Regression-Based Empirical Modeling of Thermal Conductivity of CuO-Water Nanofluid using Data-Driven Techniques

In this work, an empirical correlation to predict the thermal conductivity of CuO-water nanofluid is developed. The prime novelty of this work is to include the size of the nanoparticles and to utilize the techniques of artificial intelligence on this problem. The experimentation is carried out for the following operating range: working temperature between 302 K to 323 K, particle volume fraction between 0.1 % and 0.4 %, and a particle diameter of 40 nm and 80 nm. The results of the experimentation are benchmarked with the standard properties of water. Afterwards, three different data-driven techniques (SRM, GMDH and ANN) are applied for the correlation development of thermal conductivity. It is reported that GMDH of third polynomial power is the most appropriate yielding an R² of 0.99973, SSE of 2.208834e−06, and MSE of 1.004e−08. Extensive external validation is also carried out on these techniques to ensure the correctness of the methodology. The results of these surrogate models are compared with other models based on their performance indices of regression. Another comparative study has shown that the prediction capability of our proposed regression model has a minimum deviation of ~ 0.35 % and a maximum deviation of ~ 3.7 %.