Artificial intelligence-assisted modelling of heavy metal adsorption using cellulose-based and bio-waste adsorbents: A focus on ANN and ANFIS architectures

Abstract

This review explores the application of artificial intelligence (AI) models, specifically artificial neural networks (ANN) and adaptive neuro-fuzzy inference systems (ANFIS), in predicting heavy metal adsorption performance using bio-based adsorbents. Focus is placed on sustainable materials such as cellulose nanocrystals (CNCs), agricultural waste-derived biochar, and microbial biomass. The review compiles more than 60 studies over the past decade, analysing model structures, input-output variables, training algorithms, and validation strategies. Performance metrics reveal that most ANN models achieve R² > 0.98, with NARX-ANN reaching as high as 0.9998 in time-resolved batch adsorption simulations. ANFIS models offer added interpretability through fuzzy rule extraction, though their adoption remains limited. Optimization techniques such as particle swarm optimization (PSO) and genetic algorithms (GA) improved RMSE by 5–15%.Comparative evaluation shows variability in model generalization depending on input complexity and adsorbent type. Despite promising results, the review identifies gaps in dataset standardization, model validation, and real-world applicability under multicomponent or noisy conditions. The novelty of this review lies in its cross-comparative benchmarking of ANN and ANFIS architectures applied specifically to bio-adsorbents, and its recommendations for engineering-grade AI deployment in environmental remediation systems. Future research should incorporate deep learning, sensor integration, and regulatory-informed optimization to enhance model robustness and scalability in wastewater treatment applications.