Physics- Informed Neural Network and Experimental Investigations for Analysing Adsorption Kinetics of Desiccant Coated Energy Exchanger Under Tropical Climatic Conditions

Abstract

To improve the energy exchange abilities, and to enhance indoor air quality, desiccant coated energy exchanger (DCEE) is a capable substitute compared to conventional energy exchangers such as fixed beds and desiccant wheels. Thus, in the present study, a novel data-driven modelling methodology utilizing physics-informed neural networks (PINNs) is developed to predict the exit parameters of DCEE during adsorption. The performance characteristics of DCEE are evaluated using PINNs by considering different input and design parameters. Good agreement is obtained between the PINN and experimental results for both the steady-state and transient cases, proving the PINN method's capability in solving multiple physics-based PDEs on a single domain with maximum discrepancy of ±7.8%. The SEM results concluded that a uniform coating is formed on the fin tube. Further, the water vapor adsorption isotherm is evaluated. The experimental analysis of the adsorption kinetics of silica gel shows that the water uptake capability is about 0.35 g.g-1 .