Modelling and optimization of psychoactive pharmaceutical caffeine removal by electrochemical oxidation process: A comparative study between response surface methodology (RSM) and adaptive neuro fuzzy inference system (ANFIS)

Abstract

In this study, the modelling and optimization of psychoactive pharmaceutical caffeine removal in synthetic solution and real municipal wastewater effluent by electrochemical oxidation (EO) process was investigated using central composite design (CCD) based on response surface methodology (RSM) and adaptive neuro fuzzy inference system (ANFIS). The influence of independent variables including electrolysis time, current intensity, initial concentration of caffeine, and type of anode were studied. Results showed that anode type followed by electrolysis time are the most important variables affecting caffeine degradation. Both CCD and ANFIS models were able to successfully predict the electrochemical process behaviour, while ANFIS models performed slightly better (R2 = 0.993, RMSE = 2.694 for caffeine removal efficiency, and R2 = 0.976, RMSE = 0.261 for energy consumption). Identification of intermediates by liquid chromatography-tandem mass spectrometry (LC-MS/MS) suggests that the degradation of caffeine by the EO process follows an oxidation pathway similar to the mechanism proposed for other advanced oxidation processes. The optimal conditions determined using CCD were applied on real municipal wastewater effluent in which caffeine removal efficiency varied between 78.0 ± 4.3% and 92.5 ± 1.0% for different initial caffeine concentrations showing the effectiveness of the process. Finally, toxicity assessment with Daphnia magna showed that the EO of real municipal wastewater effluent in optimal conditions may increase the toxicity levels of the samples. Toxicity could be reduced by extending the electrolysis time or could be completely eliminated using granular activated carbon.