LAUSR

ABSTRACT Artificial neural network modeling and statistical analysis are used to optimize electrochemical removal of Reactive Black 5 using a Ti/(RuO2)0.8-(Sb2O3)0.2 electrode. Experimental design was used to analyze the influence of dye concentration (26.4–93.6 ppm), electrolyte concentration (NaCl (0.0062–0.1238 M), and current density (0.62–12.38 mA cm–2) in a batch treatment system. The response surface methodology and artificial neural network were appropriate methods to optimize the operating conditions in electrochemical degradation process. Optimization models was developed to assess the performance of the electrochemical degradation, where color removal percentage, COD removal percentage, and the energy consumption (KW/m3) were considered. A total removal of color and decrease in up to 73.77% of chemical oxygen demand within 180 min of treatment was obtained using the optimized conditions. The final neural network model, characterized by a 4–8-3 architecture, presenting performance index determination coefficient (R2) of 0.982 and mean square error of 0.0146. The neural model developed demonstrated an efficiently predictive performance and to optimize the parameters of the electrochemical oxidation process. Abbreviations: Absi (λmáx): Initial absorbance in the maximum wavelength; Absf (λmáx): Final absorbance in the maximum wavelength; ANN: Artificial Neural Network; CD: Current Density; CE: Counter Electrode; COD: Chemical Oxygen Demand; CR: Color Removal; DY: Dye Concentration; E: Energy consumption; EC: Electrolyte Concentration; I: Current Applied; MMO: Mixed Metal Oxide; MSE: Mean Squared Error; n: Amount of Experiments; PSO: Particle Swarm Optimization; R2: Determination Coefficient; RB5: Reactive Black 5; RE: Reference Electrode; RMSE: Square Root Mean Error; T: Electrolysis Time; U: Neuron Activation Value; V: Voltage; Vs: Solution Volume; XRD: X-Ray Diffraction; WE: Working Electrode; yical: Removal Percentage Obtained by the Model; yiexp: Removal Percentage Obtained Experimentally; β: Lower and Upper Limits