Experimental and Artificial Neural Network Evaluation of Frost Formation on Square Finned Tube under Natural Convection

Abstract

In this study, a test setup is constructed to investigate the frost formation on a square-finned tube under natural convection conditions experimentally. Accordingly, the impacts of relative moisture, fin surfaces, and air temperature on the frost thickness on the first, middle, and last fins are analyzed in detail. Despite the widespread use of square-finned tubes in industrial equipment, few investigations have been conducted on these fins. Due to the obstructed air path, the experimental results show that frost growth starts from the fin tips and not the pipe surface. It is also found that an increase in the mean refrigeration temperature reduces the frost porosity, but increasing relative humidity enhances frost deposition considerably. Various artificial neural networks are developed and tested using experimental data to choose the most accurate model (341 points). The final appropriate model was able to forecast the behavior of frost growth using an input layer, two hidden layers, and an output layer (4-10-10-9). For this model, the statistical error test of coefficient of determination and mean square error gave values of 0.9892 and 0.000297, respectively. Eventually, a series of valuable dimensionless correlations are developed and presented to evaluate the frost thickness on square fins.