Integrated experimental and machine learning approach for Reactive Black 5 removal using straw-derived biochars

Abstract

Wastewater contaminated with synthetic dyes poses significant environmental and health risks, and cost-effective solutions are urgently needed. Conventional adsorbents are often costly and exhibit limited efficiency, fostering an increasing pursuit for sustainable and economically viable alternatives. This study investigates the use of biochars derived from wheat straw (WSP), oil seed rape straw (OSR), and Miscanthus straw (MSP) for the removal of Reactive Black 5 dye from water. The novelty of this work lies in combining agricultural biochars with advanced data-driven approaches—including generalized linear modeling, machine learning (Random Forest, Gradient Boosting), and Monte Carlo simulations—together with ecotoxicological validation, to bridge experimental results with predictive modeling. Biochars produced at 550 °C and 700 °C were evaluated through batch adsorption experiments, FT-IR and SEM/EDS analyses, equilibrium and isotherm modeling, desorption and regeneration capabilities, and real effluent application. WSP700 achieved the highest removal efficiency (94%) at 75 g L−1, with adsorption most effective at pH 5. Although higher dosages improved removal, adsorption capacity decreased due to site aggregation. Random Forest provided the best fit for capturing non-linear behavior, whereas cross-validation and external interpolation revealed that Gradient Boosting and linear/penalized regressions offered better generalization performance. Regeneration tests showed 70% desorption efficiency after seven cycles, and ecotoxicological assays revealed a marked EC50 increase (1.131 → 2.204), indicating reduced environmental risk. Overall, the results highlight the potential of WSP and OSR biochars as efficient, regenerable, and environmentally safe adsorbents for dye removal in wastewater treatment, supporting the development of sustainable circular-economy strategies.