Experimental and Artificial Neural Network Predictions of Thermal Efficiency of Flat Plate Collector with SiO₂/Glycerol+Water Nanofluids

The thermophysical characteristics, heat transfer, friction factor, efficiency, and thermal performance factor of flat plate collectors operating with $\mathrm{SiO}_{2} / 30: 70 \%$ glycerol and water mixed nanofluids under thermosyphon conditions are experimentally determined and explained in this research. The artificial neural network-Levenberg-Marquardt technique was used to predict the acquired data. The studies were carried out between 09:00 and 16:30 hours and with volume loadings ranging from 0.25 to $1.0 \%$. This time period is split into two halves for analysis: time duration-1 (09:00-13:00 hours) and time duration-2 (13:00-16:30 hours). At $\mathbf{1. 0 \%}$ vol., the results demonstrate a $20.89 \%$ and $34.55 \%$ increase in thermal conductivity, and viscosity over the base fluid, respectively. The results also show that the maximum improvement was attained for all parameters at mid-day (13:00 hours); for example, the efficiency, heat transfer, friction factor, and Nusselt number increased to $18.49 \%, 48.88 \%, 22.98 \%$, and $53.65 \%$, respectively, over $1.0 \%$ vol. of nanofluid at a Reynolds number of 718.36, compared to base fluid. At $1.0 \%$ vol, the thermal performance was 1.095-times better than the basic fluid. Using the acquired data, the Levenberg-Marquardt algorithm guesses with excellent accuracy. Regression equations were suggested, and the determined $\mathbf{R}^{\mathbf{2}}$ values for the Nusselt number, heat transfer, friction factor, efficiency, and thermal performance were 0.99874, 0.99877, 0.99844, 0.99217, and 0.98231, respectively.