基于填充粒子群算法的双次级永磁同步直线电机优化设计
安徽大学电气工程与自动化学院 合肥 230601
Based on Filled Particle Swarm Optimization Algorithm of the Double-Secondary Permanent Magnent Synchronous Linear Motor Optimization Design
Anhui University Hefei 230601 China
责任编辑: 崔文静
收稿日期: 2015-07-24 网络出版日期: 2015-08-25
基金资助: |
|
Received: 2015-07-24 Online: 2015-08-25
作者简介 About authors

宋俊材 男 1992年生,硕士,研究方向为直线电机的本体优化设计。
提出了一种填充粒子群算法(FPSO),用以解决双次级永磁同步直线电机优化设计问题。在有限元分析的基础上,采用支持向量机拟合直线电机结构参数与运行性能参数之间的关系,建立用于优化计算的非参数模型;引入填充函数,对传统粒子群算法进行改进,并采用多峰值函数对算法进行测试,结果表明:FPSO具有良好的快速性和全局收敛性;采用FPSO对电机结构参数进行优化,得到一组最优的电机结构参数。仿真实验表明:采用该算法优化后的电机推力大、推力波动小且峰值电流小,符合电机的优化设计目标。
关键词:
This article proposes a filled particle swarm optimization(FPSO) algorithm to study the double-secondary permanent magnent synchronous linear motor optimization design. This article uses support vector machine to build the model of structure parameters and output performances on the foundation of FEM analysis. The technology of filled functions was introduced to improve the traditional particle swarm optimization algorithm, the test function proves that the FPSO can be more frequent and accurate in global optimization. Finally FPSO is applied to optimize the linear motor model. The simulation experiment shows that the optimized PMLSM has an increased thrust, and smaller thrust ripple, low peak current, the motor can reach the goal of optimization design.
Keywords:
本文引用格式
宋俊材, 董菲, 赵吉文, 李乐, 苏云升.
Song Juncai.
1 引言
本文从电机本体结构出发,对电机的结构参数进行优化设计,提高电机的运行性能。将影响电机性能的结构参数——气隙、极距、线圈长度、永磁体长度和永磁体宽度等参数作为输入,电机的推力、推力波动率和峰值电流作为输出,基于正交实验和随机试验设计,通过有限元仿真获得一定数量的样本数据,使用支持向量机,对电机结构参数与运行性能之间的映射关系进行非线性拟合,建立用于优化计算的非参数模型,最后使用FPSO对电机结构参数进行优化,在提高电机推力和运行效率的同时减小了推力波动,改善了电机的运行性能。
2 双次级永磁同步直线电机建模
2.1 双次级永磁同步直线电机有限元建模
本文研究的是双次级永磁同步直线电机,如图1所示。定子材料是电工纯铁,磁钢材料是汝铁硼N42,A、B、C三相采用Y联结,极对数是32,定子槽数是12。
图1
图2
表1 双次级永磁同步直线电机结构参数(单位:mm)
Tab.1
参 数 | 数 值 |
---|---|
永磁体长度l | 36.15 |
永磁体宽度h | 12.20 |
气隙δ | 1.60 |
极距τ | 18.35 |
线圈长度S | 7.06 |
表2 样本数据
Tab.2
电机结构参数 | 输出性能参数 | ||||||
---|---|---|---|---|---|---|---|
l | h | δ | τ | S | η(%) | F | I |
36 | 13 | 1.85 | 18.8 | 6.8 | 1.207 | 35.2 | 3.574 |
36 | 14 | 1.95 | 18.9 | 6.8 | 0.834 | 36.7 | 3.684 |
37 | 16 | 2.15 | 19.1 | 7.0 | 0.956 | 37.8 | 3.412 |
38 | 15 | 2.00 | 19.0 | 7.1 | 0.672 | 36.2 | 3.689 |
… | … | … | … | … | … | … | … |
… | … | … | … | … | … | … | … |
46 | 18 | 2.1 | 19.3 | 7.3 | 2.036 | 37.5 | 3.889 |
2.3 支持向量机建立快速计算模型
支持向量机(SVM)是一种以有限样本统计学习理论为基础的学习方法,可以有效解决小样本、高维数、非线性和局部最优等问题[4]。本文使用SVM对有限元分析产生的样本数据建立快速计算模型,其实质是探寻一个实值函数f (x),以求解任一输入x所对应的输出值y。引入遗传算法对SVM的关键参数——惩罚系数c和核参数g进行寻优[5],得到c = 1.805 9,g = 6.458 8。将有限元分析产生的200组数据分为两组,一组用于支持向量机训练,建立SVM模型,另一组用于模型的检验,通过对比原始输出数据和回归预测数据,计算模型误差率,检测模型的预测能力。本文使用推力波动率预测精度对所建立的SVM模型进行检验[6]。推力波动率模型预测精度图如图3所示。
图3
通过图3可知,原始数据与回归数据之间有较好的拟合性,该模型具有高达91%的精准度,可为后续算法对电机结构优化的逆向计算提供较好的数学支持。
3 填充粒子群算法
3.1 多目标优化及分解问题
多目标优化问题的数学形式

式中,x为n维决策变量;Ωn为决策变量的可行解空间;f (x)为m个目标函数。采用文献[7]提出的Tchebycheff分解方法,将多目标优化问题转化为多个单目标优化问题进行求解,为了使多个目标同时达到最优,FPSO寻找的是Pareto最优解。
3.2 填充粒子群算法
在填充粒子群算法中,首先使用填充函数对目标函数进行填充处理,减少目标函数极值个数;其次对粒子进行初始化,更新粒子当前位置,确定粒子个体最优位置;最终经过迭代确定群体最优位置,从而得到最优解。算法介绍如下。
3.2.1 填充函数
简单的填充方法是对目标函数f (x)进行空间变换,将函数极值区域进行填充平滑处理,提高算法的全局收敛性,填充方法示意如图4所示。
图4
对函数f (x)进行填充处理后可以得到新的目标函数

式中,λ为常量,且λ≥ε≥0,λ用于对寻优空间进行分割,小于高度λ的区域,用高度ε进行压缩和填充,大于λ的区域则维持原空间高度不变。相比于f (x),F(x)最值区域的相对高度增加,极值区域明显减少[8]。
简单的填充方法对f (x)进行填充是可行的,但在针对多峰值复杂函数时,该方法不能快速准确地对多个相距较远的极值区域进行填充。因此引入填充函数进行改进,其思想是构造辅助函数,使用优化程序跳出局部最优范围,从当前局部极值点跳到另一个局部极值点,不断重复该过程,直到达到最值时终止。根据文献[9]提出的一种高效填充函数的定义,本文采用了连续单参数的填充函数

式中,x*为f (x)的当前局部极小点;q>0,即填充因子,q充分小且选取合适的时候,填充函数的效果更好;v(t)和u(t)是连续可微的函数,其中函数v(t)可取- t,- t2,arctant,1- e- t等,u(t)可取
3.2.2 粒子初始化
(1)随机生成粒子群体S = {P1,…,PN},计算粒子适应值,并随机初始化每个粒子的速度;令每个粒子历史最优Pbi = Pi(i = 1,…,N),粒子群体最优集合β = {P1,…,PN},并随机乱序分配给N个粒子,作为各自群体最优位置。
(2)更新z*和z+,
为每个粒子个体指定一个固定权重向量,其中:上标i对应于第i个个体(i = 1,…,N,N为种群规模);下标j对应于粒子的第j维(j = 1,…,n)。
3.2.3 粒子当前位置更新
对粒子群体S中所有个体,更新个体的位置与速度,并计算粒子适应值[10]。
粒子位置更新公式为

粒子速度更新公式为

3.2.4 个体历史最优位置
(1)更新z*和z+。
(2)若Ts[ f (Pt)|(Λi,z*,z+)]<Ts[ f (Pi)|(Λi,z*,z+)], 则有Pbi = Pi,否则按步骤4操作,直至完成更新S和z*。
3.2.5 群体最优位置更新
(1)令
(2)随机打乱集合β中的N个位置乱序,随机分配给N个粒子作为它们的群体最优位置。
(3)如果满足终止条件,则输出β,否则返回步骤(2)。
3.3 算法性能测试
采用Restrain函数对算法性能进行测试,即

图5
图6
图7
测试结果表明,使用FPSO对测试函数进行寻优时,经过49次迭代后,目标函数收敛于(0,0)点,最小值为0,相比于改进前的PSO,FPSO拥有更好的快速性和全局收敛性。
4 电机优化设计
参照图8所示的直线电机,对本文所研究的双次级永磁同步直线电机进行优化设计。
图8
表3 电机结构参数对比
Tab.3
电机结构参数 | |||||
---|---|---|---|---|---|
l | h | δ | τ | S | |
优化前 | 36.21 | 13.26 | 1.85 | 18.46 | 3.15 |
优化后 | 38.26 | 14.57 | 2.12 | 18.84 | 3.04 |
图9
图10
图11
图12
通过仿真实验可以得到,优化后的电机平均推力为41.25N,推力波动率0.46%,峰值电流为3.00A,推力电流比为13.75。相比于优化前的电机,经过FPSO优化后的电机平均推力增大了10.8%,推力波动减小了34.6%,峰值电流减小了7.7%,表明经过该算法优化后的电机推力大,推力波动小,峰值电流小,效率高,电机运行性能良好。
5 结论
本文根据双次级永磁同步直线电机的结构特点,构建电机的有限元模型,采用支持向量机建立快速计算模型,引入“填充函数”对传统粒子群算法进行改进,使用改进后的算法对电机结构参数进行优化,得到一组最优的电机结构参数,建立相对应的有限元模型,进行仿真实验,实验结果表明,优化后的具有推力波动小、推力大、电流小、稳定性强及效率高等优点,达到了电机优化设计的目标。
参考文献
多目标分解随机粒子群算法及其在直线电机优化设计中的应用
[J].
DOI:10.1016/j.jenvp.2019.101360
URL
PMID:31885519
[本文引用: 1]
A foundational tool for making causal inferences is the emulation of randomized control trials via variable interventions. This approach has been applied to a wide variety of contexts, from health to economics [4, 7]. Variable interventions have long been studied in independent and identically distributed (iid) data contexts, but recently non-iid settings, such as networks with interacting agents [9, 20, 32] have attracted interest. In this paper, we propose a type of structural intervention [14] relevant in network contexts: the network intervention. Rather than estimating the effect of changing variables, we consider changes to social network structure resulting from creation or severance of ties between agents. We define the individual participant and average bystander effects for these interventions and describe identification criteria. We then prove a series of theoretical results that show existing identification theory obtains minimally KL-divergent distributions corresponding to network interventions. Finally, we demonstrate estimation of effects of network interventions via a simulation study.
A multi-objective decomposition-based stochastic particle swarm optimization algorithm and its application to optimal design for linear motor
[J].
DOI:10.1016/j.jenvp.2019.101360
URL
PMID:31885519
[本文引用: 1]
A foundational tool for making causal inferences is the emulation of randomized control trials via variable interventions. This approach has been applied to a wide variety of contexts, from health to economics [4, 7]. Variable interventions have long been studied in independent and identically distributed (iid) data contexts, but recently non-iid settings, such as networks with interacting agents [9, 20, 32] have attracted interest. In this paper, we propose a type of structural intervention [14] relevant in network contexts: the network intervention. Rather than estimating the effect of changing variables, we consider changes to social network structure resulting from creation or severance of ties between agents. We define the individual participant and average bystander effects for these interventions and describe identification criteria. We then prove a series of theoretical results that show existing identification theory obtains minimally KL-divergent distributions corresponding to network interventions. Finally, we demonstrate estimation of effects of network interventions via a simulation study.
永磁同步直线电机磁阻力分析及抑制措施
[J].
Analysis and suppression measures of magnetic resistance force in permanent magnet linear synchronous motors
[J].
初级永磁型游标直线电机设计与静态特性分析
[J].
Design and analysis of a new linear primary permanent magnet vernier machine
[J].
基于有限元法的圆筒型直线电机温度场数值计算
[J].
Numerical calculation of temperature field for tubular linear motor based on finite element method
[J].
永磁同步直线电机定位力分析与实验研究
[J].定位力是推力波动的主要组成部分,也是评价永磁直线同步电机性能的重要标志。在建立考虑端部效应的磁路分析模型基础上,采用能量法理论计算定位力,分析了影响定位力波动的关键因素,得出合适的槽型和改变边齿形状可以有效减小定位力。通过有限元仿真和实验,证明了理论计算和结论的正确性。
Detent force analysis and experiment for permanent magnet linear synchronous motor
[J].定位力是推力波动的主要组成部分,也是评价永磁直线同步电机性能的重要标志。在建立考虑端部效应的磁路分析模型基础上,采用能量法理论计算定位力,分析了影响定位力波动的关键因素,得出合适的槽型和改变边齿形状可以有效减小定位力。通过有限元仿真和实验,证明了理论计算和结论的正确性。
A v-twin support vector machine (v-TSVM) classifier and its geometric algorithms
[J].DOI:10.1016/j.ins.2010.06.039 URL [本文引用: 1]
Multi-objective optimization problem with complicated Pareto sets, MOEA/D and NSGA-Ⅱ
[J].DOI:10.1109/TEVC.2008.925798 URL [本文引用: 1]
改进多目标粒子群优化算法及其在电弧炉供电优化中的应用
[J].
Improved multi-objective particle-swarm algorithm and its application to electric arc furnace in steelmarking process
全局最优的邻域估计
[J].
In order to find the global optimum, it’s crucial to prospect its neighborhood of global optimum(NGO). Inspired by
force balance relationship of the gravity center(GC) of a sand table, a method of prospecting NGO based on GC is presented
in this paper. Firstly, space transformation technique is used such that the GC of the optimization space can be “moved” close
to the NGO. Then, the NGO whose geometric center is the “moving” GC narrows down gradually. Experiment results of
function test and engineering application show that the NGO can be determined effectively with small population size within
less iterations by using the proposed method.
Prospecting neighborhood of global optimum
[J].
In order to find the global optimum, it’s crucial to prospect its neighborhood of global optimum(NGO). Inspired by
force balance relationship of the gravity center(GC) of a sand table, a method of prospecting NGO based on GC is presented
in this paper. Firstly, space transformation technique is used such that the GC of the optimization space can be “moved” close
to the NGO. Then, the NGO whose geometric center is the “moving” GC narrows down gradually. Experiment results of
function test and engineering application show that the NGO can be determined effectively with small population size within
less iterations by using the proposed method.
/
〈 |
|
〉 |
