An artificial neural network model for predicting the performance of thermoacoustic refrigerators

Abstract

This paper proposes an artificial neural network (ANN) model for predicting the cooling temperature and performance of a thermoacoustic refrigerator (TAR). Two ANN models were presented with two different numbers of neurons in the hidden layers. The results showed that the proposed model can predict the TAR cooling temperature and performance with high accuracy. An R squared value reaching 0.97 can be obtained when using ten neurons in the hidden layer as compared to 0.77 when using five neurons. The use of two hidden layers was sufficient for the highest model accuracy when compared to one or three hidden layers. A feature analysis using the “RReliefF” algorithm was used to investigate the relative importance of operating conditions and geometric parameters to the performance. The operating frequency had the highest importance weights to the performance as compared to the stack geometric parameters. This confirms the results of previous research which showed that the frequency can affect the performance of thermoacoustic refrigerators significantly. The use of ANNs could, therefore, enhance the predictability of TAR performance using different operating conditions and geometric parameters. The feature analysis also shows the important parameters that need further optimisation to be targeted in future TAR designs. The use of ANN models for predicting TAR cooling temperature and performance can also minimise the need for conducting costly and time-consuming experiments.