改进遗传算法的ERT有限元模型拓扑结构优化  

Topology optimization of ERT finite element model based on improved GA

作　　者：肖理庆[1,2] 王化祥[1] 程红林[2] 徐晓菊[2] 

机构地区：[1]天津大学电气与自动化工程学院,天津300072 [2]徐州工程学院信电工程学院,徐州221111

出　　处：《仪器仪表学报》2012年第7期1490-1496,共7页Chinese Journal of Scientific Instrument

基　　金：国家自然科学基金(60820106002;60532020;50937005);"青蓝工程";天津市自然科学基金(11JCYBJC06900);江苏省高校自然科学研究项目(11KJD120003);徐州市科技计划项目(XM12B078)资助

摘　　要：为了提高电阻层析成像逆问题精度,利用基于区间算法与粒子群算法的改进遗传算法,优化有限元模型拓扑结构。以模型每一层半径为变量(最外层除外),分别以有限元平均质量、敏感场均匀分布时模型均方根值的倒数以及两者的乘积为适应度函数,并引入三角形最长边与最短边的比值作为惩罚函数。仿真与实际实验结果表明,相比传统按等间隔原理剖分的有限元模型及其改进模型,优化后的模型能有效提高逆问题的精度,且以敏感场均匀分布时模型均方根值的倒数为适应度函数效果最理想。In order to improve the accuracy of the inverse problem in electrical resistance tomography （ ERT）, the topology of ERT finite element model is optimized by applying an improved genetic algorithm based on interval algorithm and particle swarm optimization. The radius of each layer of the model （ except the outermost layer） is utilized as a variable ; and the average quality of the finite element, the reciprocal of the root mean square （RMS） of the model and the product of the two are chosen as the fitness functions, respectively. Also the ratio of the longest side to the shortest side of the triangle element is introduced as a penalty function. Both simulation and experiment results demonstrate that, compared with typical equidistance finite element model and its improved model, the improved genetic algorithm can effectively improve the imaging precision. In addition, the reciprocal of the RMS of the model is proved to be superior to the other two forms as the fitness function.

关 键 词：电阻层析成像 逆问题 区间算法 粒子群算法 遗传算法 惩罚函数 

分 类 号：TH701[机械工程—仪器科学与技术]