Experimental investigation of cross-flow heat exchangers with helical fins: Performance analysis via RSM and ANN

In this study, the thermal and flow characteristics of cross-flow heat exchangers with helical fins were analyzed experimentally and numerically. The Box-Behnken Design (BBD) was used to examine the effects of air velocity, air inlet temperature, and water flow rate on key performance parameters including Nusselt number (Nu), Reynolds number (Re), friction factor (f), Colburn j factor, and Stanton number (St). Experimental results showed that increasing the Reynolds number improved heat transfer, with Nu increasing by up to 35 % and f decreasing by approximately 70 %. Among the variables, air velocity (x₃) was the most dominant, while water flow rate had a minor effect. Experimental results were also compared with ANSYS Discovery simulations, which revealed a temperature deviation of 15 % and a pressure drop error of 7.9 %, highlighting the limitations of simplified turbulence models. RSM regression models showed high accuracy, especially for Reynolds number (R² = 1.00, p < 10⁻¹²), while models for Nu (R² = 0.899), f (R² = 0.971), and j (R² = 0.940) showed minor deviations due to turbulence-induced nonlinearities. Artificial Neural Networks (ANN) yielded even higher predictive accuracy, particularly for f (R² = 0.9996), Nu (error: 6.6 %), and j (error: 7.3 %), confirming their potential in thermal modeling. Overall, air velocity was the most influential parameter, and the hybrid use of RSM and ANN provided a strong framework for heat exchanger optimization. Future work should focus on AI-based optimization techniques and advanced CFD analysis.