PREDICTING THE ENERGY PERFORMANCE OF A RECIPROCATING COMPRESSOR USING ARTIFICIAL NEURAL NETWORKS AND PROBABILISTIC NEURAL NETWORKS

This work presents an analysis to predict the energy performance of a reciprocating compressor working with refrigerant R134a using artificial intelligence. The compressor is located in a vapor compression system; tests were experimentally obtained and were used to develop two models: one using an artificial neural network and another one using a probabilistic neural network. Because the relationship between the compressor input variables and the respective output variables is complex, these techniques of the area of artificial intelligence are excellent methods to model this type of compressor. The compressor input Variables were: compressor rotation speed, suction pressure, suction temperature and discharge pressure. The compressor output variables were: mass flow rate, discharge temperature and energy consumption. Computer simulations were performed to train and validate the proposed methods. In order to measure the performance of these methods, the mean squared error was computed fill each experimental test and for each model. The simulations results were used to establish the validity of the models. Finally, the main contribution of this paper is to extend the use of artificial intelligence to predict and simulate the behavior of a reciprocating compressor.