|
چکیده
|
This study deals with the intelligent modeling of reactive yellow 84 dye removal from aqueous solutions by zinc oxide (ZnO) nanoparticles based on inorganic light expanded clay aggregates (LECA). Accordingly, a set of experimental data were utilized to develop robust models by adjusting the initial feed concentration, initial pH, de-colorization time, catalyst concentration, and lamp power as input factors. The Leverage method, a common method for detecting outliers, proved all measured data were reliable. After definition of input factors, four intelligent approaches, i.e., Multi-Layer Perceptron (MLP), Gaussian Process Regression (GPR), Radial Basis Function (RBF), and Adaptive Neuro Fuzzy Inference System (ANFIS) with subtractive clustering were used to establish exact models for predicting the dye removal. Among them, the GPR based-model produced the best results with AARE, RRMSE, and R2 values of 4.87%, 6.21%, and 97.31%, respectively, for the test dataset. The remaining models also provided satisfactory outputs with AAREs between 6.59% and 9.81%. In addition, the influence of each operating parameter on the dye removal was properly described for the novel models. Finally, the most effective parameters on removal performance were determined by using a sensitivity analysis based on the GPR model.
|