Development of artificial intelligence for modeling wastewater heavy metal removal
- The application of AI models on heavy metals (HMs) simulation are reviewed.
- Heavy metal removal techniques (RT) with various material are described.
- All applied AI models for each heavy metal RT are thoroughly presented.
- A critical analysis and assessment are cultivated for the AI models advancement.
- Research gaps and prospective researches are recognized based on the survey.
The presence of various forms of heavy metals (HMs) (e.g., Cu, Cd, Pb, Zn, Cr, Ni, As, Co, Hg, Fe, Mn, Sb, and Ce) in water bodies and sediment has been increasing due to industrial and agricultural runoff. HM removal in nature is highly stochastic, nonlinear, nonstationary, and redundant. Over the last two decades, the implementation of artificial intelligence (AI) models for HM removal has been massively conducted. The divergence in the selection of predictors, target variables, the optimization, normalization of the algorithm, function, and architecture of AI models are time-consuming processes, which limit the optimal use of such models for HM removal simulation. The selection of sustainable, cost-efficient, and user-friendly treatment techniques that have minimal reverse impact on the ecosystem is immensely challenging. The focus of the established researches is to find an optimal AI models for specific removal techniques. Predictors and target variables can be sorted using several techniques, and the selection of algorithm, function, and architecture based on individual treatment techniques have been coherently ordered and argued. In this review, each element of the predictive models and their corresponding treatment processes, including its pros and cons, are discussed thoroughly. The performance matrices are also discussed in accordance with the behavior of each model. Moreover, multiple perspectives that can enlighten interested multi-domain scientists and scholars, such as AI model developers, data scientists, wastewater treatment researchers, and environmental policymakers, on the actual status of the models’ progression are summarized. A comprehensive gap and assessments are also conducted to provide an insightful vision on this topic. Finally, several research directions, which could bridge the gap in the same domain are proposed and recommended on the basis of the identified research limitations.
SOURCE SCIENCE DIRECT