A machine learning based approach for predicting Pool boiling heat transfer coefficient of CNT + GO nanoparticle coated surfaces

Abstract

The use of machine learning in the field of thermal engineering not only enhance the accuracy of predictions but also allows the investigation of parametric effects and the understanding of intricate mechanisms in boiling heat transfer. In this investigation, pool boiling experiment is performed on CNT + GO coated copper surface to generate the dataset. Twelve number of coated samples are prepared and 180 number of datasets is created from the pool boiling experiment. Five different Machine learning algorithms such as Support Vector Regression, Least Square Support Vector Regression, Gaussian Process Regression, Extreme Learning Machine, and Artificial Neural Network is used for modelling purpose. The grid search optimization technique is utilized to fine tune the hyperparameters. The predictive features for heat transfer coefficient encompassed concentration of nano fluid, substrate dipping time, wall superheat, heat flux. Among these algorithms, Gaussian process regression and extreme learning machine exhibited superior performance in accurately predicting the heat transfer coefficient. The statistical assessment of the GPR, including R2, RMSE, MSE, and MAE, yielded values of 0.9998, 0.0009, 0.00, and 0.0023 respectively. Consequently, integrating the GPR and ELM can distinguish hidden patterns and relationships inside experimental data. It can contribute to the advancements in pool boiling research.