Artificial neural network modeling for variable area ratio ejector

In this article, machine learning method is applied to model ejectors. Three-layer feed-forward neural network with sigmoid active functions was employed to estimate the outlet pressure of ejector given states of primary and secondary inlets. Well prediction results were achieved within the boundary of training dataset in experiment on ejector based multi-evaporator refrigeration system. The number of hidden layer neurons is optimized by minimizing validation error. Moreover, this research lays the foundation of optimizing system parameters and building control strategies for ejector based refrigeration system based on the machine learning methods.