Regression analysis for thermal properties of Al₂O₃/H₂O nanofluid using machine learning techniques

Area of Science:

• Materials Science
• Thermodynamics
• Computational Fluid Dynamics

Background:

• Nanofluids exhibit superior thermal properties compared to conventional fluids, presenting significant potential for diverse engineering applications.
• Accurate determination of nanofluid thermal properties is challenging due to numerous influencing factors and the necessity for extensive experimental testing.

Purpose of the Study:

• To accurately predict the thermal conductivity ratio (TCR) and dynamic viscosity ratio (DVR) of Al2O3/H2O nanofluids.
• To identify key predictor variables influencing nanofluid thermal properties.
• To evaluate the efficacy of Gaussian Process Regression (GPR) for modeling nanofluid behavior.

Main Methods:

• Utilized Gaussian Process Regression (GPR) with specific kernel functions (matern and squared exponential).
• Employed temperature, nanoparticle volume fraction, and nanoparticle size as input variables.
• Trained and validated the model using 222 experimental data sets in MATLAB.

Main Results:

• GPR models achieved high accuracy, with an RMSE of 0.000126 for TCR and 0.000045 for DVR.
• Temperature was identified as the most critical factor for enhancing the thermal conductivity ratio.
• A regression coefficient (R²) of 0.99 indicates high reliability of the predicted results.

Conclusions:

• Gaussian Process Regression provides a highly accurate and reliable method for predicting nanofluid thermal properties.
• Temperature is a crucial parameter influencing the thermal conductivity of Al2O3/H2O nanofluids.
• The developed GPR models offer a valuable tool for optimizing nanofluid applications in engineering.