【作者】 孟丽霞；

【导师】 赵剑强；

【作者基本信息】 长安大学 ， 环境科学， 2008， 硕士

【摘要】 废碱液是石油化工行业排放的含有高浓度COD、硫化物和挥发酚的高毒碱性废水,本文皆在找寻一种效果好且比较经济的,在常温常压条件下对废碱液进行处理的方法。本试验采用硫酸亚铁(FeSO4)、三氧化二铁(Fe2O3)和二氧化锰(MnO2)的混合物为催化剂在常温常压下对废碱液进行催化氧化,使大部分的硫化物(S2-)氧化成硫代硫酸盐(S2O32-)、亚硫酸盐(SO32-)和硫酸盐(SO42-),挥发酚组分氧化为二氧化碳(CO2)和水(H2O)或其它中间产物,从而COD浓度有一定程度的降低。考察了影响去除率的各种因素,试验结果表明:在反应温度为65℃,酸碱度PH为10.5,反应时间为40h,催化剂FeSO4、Fe2O3和MnO2各为1.5g/L等条件下,酚的去除率可达到89%以上,S2-的去除率可达99%,COD的去除率可达68%,催化剂再生后可循环使用,满足处理要求。本文根据人工神经网络理论,针对废碱液处理模拟预测的特点,提出了基于反应温度T、反应时间t、各催化剂投加量m和PH值4个参数的废碱液处理模拟预测BP神经网络模型。本文以MATLAB为计算平台,用神经网络工具箱对模型进行训练,通过对BP模型预测结果与实测值比较表明:训练稳定后的BP网络模型对废碱液处理效果的预测推断具有很好的精度, COD和苯酚的BP网络模拟结果与实测值相对误差最大值分别为7.3052%和8.0075%,均不超过10%,误差在可接受范围内,模型较令人满意。更多还原

【Abstract】 Spent caustic is a kind of high toxic and alkaline industrial waste water discharged from the petroleum and chemical plant, and it contains high content of COD, large quantity of sulfide and phenol. The aim of this paper is to find an effective and economic method to treat the spent caustic under room temperature and atmospheric pressure.This test chose the mixture of ferrous sulfate (FeSO4), ferric oxide(Fe2O3) and manganese(MnO2) as to be a catalyst to oxidize the spent caustic by air under room temperature and atmospheric pressure, almost all the sulfide(S2-) were oxidized to be thiosulfate(S2O32-), sulfite(SO32-), and sulfate(SO42-), the phenols were oxidized to be carbon dioxide(CO2)and water(H2O) or others’intermediate products, so the concentration of COD was reduced at the mean time. The various conditions which affect removal rates were studied, the experimental showed that the phenols removal was more than 99%, S2- and COD removal reached 99% and 68% respectively. Treatment condition is as follows: reaction temperature of 65℃, PH of 10.5, catalytic oxidation reaction time of 40h, FeSO4、Fe2O3 and MnO2 catalysts of 1.5g/l respectively. The catalysts could be circulated, the result satisfied the request of treatment.A back propagation (BP) artificial neural networks (ANN) model with 4 parameters of reaction temperature, reaction time, the catalysts amount and PH was developed for spent caustic treating prediction based on the theory and method of ANN and the characteristics of spent caustic treating. The paper adopted MATLAB as the computational platform to train the model with the Neural Network toolboxes. By comparing the prediction results of the BP model with the measured data, it is proposed that the result of BP model which was stable after trained simulating the spent caustic treating had high precision, the maximum relative errors between the prediction results of the BP model and the measured data of COD and phenols were 7.3052% and 8.0075% respectively, which were not more than 10%,the errors were acceptable, so the model was satisfied.更多还原