Research Of Hybrid Optimization Algorithms Based On Swarm Intelligence

During the past decade, swarm intelligence methods have been extensively applied in engineering, due to their remarkable features, e.g., stochastic search, parallel processing, and distributed & decentralized control. However, single swarm intelligence algorithms are usually weak at dealing with practical multi-modal, high-dimension, multi-objective, and constrained optimization problems. Therefore, study of hybrid swarm optimization strategies for targeted problems is indeed of great theoretical significance and application value. This dissertation proposes and develops several hybrid optimization methods on the basis of two typical swarm intelligence algorithms, particle swarm optimization and bacterial foraging optimization. The main contributions of the dissertation are explained as follows.Firstly, the particle swarm optimization is integrated into the clonal selection algorithm with an enhanced operator, in which the antibody suppression operators are based on the basic principles of artificial immune network. The cloning operator and super mutation operator are also redesigned for handling the multi-modal function optimization, guide ball optimization, and self-tuning PID controller parameter optimization problems.Secondly, in order to overcome the premature disadvantage of the original particle swarm optimization in high-dimension function optimization problems, a new dynamic multi-swarm optimization algorithm combining the clonal selection theory and particle swarm optimization together is proposed. This algorithm considers the sub-groups of swarm as the antibodies, and the particle swarm optimization is utilized in these sub-groups. The novel cloning, mutation, selection, and receptor editing operators are developed from the clonal selection principle. To verify the effectiveness of the proposed optimization method, seven high-dimension functions are used as the test-beds in our simulations.Thirdly, an adaptive bacterial foraging algorithm with the good nodes set-based crossover operator is presented so as to solve the constrained optimization problems, which is a fusion of the bacterial foraging optimization and genetic algorithms. The good nodes set are utilized here for initializing the bacteria population, constructing the crossover operator, and dispersing the similar individuals. A novel adaptive computational chemotaxis is incorporated into our optimization scheme, and a hybrid selection approach based on the Pareto-dominance and tournament selection is studied, which can effectively retain the best individuals in the population. We examine the hybrid algorithm using 11 well- known benchmark functions, and compare its performances with other constrained optimization methods. Three interesting engineering design examples are used to explore the efficiency of this adaptive bacterial foraging algorithm as well.Finally, the performance of the bacterial foraging optimization for attacking multi-objective optimization problems is analyzed. A hybrid multi-objective optimization algorithm is proposed by merging the cultural algorithm with the bacterial foraging method. This technique regards the bacterial foraging optimization as a search engine, and deploys the belief space, acceptance function, and influence function for manipulating with the multi-objective optimization problems. The knowledge of the belief space is applied to adaptively adjust the chemotaxis step, and the individual health factors in the bacterial foraging method are also introduced. A total of five typical multi-objective optimization problems are employed to investigate the behaviors of the proposed algorithm in our numerical simulations.