混合粒子群优化及其在嵌入式智能控制中的应用-机械设计及理论专业论文.docx

武汉理工大学博士学位论文Abstract 武汉理工大学博士学位论文 Abstract In this paper,Some hybrid intelligent computing method including Neural Network(NN)，Fuzzy Logic(FL)，Simulated AnneaIIng(SA)etc，as well as its application have been researched systematically in both theory and experiment based on particle swarm optimization algorithm． The thesis researched deeply particle swarm optimization algorithm．弧e concept of aggregation degree of particle swarm has been proposed．Tbc pmticles diversity was improved through periodically monitoring aggregation degree of the particle swarm．On the later development of the PSO algofithm,adopting mutation strategy to the best particle’s position enhances its ability ofjumping out the local minima．Through furthc!l-analysis tO the improved algorithm,the thesis proposed缸 flgofithm of mutation to individual best pofifio玛which can increase activity of particle in a grealcr degree and蜘k蝴the optimization pedbrm雒ce．Combining SA optimization technique，the hybrid PSO algorithm embedded the improved SA was proposed．11硷probability-sudden-jump chafagt盯of SA can avoid getting into local minima of PSO algofithm effectively．Meanwhile，applying parallel PSO and adopting improved SA optimization mechanism,the dissatisfactory time character of the SA algorithm could be made better．The computer simulating indicated the algorithm is锄effective optimized method．Moreover,the research showed the solution form of random opfJmizing search algoritlun flong with the neighbourhood strugtl】fc would take a great effect on the speed and precision of algorithm convergence· In the research of forwardfeed nculm network(FNN)based on PSO algofithnl， the FNN learning algodthra based 011 the improved PSO aJgodthm(PSOFh呐has been proposed,in which the derivative information is not requisite．PSOFNN has overcome the defection of the缸aditional learning algorithm easily to fall into the local minima and has a good property ofinitial robustness．In the research oflcarning algorithm 011 FNN’s variable翻玎lcn聪