GANN模型在十溴联苯醚分散液液微萃取中的应用

文章研究目的在于用遗传神经网络模型(GANN模型)快速优化水样中十溴联苯醚分散液液微萃取的萃取条件。以水样中十溴联苯醚分散液液微萃取的正交试验为训练样本,建立十溴联苯醚分散液液微萃取条件的遗传神经网络模型。比较遗传神经网络模型和BP神经网络模型的学习速度、学习精度及网络泛化能力。采用Matlab遗传算法工具箱运用遗传神经网络模型对影响萃取回收率的因素进行优化求解,获得了水样中十溴联苯醚分散液液微萃取优化后的萃取条件,并进行实验验证。文章建立的遗传神经网络模型得到的预测值与实验值平均偏差为14.41%,R2为0.8887;最佳DLLME萃取条件为10μL四氯乙烯、0.71mL丙酮、pH=5、离子强度为20%NaCl、萃取时间10min;优化后十溴联苯醚分散液液微萃取的萃取回收率和富集因子比优化前分别提高了54%和580。

Application of Genetic Algorithm Neutral Network Model in Dispersive Liquid-Liquid Microextraction Condition Optimization of Decabrominated Diphenyl Ether

The study aims to optimize dispersive liquid-liquid microextraction (DLLME)of decabrominated diphenyl ether (BDE-209) in water samples rapidly through genetic algorithm neural network (GANN) model. A GANN model was established based on an orthogonalized experiment of BDE -209 DLLME. Convergence speed,learning precision and generalization were compared between BP neural network model and GANN model. Matlab genetic algorithm toolbox was applied to seek a solution of DLLME optimization,and the optimum DLLME conditions were validated by experiments. Average errors between predicted and experimental values of extraction recovery were less than 14.41%,and R2 was 0.8887. The optimum DLLME condition of BDE-209 was 10μL of tetrachloroethylene,0.71mL of acetone,pH 5,20% of NaCl and extraction time of 10min. Results showed that the extraction recovery and enrichment factor of optimum condition had an increase of 54% and 580 respectively through GANN model.