2026-05-28T12:20:54Zhttps://riubu.ubu.es/oai/request

oai:riubu.ubu.es:10259/117372026-05-28T00:05:35Zcom_10259_6168com_10259_5086com_10259_2604col_10259_6169

Rehman, Hamid Debik, Eyup Ulucan-Altuntas, Kubra Manav-Demir, Neslihan Canci, Baris Iqbal, Mazhar Barros García, Rocío ur Rehman, Wasif Mohanty, Sanjay K Khan, Aqib Hassan Ali Bioleaching REE recovery Machine learning Waste management Metals Organic acids This review provides a comprehensive, data-driven perspective on rare earth element (REE) recoveries from various waste streams by bioleaching, integrating mechanistic insights, microbial performance data, advanced statistical and machine learning tools. A total of 77 observations across 10 waste types were analyzed via Bayesian meta-analysis, yielding an average REE recovery of 56.2 % (95 % credible interval: 51.1–61.0 %). Among the waste types, coal fly ash and electronic waste (e-waste) demonstrated the highest recoveries (76 % and 89 %, respectively). Fungi, particularly Aspergillus and Penicillium, performed better than bacteria, despite being less commonly used in bioleaching studies. Fungal-only systems typically achieved 60–76 % recovery, whereas values above 85 % were reported when fungal bioleaching was combined with chemical or physical pretreatments. Acidophilic bacteria exhibited the highest recovery efficiency among the bacterial species (66 %). The microbial consortia (combinations of fungi and bacteria) achieved up to 76 % recovery efficiency due to synergistic interactions. Importantly, many of the highest recoveries (≥90 %) reported in the literature refer to base metals such as Cu, Ni, and Zn, which are more easily solubilized than REEs; harmonizing claims requires distinguishing organism-only effects from organism + pretreatment strategies, and base metal recoveries from REE recoveries. Structural equation modeling (SEM) revealed that factors such as pH, type of waste, and process parameters, played key roles in determining REE recovery success. Among these, process variables (e.g. pH and pulp density) had the strongest direct influence (β = 0.895). Machine learning models, including support vector machine regression (SVMR) and K-nearest neighbor regression (KNNR), further highlight the importance of metal content, process parameters, and microbial presence. These models performed well, with R² values of 0.87 for SVMR and 0.787 for KNNR. Overall, this integrated approach demonstrates the potential for scaling-up bioleaching processes. By combining biological insights with predictive analytics, this integrated framework demonstrates strong foundation for industrial-scale REE recovery and supports shifting toward a more circular and sustainable economy 2026-05-27 2026-05-27 2025-12 info:eu-repo/semantics/article 2590-1230 https://hdl.handle.net/10259/11737 10.1016/j.rineng.2025.107720 eng Results in Engineering. 2025, V. 28, 107720 https://doi.org/10.1016/j.rineng.2025.107720 http://creativecommons.org/licenses/by/4.0/ info:eu-repo/semantics/openAccess Atribución 4.0 Internacional Elsevier