The aim of study was to analyze diclofenac (DCF) removal from aqueous solutions by metal-organic framework MIL-100(Fe) using multi-parameter batch experiments and artificial neural network (ANN) modeling. First, single-parameter experiments were performed in terms of initial solution pH, MIL-100(Fe) dosage, initial DCF concentration, and temperature. The DCF removal decreased with an increase in pH from 5 to 10 and became negligible at pH 12. The kinetic and equilibrium data showed that DCF removal reached an equilibrium at 12 h, with a maximum capacity of 414.6 mg/g from the Langmuir isotherm model. The DCF removal was enhanced with increasing temperature. Multi-parameter experiments (n = 56) conducted under 28 duplicate experimental conditions showed DCF removal rates between 70.8 – 90.8% with a final pH range of 4.5 – 5.4 for most of the experimental conditions. The ANN model was developed based on the multi-parameter experimental data. The optimal topology for the ANN model was determined to be 4:7:6:2 (4 input variables, 7 neurons in the first hidden layer, 6 neurons in the second hidden layer, and 2 output variables). Among the four input variables, temperature was the most important variable affecting DCF removal rate under the given experimental ranges.

研究的目的是使用多参数批处理实验和人工神经网络（ANN）模型分析金属有机框架MIL-100（Fe）从水溶液中去除双氯芬酸（DCF）的能力。首先，根据初始溶液的pH，MIL-100（Fe）剂量，初始DCF浓度和温度进行单参数实验。随着pH值从5增加到10，DCF去除量降低，在pH 12时可忽略不计。动力学和平衡数据表明，Langmuir等温模型中DCF去除量在12 h达到平衡，最大容量为414.6 mg / g。 。DCF的去除随着温度的升高而增强。多参数实验（ñ = 56）在28个重复的实验条件下进行的结果表明，在大多数实验条件下，DCF去除率在70.8 – 90.8％之间，最终pH范围为4.5 – 5.4。基于多参数实验数据开发了人工神经网络模型。将ANN模型的最佳拓扑确定为4：7：6：2（4个输入变量，第一个隐藏层中的7个神经元，第二个隐藏层中的6个神经元和2个输出变量）。在四个输入变量中，温度是在给定实验范围内影响DCF去除率的最重要变量。