Machine learning-based life cycle optimization for the carbon dioxide methanation process: Achieving environmental and productivity efficiency

Abstract

One of the eye-catching procedures for pollution reduction is converting carbon dioxide to valuable materials and chemicals as methane, which is known as carbon dioxide methanation. Although the carbon dioxide hy-drogenation process can help the environment by reducing carbon in the atmosphere, it can also release toxic emissions into the environment, which needs urgent assessment. In this research, the final goal is to determine the best path for this process in order to reach a minimum level of environmental pollution and maximum yield. Based on that, machine learning approaches including artificial neural networks and Bayesian optimization -based support vector machine kernel were employed for the carbon dioxide methanation process with Ni/ Al2O3 catalyst to find the first objective function, and then an environmental model was formulated based on life cycle assessment results in order to generate global warming potential. The multi-objective optimization prob-lem, which was based on four decision parameters (temperature, pressure, hydrogen to carbon ration, and gas velocity), was computed using a genetic algorithm, and decision-making techniques were used for finding the best solution. For the hydrogen/carbon dioxide ratio, the optimal ratio was approximately 4-4.5. Regarding temperature, the range of 340-360 degrees C has been determined, and for the optimal gas velocity, the range of 6600-7000 L/gcat.h has been computed. The best optimal pressure obtained by the CODAS, COPRAS, MOORA, MABAC, SAW, and TOPSIS methods was 7.69 bar, and the optimal pressure value computed by other methods has been around 1-2 bar.