In Journal of environmental management
Recycling printed circuit boards (PCBs) in the e-waste stream is essential for ecological protection and metal recycling for a circular economy. Efficient metal recovery from PCBs is highly dependent on the determination of the optimum combination of inputs in the recycling process. In this study, optimization and predictive modelling of the bio-Fenton process were performed employing the response surface methodology (RSM) and the artificial intelligence (AI) models for efficient enzymatic metal bioleaching from discarded cellphone PCBs. The Box-Behnken design (BBD) of RSM was chosen as the design matrix. Further, two AI models, i.e., support vector machine (SVM) and artificial neural network (ANN) were employed to predict complex metal bioleaching process. Experiments were performed based on variations of four input process parameters, namely, glucose oxidase (GOx) content (100-1000 U/L), Fe2+ content (10-50 mM), PCB pulp density (1-10 g/L), and shaking speed (150-450 rpm). Results revealed that the maximum simultaneous enzymatic metal extraction of 100% Cu, 70% Ni, 40% Pb, and 100% Zn was attained at the optimized conditions: GOx content: 300 U/L, Fe2+ content: 10 mM, pulp density: 1 g/L, and shaking speed: 335 rpm. A comparative analysis of the AI models suggested that the ANN-based model predicting more accurate results than the SVM-based model with coefficient of determination values > 0.99 for all the targeted metals. The FTIR analysis confirmed the partial disintegration of PCB polymeric base by OH radicals (OH•), which helped in liberating and exposing the embedded metals to the bio-Fenton solution. Further, the oxidation of metals by ferric ions produced from GOx-mediated oxidation of ferrous ions ensued efficient enzymatic metal bioleaching. Selective metal recovery of >99% was obtained by the chemical precipitation of bioleachate.
Trivedi Amber, Hait Subrata
2022-Nov-21
Artificial neural network, Glucose oxidase, Metal recycling, Printed circuit boards, Response surface methodology, Support vector machine
