摘要
研究吸收光谱重叠严重的苯酚、邻苯二酚、间苯二酚和对苯二酚的4组分体系,针对神经网络易陷入局部极小等缺陷,将遗传算法与神经网络相结合,用遗传算法优化神经网络的初始权值和阈值,由神经网络输出的误差构造适应度函数,建立遗传神经网络算法,用紫外分光光度法同时测定混合的苯酚、邻苯二酚、间苯二酚和对苯二酚,预测集样品的相对平均误差分别为1.548%,1.6%,2.028%和-1.004%,对自来水水样的加标回收率分别为102.92%,101.53%,105.78%和103.17%。
The four-component system of phenol, catechol, resorcinol and hydroquinone was studied by UV spectrophotometry with serious overlapping peaks. Considering the defects of backpropagation neural network （BP）, the model was set up by the optimization of initial weights and thresholds of neural network using genetic algorithm and designing fitness function by output error. The contents of phenol, cateehol, resoreinol and hydroquinone were determined simultaneously by GA-BP-ANN model and ultraviolet spectrophotometry. For phenol, catechol, resorcinol and hydroquinone, the relative mean errors in the prediction set were 1.548%, 1.6%, 2. 028% and - 1.004%, and the recovery rate of water sample were 102.92%, 101.53%, 105.78% and 103. 17%.
出处
《淮海工学院学报（自然科学版）》
CAS
2009年第4期42-45,共4页
Journal of Huaihai Institute of Technology:Natural Sciences Edition
基金
江苏省高校自然科学研究计划项目(05KJB150003)
江苏省海洋生物技术重点建设实验室开放课题(2005HS010)
淮海工学院大学生实验创新课题
