Deployment of Machine Learning Models for Predicting the Thermal Conductivity of Copper-Infused Damper Oil Nanofluids

Nanofluids, owing to their enhanced thermal conductivity, offer significant potential in diverse heat-transfer applications. However, inconsistencies in reported heat-transfer characteristics and the high cost of experimentation necessitate alternative predictive approaches. This present study employs Machine Learning (ML) techniques such as Multilayer Perceptron Artificial Neural Networks (MLP-ANN) and Optimised Gaussian Process Regression (O-GPR) to predict the thermal conductivity ratio of Cu/damper oil nanofluids. The proposed MLP-ANN model achieved a Root Mean Square Error (RMSE) of 5.868 × 10-5 and a regression coefficient (R2) of 0.99579, using eight neurons in the hidden layer with the trainlm algorithm. The O-GPR model, trained with multiple kernel functions, performed best with a non-isotropic squared exponential kernel, yielding an RMSE of 2.2495 × 10-5 and an R2 of 0.99. Comparison with a mathematical correlation showed percentage error ranges of –1.29 to +1.45 (MLP-ANN), –0.93 to +0.89 (O-GPR), and –1.06 to +1.64 (mathematical model), indicating superior accuracy from the O-GPR model. The findings highlight that O-GPR can reliably predict the thermal conductivity of copper-infused damper oil nanofluids, minimising the need for costly experimental procedures. Major Findings: From the present study, it is found that nanofluid thermal conductivity prediction is feasible using ML models. O-GPR outperforms MLP-ANN in terms of prediction accuracy, and error percentage analysis confirms O-GPR as the most reliable model.