ANN vs. MARS modeling: Experimental performance enhancement of MgO-TiO2/water binary and TiO2/water mono nanofluids in a plate-fin heat exchanger

This study has experimentally examined the cooling performance/heat transfer rate, effectiveness, and UA product of a PFHE (plate-fin heat exchanger) used on a motorcycle, employing MgO-TiO₂/water binary and TiO₂/water mono nanofluids, in addition to pure water, at various concentrations (0 %, 0.00645 %, 0.0125 %, 0.025 % and 0.05 %), inlet temperatures (70 °C and 80 °C), and flow rates (6.5 LPM, 9.5 LPM and 12.5 LPM). The maximum heat transfer rate, effectiveness, and UA product values observed were 673.868 W, 0.856, and 27.768, respectively, with a 0.025 % concentration of MgO-TiO₂/water binary nanofluid at an inlet temperature of 80 °C and a flow rate of 6.5 LPM. The heat transfer rate, effectiveness, and UA product values for pure water (0 %) under the same conditions were 591.152 W, 0.809, and 23.534, respectively. By applying the acquired data, both ANN and MARS were employed to predict effectiveness, and a comparison was established between the two methods. According to the best result for ANN (T_in = 80 °C and MgO-TiO₂/water), the MARS result indicates an MSE of 1.17 × 10⁻⁵, RMSE of 0.0034, SSE of 0.0002, MAPE of 0.3411, and an R² of 0.9921. By comparison, the ANN results indicate an MSE of 2.52 × 10⁻⁶, RMSE of 0.0016, SSE of 3.78 × 10⁻⁵, MAPE of 0.0876, and an R² of 0.9983. According to the best result for MARS (T_in = 80 °C and TiO₂/water), the ANN result shows an MSE of 6.03 × 10⁻⁵, RMSE of 0.0025, SSE of 9.05 × 10⁻⁵, MAPE of 0.1359, and an R² of 0.9961. On the other hand, the MARS results exhibit a MSE of 3.97 × 10⁻⁶, RMSE of 0.0020, SSE of 5.96 × 10⁻⁵, MAPE of 0.2003, and an R² of 0.9974.