Utilizing deep learning algorithms and artificial neural networks to forecast the viscosity, thermal conductivity, and electrical conductivity of Fe3O4/TiO2 magnetic hybrid nanofluid

Machine learning provides a powerful approach for predicting the complex thermophysical properties of nanofluids. This study employs a suite of machine learning algorithms to forecast the viscosity, thermal conductivity, and electrical conductivity of a Fe₃O₄/TiO₂ magnetic nanofluid, using experimental data over the temperature range of 10-50 °C and volume fractions of 0-0.3%. Among Gaussian Process Regression, Multiple Linear Regression, Support Vector Regression, Multilayer Perceptron, and Multiple Polynomial Regression (MPR), the MPR model demonstrated superior performance, achieving a correlation coefficient above 0.99 and the lowest error metrics (e.g., Root Mean Square Error of 0.0216 for viscosity). Subsequent multi-objective optimization using the Multi-objective Grey Wolf Optimizer (MOGWO) generated a Pareto front of optimal solutions. The most balanced solution, identified using entropy-based weighting, corresponded to a configuration of 60 wolves and 300 iterations. This integrated framework accurately predicts the thermophysical properties and identifies optimal trade-offs for engineering applications.