Design optimization and LSTM-based prediction of phase change release process in graded metal foam tank: role of eccentric active rotation

AbstractTo enhance the heat release process in phase change energy storage (PCES) units, this study introduces a novel approach combining gradient metal foam (MF) structures with eccentric rotation. A horizontal PCES unit embedded with three concentric layers of gradient MF is designed. Numerical simulations are implemented using a local thermal non-equilibrium formulation coupled with the enthalpy-porosity technique. The effects of various pore gradient configurations and unit eccentricity on the liquid fraction, temperature distribution, and heat release characteristics are systematically investigated. Based on the Taguchi method, the optimal configuration is 10 PPI gradient distribution, 200 mm eccentric displacement, and 0.20 rpm rotation speed. Quantitative comparison reveals that this optimized structure reduces solidification duration by 55.04% relative to the baseline Case 14 (Negative pore gradient, 30 PPI gradient distribution, 100 mm eccentric displacement, and 0.05 rpm rotation speed), with a marginal improvement in total energy release but a substantial 123.66% increase in average heat release rate. In addition, a LSTM deep learning model is developed to efficiently predict the time-varying changes of liquid composition, temperature, instantaneous heat release rate, and total released energy. The LSTM-based methodology achieves significant computational economy while maintaining prediction accuracies of 99.99% for liquid fraction, 99.92% for temperature profiles, and 98.84% for total released energy.