基于BP神经网络的间硝基苯磺酸钠CWPO降解条件优化

Optimization of CWPO degradation conditions of meta-nitro benzene sulfonic acid sodium salt based on BP artificial neural network

为了对电镀废水中的间硝基苯磺酸钠催化湿式过氧化氢氧化(CWPO)降解条件进行模拟及优化,建立了间硝基苯磺酸钠CWPO降解过程BP神经网络模型。经验证,模型预测值与试验值的平均相对偏差为0 81%,相关系数r和Nash-Suttcliffe模拟效率系数NSC分别为0.992 5和0.983 9。相对灵敏度分析表明,影响间硝基苯磺酸钠去除率(以TOC表示)的顺序从大到小为:反应温度、间硝基苯磺酸钠质量浓度、pH值、H_2O_2用量、反应时间、初始氧分压、催化剂用量。结合遗传算法以TOC去除率最高作为优化目标,分别对降解条件进行优化。经对比,带成本约束的优化降解结果(99.36%)比试验中的TOC去除率平均值(85.51%)提高了 10%以上,同时,优化后的降解成本(2.03元)相比无成本约束条件下的降解成本(2.38元)降低了近15%(0.35元)。

This article is intended to present our work on the optimization of CWPO degradation conditions of recta-nitro benzene sulfonic acid sodium salt based on BP artificial neural network, which is used to simulate and optimize the degradation conditions of Catalytic Wet Hydrogen Peroxide Oxidation degradation process. As what is done, we have established BP artificial neural network and investigated the effect of the amount of catalyst, initial pressure of oxygen, initial concentration of meta-nitro benzene sulfonic acid sodium salt on the process of degradation, and selected the amount of H2O2, reaction temperature, reaction time and pH, 130 groups of experimental data as the training samples for the BP neural network model. In so doing, we have established the other 20 groups as validation samples to test the reliability of the model. The relative deviation between the values predicted by the above mentioned neural network model and the experimental ones ranged from 0.08% to 5.47%, and the average relative deviation （0.81% ） with the correlation coefficient （r = 0.996 2） and Nash-Suttcliffe （NSC = 0.983 9）. Then we began to the factors＇ order in significance on the TOC removal rate by the factors＇ sensitivity analysis and discovered the influence degree of these factors on TOC removal rate which can be shown as follow： reaction temperature 〉 initial concentration of meta-nitro benzene sulfonie acid sodium salt 〉 pH 〉 amount of H2O2 〉 reaction time 〉 initial pressure of oxygen 〉 amount of catalyst. Finally, with the optimization goals of the highest TOC removal rate in mind, we have succeeded in combining the BP model with the genetic algorithm to optimize the result of the meta-nitro benzene sulfonic acid sodium salt degradation and the experimental cost through the optimization of the experimental conditions. The mean TOC removal rate （99.36%） has been optimized by overcoming a series of cost constraints as well as noncost constraints.