Performance evaluation of nanotubular halloysites from weathered pegmatites in removing heavy metals from water through novel artificial intelligence-based models and human-based optimization algorithm

The efforts of this study aimed to evaluate the feasibility of the nanotubular halloysites in weathered pegmatites (NaHWP) for removing heavy metals (i.e., Cd²⁺, Pb²⁺) from water. Furthermore, two novel intelligent models, such as teaching-learning-based optimization (TLBO)-artificial neural network (ANN), and TLBO-support vector regression (SVR), named as TLBO-ANN and TLBO-SVR models, respectively, were proposed to predict the Cd²⁺ and Pb²⁺ absorption efficiencies from water using the NaHWP absorbent. Databases used, including 53 experiments for Pb²⁺ absorption and 56 experiments for Cd²⁺ absorption from water, under the catalysis of different conditions, such as initial concentration of Pb²⁺ and Cd²⁺, solution pH, adsorbent weight, and contact time. Subsequently, the TLBO-ANN and TLBO-SVR models were developed and applied to predict the efficiencies of Cd²⁺ and Pb²⁺ absorption from water, aiming to evaluate the role as well as the effects of different conditions on the absorption efficiencies using the NaHWP absorbent. The standalone ANN and SVM models were also taken into consideration and compared with the proposed hybrid models (i.e., TLBO-ANN and TLBO-SVR). The results showed that the NaHWP detected in a Kaolin mine (Vietnam) with 70% nanotubular halloysites is a potential adsorbent for water treatment to eliminate heavy metals from water. The two novel hybrid models proposed, i.e., TLBO-ANN and TLBO-SVR, also yielded the dominant performances and accuracies in predicting the Cd²⁺ and Pb²⁺ absorption efficiencies from water, i.e., RMSE = 1.190 and 1.102, R² = 0.951 and 0.957, VAF = 94.436 and 95.028 for the TLBO-ANN and TLBO-SVR models, respectively, in predicting the Pb²⁺ absorption efficiency from water; RMSE = 3.084 and 3.442, R² = 0.971 and 0.965, VAF = 96.499 and 96.415 for the TLBO-ANN and TLBO-SVR models, respectively, in predicting the Cd²⁺ absorption efficiency from water. Furthermore, the validation results also demonstrated these findings in practice through 23 experiments with the accuracies of 98.3% and 98.37% for the TLBO-ANN and TLBO-SVR models, respectively, in predicting the Pb²⁺ absorption efficiency from water; the accuracies of 98.3% and 97.46% for the TLBO-ANN and TLBO-SVR models, respectively, in predicting the Cd²⁺ absorption efficiency from water. Besides, solution pH was evaluated as the most critical parameter that can be adjusted to enhance the performance of the absorption of the heavy metals in this study. By using the NaHWP absorbent and the novel proposed intelligent models developed, heavy metals can be eliminated entirely from water, providing pure water/clean freshwater without any risk of adverse health effects for the short term or long term.