Exergo-energetic evaluation and predictive modeling of a double helical coil air heater for sustainable energy applications

Purpose The purpose of this study in residential heating that focuses on passive and surface engineering strategies to produce significant benefits with minimal geometric modifications. The current double helical coil air heater (DHCAH) synergistically integrates geometric modification and the inherent advantages of helical coil designs. Design/methodology/approach This study experimentally investigates the energetic and exergetic performance of a DHCAH under varying geometrical and flow parameters, including helical tube diameter, fluid flow rate, air velocity and inlet temperature. Nusselt number, friction factor, JF factor, effectiveness, exergy efficiency and sustainability index are evaluated. Subsequently, artificial neural network (ANN) and adaptive physics-guided neural network (PGNN) models are developed and assessed using repeated five-fold cross-validation. Unlike the purely data-driven ANN, the PGNN incorporates experimentally derived thermofluid and thermodynamic consistency relationships through adaptive physics-guided loss terms, ensuring physically admissible predictions without directly solving governing partial differential equations. Findings Experimental results show that the tube-side Nusselt number increases with hot-water flow rate, whereas the shell-side Nusselt number decreases. The 6.4 mm helical tube generally provides higher tube-side heat transfer and lower friction factor, while the 12.7 mm tube achieves higher effectiveness and JF factor. ANN and PGNN were evaluated using repeated five-fold cross-validation with four random seeds (20 runs/model). PGNN achieved higher mean R2 for six of seven outputs, while ANN performed better for shell-side Nusselt number. The largest improvement occurred for JF factor, with R2 increasing from 0.6630 to 0.7538. PGNN also generally reduced NRMSE and MAPE. Originality/value An air heater comprising a double helical tube is a novel approach. Experimental work has been carried out to find out the exergo-energetic behavior of the DHCAH. Also, machine learning approaches (ANN and PGNN) have been implemented to predict, validate and compare the synergistic performances.