Synthesis, stability, thermophysical properties and AI approach for predictive modelling of Fe₃O₄ coated MWCNT hybrid nanofluids

Stability and thermophysical properties of water-based magnetite (Fe₃O₄) material coated on multiwalled carbon nanotubes hybrid nanofluids was investigated. The in-situ growth approach was coupled with the chemical reduction method to make Fe₃O₄ coated multiwalled carbon nanotubes, and X-ray diffraction, vibrating sample magnetometer, and scanning electron microscopy were used to validate these findings. The experiments were conducted for different particle volume loadings (0.05% to 0.3%). Highest stability value of –48 mv was achieved for ϕ = 0.05%. At, ϕ = 0.3% of nanofluid, the thermal conductivity was improved to 13.78%, and 28.33% at temperatures of 20 °C and 60 °C against water. Similarly, at ϕ = 0.3% of hybrid nanofluid, the viscosity has enhanced to 27.83%, and 50% at temperatures of 20 °C and 60 °C against water. Using the experimental data, sensitivity analysis was used to build Multi-Layer Perceptron Artificial Neural Networks (MLP-ANN) with appropriate topologies and training techniques. MLP-ANN was employed to establish the relationship between the inputs (temperature and mixture concentration) and the outputs (density, thermal conductivity, viscosity and, specific heat) for water-based magnetite (Fe₃O₄) material coated on multiwalled carbon nanotubes hybrid nanofluids. The model performances were evaluated using the coefficient of correlation (0.9938–0.9999), coefficient of determination (0.9854–0.9996), root mean squared error (0.0072–0.2626), mean absolute percentage error (0.001%-2.09%), and Nash-Sutcliffe efficiency (0.9856–0.9999). The model's uncertainty was measured with Theil's U2 (0.035–0.267). The results revealed that the MLP-ANN could consistently emulate the experimental testing conditions proficiently, even for diverse temperatures and concentrations, with significant accuracy.