Artificial neural network modeling for steam ejector design. (5th March 2022)

• Record Type: Journal Article
• Authors: Zhang, Kun; Zhang, Zhen; Han, Yuning; Gu, Yinggang; Qiu, Qinggang; Zhu, Xiaojing
• Is Part Of: Applied thermal engineering. Volume 204(2022)
• Publication Date: 2022-03-05
• Subjects: Steam ejector -- Artificial neural network -- Training algorithm -- System stability; Heat engineering -- Periodicals; Heating -- Equipment and supplies -- Periodicals; Periodicals; 621.40205
• DOI: 10.1016/j.applthermaleng.2021.117939
• Languages: English
• ISSNs: 1359-4311

Abstract

Highlights: Three Artificial Neural Network models were applied to ejector design. Levenberg-Marquardt was found to yield the best accordance and least error. ANN reveals superiority on the assistance for ejector design and optimization. Abstract: An artificial neural network (ANN) model for a steam-centered ejector was established and the effect of different training algorithms on the prediction effectiveness of the ANN model was discussed, which found that the ANN model produces better results than the conventional thermodynamic model on the fitting and prediction of experimental data. The Levenberg-Marquardt(LM) trained model yielded the best results among three chosen ANN models, with the experimental accordance improvement of 68% and the prediction error within 15% under given operating conditions. The LM model made the prediction for a steam ejector in a certain system that the outlet area ratio exhibits a smaller effect on the system operation, compared with the entrainment ratio and throat area ratio, which assists to optimize system design and maintain operation stability.