Data-driven Modelling of EHD-Assisted Melting

Solid-liquid phase change materials have poor thermal conductivities that affects the performance of latent heat thermal energy storage systems. One solution is heat transfer enhancement by the induction of electrohydrodynamic (EHD) flow. In this work, we present a data-driven approach to EHD-assisted melting. The geometry is a heated square cavity with a circular electrode at the centre for charge injection. The electric Rayleigh number (T), Rayleigh number (Ra) and Stefan number (St) are fed as inputs to the data-driven setup to analyse the role played by electroconvection, buoyancy-driven convection and pure conduction. A ANN is trained using data from simulations. This ANN captures the relationship between nondimensional numbers and the growth of total liquid fraction. This network is then used to generate a large number of samples for a sensitivity analysis to understand the influence of nondimensional numbers. It is observed that the sensitivity of the melting rates to T and Ra become non-trivial only when the liquid fraction (LF) exceeds 0.2. Further, this threshold LF below which electroconvection does not assist the melting rate remains independent of the buoyancy force strength. Second-order sensitivity indices are significant only for the T and Ra, implying interactions between buoyancy-driven convection and electroconvection