The performance analysis of a variable geometry ejector utilizing CFD and artificial neural network

Abstract

Currently, ejector-based refrigeration applications are attracting much attention as ways of improving refrigeration systems and energy conversion applications. The ejector performance is influenced mainly by the operating conditions and the geometry of the system, so the geometry parameters need to be optimized for optimal performance. However, due to the fluctuating operating conditions no fixed geometry will satisfy the whole range of performance requirements. Thus, a dynamic geometrical ejector is more feasible. A CFD model designed to test the functionality of a Variable Geometry Ejector with a movable spindle and variable nozzle exit position is presented, utilizing R1234yf due to its environmental advantages. In instances where the primary temperature falls below the ideal threshold, the optimal outcome was achieved by positioning the nozzle exit at 0 mm. In contrast, if the primary temperature is equal to or exceeds the optimal level, a negative value of the nozzle exit position can enhance the entrainment ratio, raising the critical temperatures by up to 11.8%. Not utilizing a spindle resulted in a higher entrainment ratio, but it also yielded the smallest stable operating range among the spindle-equipped scenarios. Further, adopting a spindle with a nozzle position of 0 mm led to a significant improvement in critical temperature, increasing it by 25.8%. A Design of Experiment analysis is conducted to show the most influential factors on the primary, and secondary mass flow rates and the entrainment ratio. An artificial neural network that merges the operating conditions and geometrical parameters succeeds in predicting the ejector entrainment ratio with R2 = 99.81%.