Applied Chemistry Department, Faculty of Petroleum and Gas (Gachsaran), Yasouj University, Gachsaran, 75813-56001, Iran.
Chemistry Department, Yasouj University, Yasouj 75918-74831, Iran.
J Colloid Interface Sci. 2017 Nov 1;505:278-292. doi: 10.1016/j.jcis.2017.05.098. Epub 2017 May 27.
Two machine learning approach (i.e. Radial Basis Function Neural Network (RBF-NN) and Random Forest (RF) was developed and evaluated against a quadratic response surface model to predict the maximum removal efficiency of brilliant green (BG) from aqueous media in relation to BG concentration (4-20mgL), sonication time (2-6min) and ZnS-NP-AC mass (0.010-0.030g) by ultrasound-assisted. All three (i.e. RBF network, RF and polynomial) model were compared against the experimental data using four statistical indices namely, coefficient of determination (R), root mean square error (RMSE), mean absolute error (MAE) and absolute average deviation (AAD). Graphical plots were also used for model comparison. The obtained results using RBF network and RF exhibit a better performance in comparison to classical statistical model for both dyes. The significant factors were optimized using desirability function approach (DFA) combined central composite design (CCD) and genetic algorithm (GA) approach. The obtained optimal point was located in the valid region and the experimental confirmation tests were conducted showing a good accordance between the predicted optimal points and the experimental data. The properties of ZnS-NPs-AC were identified by X-ray diffraction; field emission scanning electron microscopy, energy dispersive X-ray spectroscopy (EDS) and Fourier transformation infrared spectroscopy. Various isotherm models for fitting the experimental equilibrium data were studied and Langmuir model was chosen as an efficient model. Various kinetic models for analysis of experimental adsorption data were studied and pseudo second order model was chosen as an efficient model. Moreover, ZnS nanoparticles loaded on activated carbon efficiently were regenerated using methanol and after five cycles the removal percentage do not change significantly.
两种机器学习方法（即径向基函数神经网络（RBF-NN）和随机森林（RF））被开发出来，并与二次响应面模型进行了评估，以预测超声辅助下从水溶液中去除灿烂绿（BG）的最大去除效率与 BG 浓度（4-20mg/L）、超声时间（2-6min）和 ZnS-NP-AC 质量（0.010-0.030g）的关系。所有这三个（即 RBF 网络、RF 和多项式）模型都使用四个统计指标，即决定系数（R）、均方根误差（RMSE）、平均绝对误差（MAE）和绝对平均偏差（AAD）与实验数据进行了比较。还使用图形图进行了模型比较。与经典统计模型相比，RBF 网络和 RF 的使用结果在两种染料的情况下都表现出更好的性能。使用期望函数方法（DFA）结合中心复合设计（CCD）和遗传算法（GA）方法对显著因素进行了优化。获得的最优点位于有效区域内，进行了实验验证测试，预测最优点与实验数据之间存在良好的一致性。通过 X 射线衍射、场发射扫描电子显微镜、能量色散 X 射线能谱（EDS）和傅里叶变换红外光谱对 ZnS-NPs-AC 的性质进行了鉴定。研究了用于拟合实验平衡数据的各种等温模型，并选择了 Langmuir 模型作为有效模型。研究了用于分析实验吸附数据的各种动力学模型，并选择了准二级动力学模型作为有效模型。此外，ZnS 纳米粒子在活性炭上的负载物有效地用甲醇再生，并且在五个循环后，去除率没有显著变化。
