Regression analysis for thermal properties of Al(2)O(3)/H(2)O nanofluid using machine learning techniques

Nanofluids possess higher thermal properties than the other conventional base fluids. Many investigators suggested that the nanofluids have the potential to apply in various engineering fields. In real time situation it is challenging to determine the thermal conductivity of nanofluids with accuracy as they have many depending factors. Moreover, numerous experimental tests are required to acquire the thermal conductivity of nanofluids accurately. In this research paper, thermal conductivity ratio and dynamic viscosity ratio of Al(2)O(3)/H(2)O nanofluid are predicted accurately by using Gaussian Process Regression (GPR) methods. The input predictor variables used in this model are temperature, volume fraction and size of the nanoparticles. 222 experimental data sets are taken to predict the thermal conductivity ratio (TCR), dynamic viscosity ratio (DVR) and also the effectiveness of the predictor variables in predicting the response variables are extensively studied and found that the temperature is the crucial factor to enhance the thermal conductivity ratio. The proposed modeling is performed by using MATLAB software. The predictions were evaluated by various evaluation criterions. It is observed that an optimized Gaussian process regression (GPR) method with matern kernel function shows an accurate agreement with experimental data with Root Mean Square Error (RMSE) value of 0.000126 for TCR and squared exponential kernel function show good agreement with experimental data with Root Mean Square Error (RMSE) value of 0.000045 for DVR. Regression coefficient value (R(2)) is 0.99; nearer to one hence the predicted results are reliable.