• 关键词：咖啡因  吸附  沉积物  影响因素  神经网络  线性方程

• 基金项目：中央高校基本科研业务费专项资金（No.20133800031650007）

• 余绵梓
• 1. 中山大学环境科学与工程学院, 广州 510275;2. 广东省环境污染控制与修复技术重点实验室, 广州 510275

• 袁啸
• 1. 中山大学环境科学与工程学院, 广州 510275;2. 广东省环境污染控制与修复技术重点实验室, 广州 510275

• 李适宇
• 1. 中山大学环境科学与工程学院, 广州 510275;2. 广东省环境污染控制与修复技术重点实验室, 广州 510275

• 胡嘉镗
• 1. 中山大学环境科学与工程学院, 广州 510275;2. 广东省环境污染控制与修复技术重点实验室, 广州 510275

• 李思妍
• 中山大学环境科学与工程学院, 广州 510275

• 赵雨晨
• 中山大学环境科学与工程学院, 广州 510275

• 摘要：咖啡因是一种在环境中广泛存在的药物，其在水系环境的迁移和分布主要受到吸附行为的影响.在实验室条件下，近似模拟自然河流水/泥界面，应用中心复合实验设计，考察了温度、pH、有机质含量和转速等对咖啡因吸附的影响；利用实验所得数据，分别拟合和验证了基于线性方程和神经网络的咖啡因吸附模型，通过对比拟合和验证结果得到适用于自然河流的咖啡因吸附模型.实验结果表明，咖啡因在沉积物中的吸附呈现先快速后缓慢的过程，30 h内吸附比例超过90%；咖啡因的吸附是放热反应，低温有利于咖啡因的吸附；而转速增大能促进咖啡因的吸附反应；pH和有机质含量的影响较小.模型模拟结果表明，两种模型均能较好地拟合吸附实验结果，但神经网络模型的拟合程度和精度均优于线性方程模型；且交叉验证结果表明，利用不同组数据进行训练，神经网络模型均取得了优于线性方程的拟合结果.因此，在所考察的因素和浓度范围内，神经网络模型较好地预测了自然河流沉积物中咖啡因的吸附行为.

• Abstract：Caffeine is a pharmaceutical product that is widely present in the environment. Its transportation and distribution in the aquatic environment are mainly determined by the sorption process. A water/sediment interface was roughly simulated in laboratory, and a central composite design was applied to investigate the influence of temperature, pH, organic content, and rotation speed on the sorption of caffeine. Based on the experimental data, a multi linear regression equation and a neural network model are applied to fit and the favorable sorption model for caffeine in the river systems was evaluated. The results show that the sorption process of caffeine onto sediment is initialed with a rapid sorption process and then a slow sorption until the equilibrium state is achieved. Over 90% of the total caffeine is adsorbed within 30 hours. The sorption of caffeine is an exothermic reaction, and the proportion of sorbed caffeine increases with decreasing temperature and increasing of rotation speed. while the changes of organic content and pH have little effect on the sorption of caffeine. Both models well fitbut the regression level and accuracy of the neural network model are better than those of the linear equation model. The results of cross validation based on different divisions of data also indicated that the neural network model simulates the sorption process of caffeine better than the linear equation model. Therefore, within the ranges of concentration of the investigated factors, the neural network model was able to predict the sorption of caffeine onto sediment in natural rivers reasonably.

• Key words：caffeine  sorption  sediment  parameters  neural network  linear equation

• 
  摘要点击次数： 1224 全文下载次数： 1851