Taguchi-based and machine learning optimization of methylene blue removal and antioxidant activity of Lavandula stoechas L. biomass

Abstract

This study utilized natural and economical biomass of Lavandula stoechas L. to eliminate methylene blue dye from water and investigate its ability to be an antioxidant source within a zero-waste framework. A Taguchi design of experiments (L25) was used to evaluate the impacts of four factors: beginning concentration (10-50 mg/L), adsorbent dosage (0.01-0.15 g/50 mL), contact time (0-150 min), and pH (2-12), across four distinct levels. Machine learning models were integrated after the Taguchi analysis to validate the robustness of the optimal parameter combination, predict adsorption efficiency under different experimental conditions, and quantitatively assess the relative importance of process variables. The machine learning results supported the Taguchi and ANOVA findings by confirming the dominant influence of pH and adsorbent dosage on adsorption performance. The adsorption raw data were analyzed using non-linear Langmuir, Freundlich, Dubinin-Radushkevich, Temkin, and Sips isotherm models. The maximum adsorption capacities were determined to be 49.09 mg/g and 47.88 mg/g, respectively, based on the Langmuir and Sips isotherms. The study concluded that the adsorption kinetics adhered to the pseudo-second order kinetic model, and that many mechanisms, including intraparticle and film diffusion, were significant in the adsorption as per the Weber Morris and Boyd models. Reusability experiments demonstrated that L. stoechas L. biomass yielded highly successful results for a maximum of five cycles. All findings demonstrated that L. stoechas L. biomass, regarded as a zero-waste biomass, serves as a promising and eco-friendly adsorbent for the elimination of cationic MB dye.