Forecasting the thermal conductivity of a nanofluid using artificial neural networks

In this study, the influence of incorporating MWCNT on the thermal conductivity of paraffin was evaluated numerically. Input variables including mass fraction (0.005-5%) and temperature (25-70 degrees C) were introduced as input and nanofluid thermal conductivity was considered as an output parameter. Thermal conductivity was modeled numerically through two techniques. In the first technique, an algorithm is applied to find the best artificial neural network (ANN) and in the second technique, a response surface methodology (RSM) on data points has been implemented. Eventually, the mean square error, correlation coefficient and maximum margin of deviation in both techniques have been compared. Calculations revealed that the ANN containing hidden layer with six neurons has priority over other ANN. The correlation coefficient for ANN and RSM was 0.993 and 0.972 which imply that ANN method has more accuracy than RSM technique.