Improved Particle Swarm Optimization And Optimization Of Mechanical Applications

Particle swarm optimization (PSO) is a new random optimized algorithm based on swarm intelligence. As it is simple and effective, need less parameters, and easier to realize, it attracts more and more attention , and becomes a new research hotspot recently. Particle swarm optimization has been widely used in function optimization, neural network trace editor training, fuzzy control systems, etc. However, its systematic and normalized theory has not been fully developed. It has important practical significance for improved research and extension in application fields.Based on the study on basic PSO, grey self-linkage PSO is presented to optimize high dimension single-mode and multi-mode functions; combined with dynamical cooperative mechanism, an improved algorithm is presented to optimized multi-objective problems. Each improved algorithm is tested by testing functions and engineer examples to demonstrate the effectiveness of the algorithms.Firstly, the principle and model of basic and standard particle swarm optimization are introduced. The convergence and topology structure are analyzed. A new particle swarm optimization based on grey self-linkage analysis is presented. In order to overcome premature of standard particle swarm optimization, the proposed algorithm considered the interrelations between problem dimensions, which performing more frequent simultaneous updates on subsets of particle position components that are strongly linked and using new speed-location update formula. Compared with other improved PSO algorithms, the testing results indicate that the new algorithm has better probability of convergence rate and accuracy for single-mode functions. For multi-mode functions, the present algorithm is also efficient, but the later search performance needs to be improved in the future work.Secondly, for multi-objective problems, in order to get well-distributed Pareto front and more non-inferior solutions, overcome the homoplasy of basic PSO, combined with dynamical cooperative mechanism, a new kind of dynamical cooperative multi-objective particle swarm optimization algorithm is proposed. Using a new kind of criterion judging the stagnation of the population, the sub-population is adaptively added or deleted during the running of the algorithm, which makes the number of the sub-populations vary dynamically. Pareto efficient solutions are saved based on external sets and elites to keep the tactics. This is used to guide the evolution of the whole particle swarm. The test results show that by the exchanges of information among the sub-populations, the whole particle swarm distributes uniformly and avoids local optimum, and the diversity of the solution is ensured.Finally, two mechanical examples show that the improved algorithms in this paper are efficient. Compared with traditional optimized methods, the presented algorithms have applicability and validity.