Sparsity-enhanced optimization for ejector performance prediction

Within a model of the ejector performance prediction, the influence of ejector component efficiencies is critical in the prediction accuracy of the model. In this paper, a unified method is developed based on sparsity-enhanced optimization to determine correlation equations of ejector component efficiencies in order to improve the prediction accuracy of the ejector performance. An ensemble algorithm that combines simulated annealing and gradient descent algorithm is proposed to obtain its global solution for the proposed optimization problem. The ejector performance prediction of a 1-D model in the literature is used as an example to illustrate and validate the proposed method. Tests results reveal that the maximum and average absolute errors for the ejector performance prediction are reduced much more when compared with existing results under the same experimental condition. Furthermore, the results indicate that the ratio of geometric parameters to operating parameters is a key factor affecting the ejector performance.