Artificial neural network-based modeling of Malachite green adsorption onto baru fruit endocarp: insights into equilibrium, kinetic, and thermodynamic behavior
Marielle Xavier NascimentoBruna Assis Paim dos SantosManoel Marcos Santiago NassardenKezya dos Santos NogueiraRenata Gabriele da Silva BarrosRossean GolinAdriano Buzutti de SiqueiraLeonardo Gomes de VasconcelosEduardo Beraldo de Moraisa Postgraduate Program in Water Resources,Federal University of Mato Grosso,Cuiabá,Brazilb Department of Chemistry,Federal University of Mato Grosso,Cuiabá,Brazilc Department of Sanitary and Environmental Engineering,Federal University of Mato Grosso,Cuiabá,Brazil
DOI: https://doi.org/10.1080/15226514.2024.2354411
2024-05-19
International Journal of Phytoremediation
Abstract:In this study, artificial neural network (ANN) tools were employed to forecast the adsorption capacity of Malachite green (MG) by baru fruit endocarp waste (B@FE) under diverse conditions, including pH, adsorbent dosage, initial dye concentration, contact time, and temperature. Enhanced adsorption efficiency was notably observed under alkaline pH conditions (pH 10). Kinetic analysis indicated that the adsorption process closely followed a pseudo-second-order model, while equilibrium studies revealed the Langmuir isotherm as the most suitable model, estimating a maximum adsorption capacity of 57.85 mg g −1 . Furthermore, the chemical adsorption of MG by B@FE was confirmed using the Dubinin–Radushkevich isotherm. Thermodynamic analysis suggested that the adsorption is spontaneous and endothermic. Various ANN architectures were explored, employing different activation functions such as identity, logistic, tanh, and exponential. Based on evaluation metrics like the coefficient of determination ( R 2 ) and root mean square error (RMSE), the optimal network configuration was identified as a 5–11-1 architecture, consisting of five input neurons, eleven hidden neurons, and one output neuron. Notably, the logistic activation function was applied in both the hidden and output layers for this configuration. This study highlights the efficacy of B@FE as an efficient adsorbent for MG removal from aqueous solutions and demonstrates the potential of ANN models in predicting adsorption behavior across varying environmental conditions, emphasizing their utility in this field. The innovative aspect of this study lies in the utilization of a new and effective adsorbent for the removal of Malachite Green (MG), derived from the fruit endocarp of baru ( Dipteryx alata Vog.). The baru fruit endocarp, typically discarded as solid waste during processing, was found to possess favorable characteristics for adsorption processes and provides an adsorption capacity that exceeds that of most other similar adsorbents. Additionally, integrating Artificial Neural Networks (ANNs) enables accurate modeling of the adsorption process, eliminating the need for extensive laboratory experiments. This contributes significantly to wastewater treatment research, enhancing effectiveness and sustainability in unwanted dye removal.
environmental sciences