粒子群算法改进及其在旋风分离器结构优化中的应用

发布时间：2018-03-18 17:54

本文选题：粒子群算法　切入点：拓扑结构　出处：《太原理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：基于启发式的群智能随机演化计算,Kenndey和Eberhart通过模拟自然界中鸟群和鱼群在捕食过程中的群体协作与竞争行为,于1995年提出了粒子群优化算法。相比于其他群智能算法,粒子群优化算法在解决多目标优化、动态寻优等问题上,具有结构简单、易编程实现等特点,经过二十多年的不断发展,逐渐形成了一套完整的理论研究体系,已经成为国际进化计算领域的重要研究方向。粒子群优化算法一经提出,便受到广泛的关注和应用,但到了进化后期算法存在着种群多样性丧失、易陷入局部极值等问题,导致后期收敛速度减缓,优化精度不足。为了在提高寻优精度的同时,加快算法的收敛速度,本文从种群拓扑结构和进化学习机制两方面分别对算法进行了改进,并将改进算法应用于旋风分离器的结构参数优化。本文研究的主要内容如下:1、为增强粒子群算法种群内不同个体间的信息交流能力,考虑从种群拓扑结构入手,提出了一种基于混沌拓扑结构的全信息变异粒子群优化算法(CFMPSO)。该算法在进化过程中周期性地对种群的拓扑结构进行混沌重组,并对各粒子邻域中的最优个体进行变异,通过全信息策略对不同变异粒子信息进行充分利用,以增强其交流能力,改善了粒子群算法的收敛性能,加快了进化速度。实验结果表明,CFMPSO算法在解决大多数测试函数时能获得较好的寻优精度,并且其函数计算次数少、寻优速度快。2、通过深入研究算法本身的特性,对种群的寻优机制进行数值分析,推导出标准粒子群算法近似于一种比例-积分(PI)控制,由于其固有积分属性的存在,使算法收敛速度减慢。通过添加微分项,提出微分策略的快速粒子群优化算法,利用该策略来加快标准粒子群及其改进算法的收敛速度,提高优化效率。对D-SPSO和D-FIPS进行仿真实验对比,结果表明微分控制策略改进算法的函数计算次数少、寻优速度快。证明引入该策略的改进粒子群优化算法在解决进化过程中出现的寻优速度慢和优化效率低的问题上,取得了良好的效果。3、本课题组将改进粒子群算法用于旋风分离器的结构尺寸优化,以旋风分离器满足较小压力损失(35)p和较大分离效率η作为优化目标,采用四因素三水平试验条件,通过Box-Behnken设计试验,得到其结构尺寸的回归方程,并利用改进粒子群算法对结构回归方程进行优化。经过对各算法优化结果分析,表明各改进算法在得到较为理想的优化结果同时,又具有较快的优化速度。
[Abstract]:Based on heuristic stochastic evolution algorithm of swarm intelligence, Kenndey and Eberhart proposed a particle swarm optimization algorithm in 1995 by simulating the cooperative and competitive behaviors of birds and fish during predation in nature, compared with other swarm intelligence algorithms. Particle swarm optimization (PSO) has the advantages of simple structure and easy programming to solve the problems of multi-objective optimization and dynamic optimization. After more than 20 years of continuous development, a complete theoretical research system has been gradually formed. Particle swarm optimization (PSO) has been widely concerned and applied since it was put forward, but at the late stage of evolution, it has some problems such as loss of population diversity, easy to fall into local extremum and so on. In order to improve the accuracy of optimization and improve the convergence speed of the algorithm, this paper improves the algorithm from two aspects: population topology and evolutionary learning mechanism. And the improved algorithm is applied to optimize the structure parameters of cyclone separator. The main contents of this paper are as follows: 1. In order to enhance the ability of information exchange among different individuals in the population of particle swarm optimization, the topological structure of the population is considered. In this paper, a full information mutation particle swarm optimization algorithm based on chaotic topology is proposed. The algorithm periodically recombines the topological structure of the population in the evolution process, and mutates the optimal individual in the neighborhood of each particle. In order to enhance its communication ability and improve the convergence performance of particle swarm optimization algorithm, the full information strategy is used to make full use of different mutation particle information. The experimental results show that the CFMPSO algorithm can obtain better optimization accuracy in solving most of the test functions, and its function calculation times are less, and the optimization speed is higher. 2. Through the in-depth study of the characteristics of the algorithm itself, The optimization mechanism of population is analyzed numerically, and the standard particle swarm optimization algorithm is deduced, which is similar to a kind of scale-integral PI-based control. Due to the existence of its inherent integral property, the convergence speed of the algorithm is slowed down. By adding differential terms, the convergence rate of the algorithm is reduced. A fast particle swarm optimization algorithm with differential strategy is proposed, which is used to accelerate the convergence speed of standard particle swarm optimization and its improved algorithm, and to improve the optimization efficiency. The simulation results of D-SPSO and D-FIPS are compared. The results show that the improved algorithm of differential control strategy has less function calculation times and faster searching speed. It is proved that the improved particle swarm optimization algorithm introduced this strategy can solve the problems of slow optimization speed and low optimization efficiency in the course of evolution. The improved particle swarm optimization (PSO) algorithm is applied to the optimization of the structure size of the cyclone separator. The cyclone separator satisfies the lower pressure loss of 35p and the greater separation efficiency 畏 as the optimization objective. Using four factors and three levels of test conditions, the regression equation of structure size is obtained by Box-Behnken design test, and the structure regression equation is optimized by using improved particle swarm optimization algorithm. The results show that the improved algorithms can get better optimization results and faster optimization speed at the same time.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TQ051.8;TP18

【参考文献】