Exploring novel heat transfer correlations: Machine learning insights for molten salt heat exchangers

Abstract

The utilization of molten salts in heat transfer applications, specifically within shell-and-tube heat exchangers, has garnered significant attention for its potential in sustainable energy solutions. this study employs advanced machine learning algorithms, including decision tree regressor, support vector regressor, extreme gradient boosting, and random forest, to not only predict the heat transfer behavior of molten salts but also unravel the complex mechanisms underlying this process. Achieving a remarkable accuracy score of 0.985, the Support Vector Regressor leads the predictive models, closely followed by random forest (0.982), Decision Tree Regressor (0.974), and Extreme Gradient Boosting (0.965). The incorporation of Shapley Additive exPlanations values accentuates the Reynolds number’s pivotal role, elucidating a robust correlation with the Nusselt value. These insights transcend mere prediction, offering a profound understanding that can significantly impact the design and optimization of molten salt heat exchangers. The applications of molten salts extend across various sectors, including concentrated solar energy and thermal storage, solidifying their position as a versatile and effective solution in the pursuit of sustainable and efficient energy systems.