پیش بینی ویسکوزیته برای شش مایع خنک کننده با استفاده از شبکه های عصبی مصنوعی مختلف

The viscosities of six environmentally friendly pure refrigerants with low GWP are predicted based on three artificial neural network (ANN) models: back propagation neural network (BPNN), radial biased function neural network (RBFNN) and adaptive neuro fuzzy interface system (ANFIS). A total of 1089 experimental data are used to train and test the models. Temperature, pressure and density are considered as input variables of networks. The optimal parameters are obtained through the stepwise searching method. The predicted values using the three optimized ANN models with values of experimental data are compared. Moreover, the viscosity of the six refrigerants in saturated liquid state are predicted using all three models in a wide temperature range. The results show that the deviations of almost all data are less than 5.0% and the ANFIS has the best performance.