Oxidation kinetics of water contaminants: New insights from artificial intelligence

Abstract

Degradation of water contaminants through the advanced oxidation processes (AOP) has become the focus of strategists of environmental science and technology. Hydroxyl radicals are regarded as promising oxidants for the efficient decomposition of the organic contaminants. Nevertheless, understanding and monitoring the kinetics of the hydroxyl radical reaction has remained cumbersome to be deciphered by the aid of mathematical and statistical analyses. Herein, a new stochastic gradient boosting (SGB) decision tree technique based on the quantitative structure–property relationship (QSPR) method was developed to model and capture an image of the degradation rate constant of the hydroxyl radicals. An artificial model was constructed, trained, and tested on the bedrock of 457 different cases of water contaminants from 27 chemical structures. Several statistical techniques, including outlier detection, regression, sensitivity, and error analyses were served to validate the reliability of the proposed model. The outcomes showed that the developed model could appropriately estimate the logarithmic hydroxyl radical rate constants of numerous water contaminants with the promising R‐squared of 0.97, and a quite low mean absolute relative error of 0.86%. Moreover, a sensitivity analysis revealed that Burden eigenvalue was found to be the most effective parameter as the input of the model. Finally, a comparison study was performed between the proposed QSPR and the models previously suggested, where the superiority of the present model was uncovered.

Prediction of reaction rate constants of hydroxyl radical with chemicals in water

“…Therefore, the optimal SVM model cannot be used for the prediction these compounds. The reason is that they have dissimilar molecular structures with other chemicals in the training set (Keivanimehr et al, 2020; Yu, 2020a, 2020b). Furthermore, Figure 2 shows that there are 76 organic compounds with leverages h > h * of 0.03 (here h * is warning leverage and calculated with h * = 3 × ( p + 1)/ n , p, and n are, respectively, the numbers of descriptors and compounds in training set) (Liao et al, 2019; Liu, Bai, et al, 2020; Liu, Deng, et al, 2020; Yu et al, 2019; Yu et al, 2019).…”

On the Investigation of Effective Factors on Electronic Structure Properties of Transition Metal Complexes: Robust Modeling Using GPR Approach

“…A portion of data may show inconsistency with the dataset, with some data being suspected. Such data points majorly imply empirical errors [69,70]. It is necessary to identify suspected data points since they would diminish predictive performance [71].…”

Recent Advances in Fly-Ash-Based Geopolymers: Potential on the Utilization for Sustainable Environmental Remediation

“…Research efforts on the optimization procedures with a view to arrive at a set of conditions that yields optimal immobilization or adsorption effects and delineates the interaction between the synthesis and process parameters are also important. These should include process optimization of experimental approaches using methodologies such as Taguchi and response surface methods and artificial intelligence (machine learning or neural network) to ascertain new insights under realistic operating conditions and techno-economic analysis. − Such studies should help in their cost–benefit evaluations and comparison to available/emerging technologies, aiding progress toward scale-up and industrial applications. These represent knowledge gaps and opportunities to be exploited moving forward.…”