TY - INPR KW - Science & Technology KW - Life Sciences & Biomedicine KW - Environmental Sciences KW - Environmental Sciences & Ecology KW - E -waste recycling KW - Resource recovery KW - Mathematical optimization KW - Circular economy KW - Machine learning KW - LITHIUM-ION BATTERIES KW - VALUABLE METALS KW - ELECTRONIC-WASTE KW - COMPLEX-MIXTURES KW - MODEL-REDUCTION KW - RECOVERY KW - SIMULATION KW - PYROLYSIS KW - ENERGY KW - COPPER TI - Towards circular economy of wasted printed circuit boards of mobile phones fuelled by machine learning and robust mathematical optimization framework UR - http://dx.doi.org/10.1016/j.rcradv.2024.200226 Y1 - 2024/11// PB - ELSEVIER VL - 23 A1 - Ashraf, Waqar Muhammad A1 - Jadhao, Prashant Ram A1 - Panda, Ramdayal A1 - Pant, Kamal Kishore A1 - Dua, Vivek N2 - Estimating the operating conditions using conventional process analysis techniques for the maximum metal extraction from the wasted printed circuit boards (WPCB) can provide sub-optimal solutions leading to the low yield of the process. In this paper, we present a closed-loop methodological framework built on machine learning and robust mathematical optimization technique, that offers the mathematical rigour, to determine the optimum operating conditions for the maximum Cu and Ni recovery from the WPCB. Alkali leaching based novel metals recovery process from the WPCB is designed, and the experiments are conducted to collect the data on the percentage recovery of Cu and Ni against the operating levels of the process input variables (ammonia concentration (NH3 conc. (g/L)), ammonium sulfate concentration ((NH4)2SO4 conc. (g/L)), H2O2 concentration (H2O2 conc. (M)), time (h), liquid to solid ratio (L/S ratio, (mL/g)), temperature (Temp. (°C)), and stirring speed (rpm)). The experimental data is deployed to construct the functional mapping between the nonlinear output variables of metals recovery process with the hyperdimensional input space through artificial neural network (ANN) based modelling algorithm ? a powerful universal function approximator. Well-predictive ANN models for Cu and Ni recovery are developed having co-efficient of determination (R2) value more than 0.90. Partial derivative-based sensitivity analysis is then carried out to establish the order of the significance of the input variables that is backed by the domain knowledge, thus promotes the interpretability of the trained ANN models. The hybridization of ANN with NLP (nonlinear programming) framework is implemented for the determination of optimized operating conditions to extract maximum Cu and Ni under separate and combined model of metal extraction. The robustness of the determined solutions is verified, the determined optimized solutions for the metal recovery are validated in the lab, and the maximum metal recovery, i.e., 100 % Cu and 90 % Ni is extracted from the WPCB. This research demonstrates the effective utilization of ANN model-based robust optimization approach for the metal recovery from the WPCB that supports the circular economy for the metal extraction industry. JF - Resources, Conservation & Recycling Advances EP - 13 AV - public ID - discovery10196894 SN - 2667-3789 N1 - © 2024 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). ER -