Optimization of smelting process of magnesium based on machine learning

The traditional magnesium reduction process consumes a significant amount of energy, which contradicts China's green and low-carbon development goals. Therefore, exploring more energy-efficient methods is crucial for environmental protection. The reduction rate of magnesium is influenced by several factors, including gas flow rate, briquetting pressure, ferrosilicon content, reduction temperature, and reduction time. In this manuscript, data analysis utilizing a machine learning algorithm: support vector machine (SVM)-was employed to predict the magnesium reduction rate. Given that energy-saving processes are a primary objective for enterprises, the processing was optimized using the particle swarm optimization (PSO) algorithm based on the SVM model, while maintaining a constant magnesium reduction rate. This optimization aims to reduce energy and gas consumption during the magnesium smelting process. Experimental verification of the magnesium reduction rate under the optimized processing conditions demonstrated that the application of machine learning algorithms can lead to resource savings in the magnesium reduction process. To further evaluate the environmental benefits of the optimized process, a Life Cycle Assessment (LCA) focusing on energy consumption and carbon dioxide (CO2) emissions was conducted. The LCA results indicate that the optimized process significantly reduces the life cycle energy consumption (reduced by 5.33%) and CO2 emissions (reduced by 3.63%) compared with the initial process, providing precise environmental performance data for the promotion and application of magnesium alloys in lightweight structures.