Data-driven analysis of molten-salt nanofluids for specific heat enhancement using unsupervised machine learning methodologies

High specific heat molten-salt is essential for sensible heat thermal energy storage. Current scientific researches focus on Molten-salt nanofluid as a potential solution. However, the causality between system parameters introduced in nanofluid preparation and specific heat enhancement is not clearly understood. Since difficulties are associated with identifying the explicit relations due to complex molecular interactions between molten-salt and nanoparticles, we inquired whether there is a common pattern/clusters in the nanofluid samples reported in earlier studies. The data-driven correlations among samples are explored by employing unsupervised machine learning methods: Hierarchical cluster analysis (HCA) and Principal component analysis (PCA). Three principal components, capturing 81.3% variation of the entire dataset, revealed that the descending order of contribution of the system parameters in the specific heat enhancement percent is concentration, temperature, density ratio, and nanoparticle size. The multivariate clusters emerging from HCA showed the interdependency of density ratio on the temperature, which significantly affects nanofluid's stability at higher concentration, causing a decrease in specific heat enhanced percent. Furthermore, the variation in nanoparticle size was found to have a negligible effect on specific heat enhancement.

Automated summary

This study investigated the relationship between system parameters and specific heat enhancement in molten-salt nanofluids. Machine learning methods revealed the order of influence of different parameters on specific heat enhancement percentage, such as concentration, temperature, and density ratio.