摘要
目的比较微遗传算法与传统优化方法确定最优试验条件的效果。方法以琥珀酸维生素E、泊洛沙姆-188、0.1molNaOH作为影响因素;以粒径、ζ电位、克拉霉素(clarithromycin,CLA)乳剂相分布率作为评价指标,在Box-Behnken试验设计给定的条件下,通过最小二乘法拟合二阶响应面方程试验模型,分别运用Matlab2009a软件的函数模块、图形模块编制的遗传算法程序和ExpertDesign软件(Version7.1.6.0)的Box-Behnken设计的传统优化方法,筛选出最优试验条件,比较两种优化方法的效果。结果当评价指标和因素之间呈非线性回归关系时,Box-Behnken设计充分考虑到各影响因素间的交互作用,同时在中心点进行重复试验,提高了实验的准确性。经微遗传算法优化后试验因素琥珀酸维生素E、泊洛沙姆-188、0.1MNaOH最佳条件分别为:68.25%、0.52%、15.26%,评价指标粒径、ζ电位、CLA乳剂相分布率分别为:135.75nm、33.67mV、98.12%;经传统优化方法优化后试验因素琥珀酸维生素E、泊洛沙姆-188、0.1MNaOH最佳条件分别为:76.0%、0.5%、15.4%,评价指标粒径、ζ电位、CLA乳剂相分布率分别为:135.75nm、31.04mV、97.33%。结论对于三水平多因素的试验,采用Box-Behnken设计试验设计可以大大减少设计所花费的时间和成本,提高试验效率。微遗传算法相对于传统优化方法,可以从全局的角度搜索试验条件的最优组合,进行多个评价指标优化效果理想、程序可行、计算时间短、可以应用于解决多维解空间的实际问题。
Objective To compare the effects between micro- genetic algorithms （micro-GA） and conventional optimization method. Methods The effects of selected variables include tocopherol succinate,poloxamer 188,0. 1 M NaOH. , and the evaluation indicators included the particle size, ξ potential, the oil distribution of clarithromycin. Utilizing Box-Behnken Experimental Design to conduct the experiments and building up the model by least square method, the best experimental conditions were respectively obtained by micro-genetic procedure based on Matlab 2009a or Expert Design Soft（ Version 7.1.6. 0）. The best experimental conditions of two methods were analyzed. Results When it reveal the non-linear re- gression relationships between the evaluation indicators and factors, Box- Behnken Experimental Design fully took into account the interaction be- tween factors and repeated the test in the center to improve the accuracy. Optimized by micro-GA, tocopberol succinate, poloxamer 188, 0. 1 M NaOH were observed to be 68.25% ,0.52% ,15.26% ,the particle size,ξ potential, the oil distribution of clarithromycin were observed to be 135.75nm、33.67 mV、98. 12%. Optimized by the traditional optimization, tocopherol succinate, poloxamer 188,0. 1 M NaOH were observed to be 76.0% 、0. 5% ,15.4% ,the particle size,ξ potential,the oil distribution of clarithromycin were observed to be 135.75nm、31.04mV、97.33%. Con- clusion For the three-level multi-factor experiments, Box-Behnken Ex- perimental Design could greatly reduce design time and cost spent,improve the experimental efficiency. Compared with traditional optimization, micro- GA could search the optimal combination of test conditions from a global point of view. The micro-GA optimization of multiple evaluation results was satisfaction, the procedure was feasible, the computation time was shortand could be applied to solve multi-dimensional practical problems.
出处
《中国卫生统计》
CSCD
北大核心
2012年第3期337-340,344,共5页
Chinese Journal of Health Statistics
基金
国家自然科学基金项目(30872183)
山西省自然科学基金项目(2007011087)
关键词
微遗传算法
BOX-BEHNKEN设计
最优试验条件
优化效果
Micro-genetic algorithms
Box-behnken ex- perimental design
Optimal experimental condition
Optimization effect
