Machine Learning-Augmented
Real-Time Prediction and Analysis of Lead Adsorption Behaviour at Different Temperatures
Abraham Nivya Mary, Pawels Renu and Madhu G.
Res. J. Chem. Environ.; Vol. 30(1); 45-57;
doi: https://doi.org/10.25303/301rjce045057; (2026)
Abstract
Contamination of drinking water sources with lead is one of the most significant
environmental and public health problems and, consequently, highly efficient absorption-based
elimination strategies are highly sought after. Here, a machine-driven real-time
lead adsorption behavior for a series of temperatures is reported based on modeling
of the adsorption isotherm models. The experimental data is reproduced, followed
by the development of a novel prospective predictive framework. Mechanism of adsorption
was screened over various parameters: concentration of adsorbent (10–60 mg/L), pH
(4–9), contact time (30–180 min), adsorbent meal load (1–6 g/L) and temperature
(100 °C, 150 °C, 200 °C) in a bid to determine their effect on adsorption effectiveness.
Three isotherm equations, namely Dubinin–Radushkevich (D-R), Redlich–Peterson (RP)
and Sips, were used to solve the adsorption process. The model that fits the experimental
data better i.e. Sips model with the highest R² value (R² 0.9995), was selected
here, as the nearest fit. In order to improve the real-time predictive accuracy,
a Random Forest Regressor (RFR) model was constructed with experimental data and
resultant highest prediction accuracy (Mean Squared Error (MSE) 0.00009, Root Mean
Squared Error (RMSE) 0.00937, R² 0.97513). These results validate the applicability
of the model to predict adsorption of a set of experimental conditions.
Results show that the highest adsorption yield occurs under the elevated temperature,
optimal pH environment and longer contact time, while excess adsorbent dosing causes
adsorption saturation effect and decreases the adsorption enhancement. In this study,
the feasibility of machine learning assisted adsorption modelling for optimization
of the real-time water treatment processes is shown.
