Determination of Condensation Heat Transfer Characteristics of R134A by Means of Artificial Intelligence Method

The present study investigates the best artificial neural network (ANN) approach to estimate the measured convective heat transfer coefficient of R134a flowing downward inside a vertical smooth copper tube having an inner diameter of 8.1mm and a length of 500mm during annular flow numerically. R134a and water are used as working fluids in the tube side and annular side of a double tube heat exchanger, respectively. Experimental data, used as the ANN training set, came from intube condensation tests including three different mass fluxes of R134a such as 260, 340 and 456 kg m⁻²s⁻¹, two different saturation temperatures of R134a such as 40 and 50 °C and heat fluxes ranging from 10.83 to 50.89 kW m⁻². Accuracy of the dataset was proven in many papers in the literature. The quality of the refrigerant in the test section is calculated considering the temperature and pressure obtained from the experiment. The pressure drop across the test section is directly measured by a differential pressure transducer. Measured values of test section such as mass flux, heat flux, the temperature difference between the tube wall and saturation temperature, average vapor quality are assigned as input of the ANNs, while the experimental condensation heat transfer coefficient and measured pressure drop are specified as the output in the analysis. The artificial neural network (ANN) methods of multi-layer perceptron (MLP), radial basis networks (RBFN), generalized regression neural network (GRNN) and adaptive neuro-fuzzy inference system (ANFIS) were used to decide the best approach for modeling condensation heat transfer characteristics of R134a. 183 data points obtained in the experiments are divided into two sets randomly. Sets of test and training/validation are including 33 and 120/30 data points respectively. In training phase, 5-fold cross validation is used for determine the best value of ANNs control parameters. The ANNs performances were measured by mean relative error criteria with the usage of unknown test sets. The performance of the method of multi layer perceptron (MLP) with 5-13-1 architecture and radial basis function networks (RBFN) with the spread coefficient (sp) of 40000 were found to be superior to other methods and architectures by means of satisfactory results with their deviations within the range of ±0.58% for the estimated condensation heat transfer coefficient and ±1.74% for the estimated pressure drop respectively.