导读： 中文标题 机械优化设计中几类全域智能算法研究 英文标题 Several Global Intelligence Algorithm Research in Machine Optimization Design 作者 童金旺 专业与研究方向 机械设计及理论 现代设计方法 指导教师 张鹏 申请学位等级 硕士 机构 西南石油大学 论...
中文标题
机械优化设计中几类全域智能算法研究
英文标题
Several Global Intelligence Algorithm Research in Machine Optimization Design
作者
童金旺
专业与研究方向
机械设计及理论 现代设计方法
指导教师
张鹏
申请学位等级
硕士
机构
西南石油大学
论文完成年度
2008
中图分类号
TH11
中文关键词
群体智能；全域优化；微粒群算法；雁群算法；量子微粒群；混沌微粒群；参数优化；钻头侧向力优化
英文关键词
Swarm Intelligent; Global Optimization; Particle Swam Optimization; Geese Swarm Optimization; Quantum Particle Swarm Optimization; Chaos optimization algorithm;
中文摘要
随着人类生存与活动空间的扩大以及认识与改造世界范围的拓宽，人类需要对客观世界的规律有更全面更深入的理解，传统的优化方法在处理复杂优化问题时，如高维、多极值、复杂形态函数时，在求解的精度和求解所需时间上，往往很难满足实际工程的需要。20世纪80年代，随着蚁群算法、微粒群算法等群体智能优化算法的相继出现，这些矛盾得到了一定程度上的缓和。群体智能优化算法的概念源于对蜜蜂、蚂蚁、大雁等群居生物群体行为的观察和研究，通过模拟群居生物的觅食、迁徙过程而形成的一种迭代优化方法。微粒群优化算法(Particle swarm optimization, PSO算法)是Kennedy和Eberhart于1995年借鉴鸟群觅食行为提出的，其概念简单，实现容易，收敛速度较快，具有较强的全局搜索能力，在近几年得到了快速的发展，并在一些领域获得了成功应用。但作为一种新兴的智能优化算法，同时也存在一些诸如易陷入局部极值等缺点。本文重点研究了微粒群算法的参数特性，并对微粒群算法提出了一些改进措施，主要研究成果如下：
(1) 简要地回顾了全域群体智能优化算法产生的理论背景和思想源泉，以及与传统优化算法相比的优缺点。总结了三种群体智能优化算法——蚁群算法、文化算法和人工鱼群算法的理论基础、算法概念和迭代过程。
(2) 阐述了微粒群算法的产生、成长过程。研究微粒群算法的收敛性条件，导出了影响微粒群算法收敛参数的 和 的关系式，并给出了收敛范围。讨论了参数之间的相互关系以及怎样影响算法效率的，并从统计学的角度，分别对惯性因子 ，最大速度 ，加速常数 、 进行了单因子方差实验分析，探讨了不同参数设置水平与算法性能之间的基本联系及其怎样影响函数最优值计算结果，结合具体的实验函数，给出了这些参数的合理取值的范围。
(3) 由于基本微粒群算法具有容易早熟收敛、局部寻优能力较差、收敛速度比较慢等缺点，本文在分析微粒群算法各参数如何影响算法性能的基础上，对各参数提出了一系列的改进措施，并与其它优化算法想结合，形成了混合优化算法，重点研究了混沌微粒群优化算法和量子微粒群优化算法。通过实验证明，这些改进措施在一定程度上提高了微粒群算法的优化性能。
(4) 建立PDC钻头的受力模型，将微粒群算法应用于PDC钻头侧向力的优化，优化后的刀翼式PDC钻头和散布式PDC钻头侧向力均有明显的减小，改善了钻头因涡动而产生损坏。
(5) 针对微粒群算法在求解函数优化问题时出现的一系列问题，本文提出了雁群算法，详细介绍了雁群算法的提出的思想根源和背景知识，给出了算法的迭代公式和迭代步骤，并结合具体实例与微粒群算法相比较，对于某些优化问题，其优化性能比微粒群算法有较大的提高，在一定程度上弥补了微粒群算法的不足。
英文摘要
Along with the human survival and activity space's expansion, as well as understanding and transformation of the worldwide scale's broadening, mankind needs a more comprehensive understanding to the objective laws of the world. The traditional optimization methods in dealing with the complex optimization problems, such as high-dimensional, multi-polar value, complex function, and the accuracy of solution and the time required in solution, is very often difficult to satisfy the needs of practical project. In the 1980s, along with the community intelligence optimization algorithm's appearances, such as Ant Colony algorithm, Particle Swarm Optimization algorithm, and so on, relaxing these Contradictions in certain degree. Intelligent optimization algorithm's concept stems from the observation and research of the groups of the organisms, such as bees, ants, geese, and so on. They are the formation of iterative optimization method which through simulate the groups of organisms feeds and migrations.Particle Swarm Optimization algorithm (PSO algorithm) was learn from the behavior of birds look for food by Kennedy and Eberhart in 1995, which has simple concept, easy to realize, faster convergence speed and has strong overall search capabilities. In the past few years, it was obtained fast development, and success application in some areas, but as a new intelligent optimization algorithm, also exist some shortcoming such as easily to fall into local extreme and so on. This article focuses on the parameters of particle swarm algorithm characteristics, and proposed some improvement for standard PSO algorithm, the main research results are as follows:
(1) This article briefly reviewed theoretical background and ideological source of the entire domain group intelligent optimization algorithms, and compared with the traditional optimization algorithm. This article summarized three community intelligent optimization algorithms’ theoretical foundation, the concept of algorithm and iterative process: Ant Colony algorithm, Cultural algorithm and Artificial Fish algorithm.
(2) This article described PSO algorithm’s the process of appearance and growing, and research PSO algorithm convergence of conditions, derived the relationship between Parameter and which influence convergence of PSO algorithm, and give the scope of the convergence. Discussed relationship among all parameters and these parameters how mutual influence convergence of PSO algorithm. From the view of statistics, this article analysis the inertia factor , maximum speed , acceleration constants and , which carried on the single-factor variance experiment. Discussed relationship between different parameters levels and algorithm performance, and how influence the result of optimal value of the results. At last, this article gives the scope of these parameters reasonable value by the experiment function.
(3) As standard Particle Swarm algorithm has a lot of shortcomings such as easy to premature convergence, limited capacity in local optimization, and slowly convergence speed, this article put forward a series of improvement based on the analysis of the PSO algorithm parameters how influence the performance of algorithms, and combined with other optimization algorithm to form a hybrid optimization algorithm. This article focused on the chaotic Particle Swarm Optimization algorithm and Quantum Particle Swarm Optimization algorithm. Experiments showed that these improvements the optimized performance of PSO algorithm to a certain extent.
(4) Establishes the PDC bit's stress model, applies the particle group algorithm in the PDC bit lateral force optimization, after the optimized blade PDC bit and the dispersion PDC bit lateral force has obvious reduction, improved the bit to have the damage because of the whirling motion.
(5) In view of use the PSO algorithm in solving function optimization questions appear a series of questions problems, this article proposed the Geese Swarm Optimization algorithm. Introduce the ideological root causes and background knowledge in detailed, give the iterative formula and the iteration step. Compared with PSO algorithm in some optimization experiment function, its performance better than Particle Swarm Optimization algorithm in some area, supplement the PSO algorithm deficiencies to a certain extent.
目录/全文下载
作者要求对全文保密至2013年7月，当前无法提供下载！
