基于RSM和BP-GA的甲基橙脱色条件的优化

Optimization of decolorization conditions of methyl orange based on RSM and BP-GA

通过单因素实验,考察温度、pH及甲基橙初始质量浓度对黄浆水中甲基橙脱色时间的影响。在此基础上,以响应面法(RSM)分析和前反馈(BP)人工神经网络(ANN)对甲基橙脱色进行建模,并采用遗传算法(GA)对BP神经网络模型进行优化。结果表明,BP-GA和RSM模型对黄浆水中甲基橙脱色的实验数据均能进行拟合和预测,而BP神经网络比RSM具有更好的拟合能力和预测能力。通过插值算法曲线拟合、BP-GA模型寻优以及实验验证,当黄浆水中甲基橙质量浓度为5~10 mg/L时,优化脱色条件为温度34.4℃,pH 5.5,甲基橙初始质量浓度5 mg/L,脱色时间约16 min。

The effects of temperature,initial pH and mass concentration of methyl orange on decoloriza⁃tion time of methyl orange in yellow seriflux were investigated by single factor experiment.On this basis,methyl orange decolorization was modeled by response surface model(RSM)and back-propagation(BP)artifi⁃cial neural network(ANN),and genetic algorithm(GA)was used to optimize BP model.The results showed that BP-GA and RSM model could fit and forecast the experimental data about decolorization of methyl or⁃ange,and BP model had better fitting ability and prediction ability than RSM.Through curve fitting of interpola⁃tion algorithm,optimization of BP-GA and experimental verification,when the mass concentration of methyl orange in yellow slurry was 5~10 mg/L,the optimum decolorization conditions were as follows:temperature 34.4℃,pH 5.5,initial mass concentration of methyl orange 5 mg/L,decolorization time 16 min.