Three-dimensional graphene network supported nickel-cobalt bimetallic alloy nanocatalyst for hydrogen production by hydrolysis of sodium borohydride and developing of an artificial neural network modeling to forecast hydrogen production rate

Sodium borohydride (NaBH4) is regarded as the most viable chemical for hydrogen pro-duction via hydrolysis thanks to its high theoretical hydrogen content, possible hydrogen evolution even at a low operation temperature, and producing environmentally-friendly products. However, the engineering of a high-performance catalyst is still needed to boost the kinetics of hydrolysis. Herein, nickel and cobalt decorated three-dimensional graphene (Ni-Co@3DG) nanostructure was fabricated via facile production pathway and successfully employed as the catalyst in the NaBH4 hydrolysis reaction for the first time. The influence of the different parameters, including reaction temperature, NaBH4 concentration, and catalyst loading, were examined to determine the optimum operating conditions for efficient hydrogen production. Additionally, this work differed from other works since the performance of the different artificial neural network (ANN) models were evaluated to find out the optimal ANN architecture to forecast the H-2 production rate. The physicochemical characterizations offered the fabricated nanocatalyst had a large specific surface area (885 m(2).g(-1)), and uniformly distributed Ni-Co bimetallic alloys, thereby enhancing the electrochemically active surface area for hydrolysis of NaBH4. The findings proved the superior catalytic activity of Ni-Co@ 3DG towards NaBH4 hydrolysis (initial concentration of 0.5 M) with the hydrogen production rate of 82.65 mmolming.. cat1at 25, and catalyst loading of 0.05 g. The reusability evaluations revealed that the Ni-Co@ 3DG catalyst could retain 95.96% of its initial activity after five successive utilizations. The computational results demonstrated that the best performance metrics were obtained for the single-layer ANN model consisting of 15 neurons in the hidden layer trained using the Bayesian Regulation backpropagation algorithm with the tansig-purelin transfer function combination in the hidden and output layers, respectively. The results demonstrated the ANN forecasted data and experimental results were in accordance, implying the optimized ANN architecture could be utilized for the prediction of the H-2 production rate of the catalyst. (C) 2022 Institution of Chemical Engineers. Published by Elsevier Ltd. All rights reserved.