: https://doi.org/10.7511/jslx201603001
2016-01-01
Abstract:In recent years ,a new potential tool for global optimization ,namely ,chaos optimization algorithm (COA) ,which is based on nonlinear dynamics characteristics such as the sensitivity on initial value ,pseudo‐randomness ,ergodicity and self‐similar fractals of chaos etc ,has been widely applied in various disciplinary areas of science and technology .According to the development history of COA ,this paper reviews four kinds of major COAs focusing on their basic ideas and the research situations of engineering application :chaotic neural network optimization method ,the first type of hybrid COA (based on chaotic search) ,the second type of hybrid COA (chaotic sequence instead of random sequence ) and chaos and fractal optimization algorithm .Chaotic map was firstly introduced to neural network ,and chaotic neural network optimization method was developed ,which can solve the complex problems of global optimization ,such as combinatorial optimization .The heuristic random algorithms like genetic algorithm and particle swarm algorithm have the capability of global searching ,but they are prone to prematurity and falling into local optima .After the chaotic search concept appeared ,researchers estab‐lished the first type of hybrid COA w hich embeds chaotic search into heuristic algorithm ,and can over‐come effectively the shortcoming of premature convergence of original heuristic algorithm .Subsequently , replacing random parameters of heuristic algorithm with chaotic sequences of chaotic maps ,the second type of hybrid COA is formed .Hybrid chaos optimization algorithms facilitate to achieve rapid global convergence and improve the computational accuracy .Finally ,using the chaotic and fractal characteris‐tics ,the chaos and fractal optimization algorithm ,as a novel approach proposed from the fractal theory ,is proposed and can search for all the global optimum solutions of the optimization problem .In addition , several research directions of COA are presented ,including enhancing parameter design of COA ,handing large scale optimization and multi‐objective optimization problem ,and utilizing the surrogate models and so o n .