引用本文:
【打印本页】   【HTML】 【下载PDF全文】   查看/发表评论  【EndNote】   【RefMan】   【BibTex】
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 718次   下载 81 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于神经网络的预膜雾化模型
王志凯1,2,赵鸿华1,刘逸博1
1.中国航发湖南动力机械研究所,湖南 株洲 412002;2.西北工业大学 动力与能源学院,陕西 西安 710129
摘要:
为了实现对预膜雾化质量的快速预估,基于神经网络建立了以结构参数和工况参数为输入,索太尔平均直径(SMD)值为输出的预膜雾化模型。使用试验数据训练和测试网络模型,结果显示,模型在较宽的工作范围内具有较高的精度和较好的泛化能力。SMD预测值随输入参数的变化规律符合试验结论,可作为给定工况参数下预膜式雾化装置的结构参数优化依据。神经网络能够较好地学习预膜雾化过程中的隐含规律。随着供油压差的增大,预膜器逐渐出现阻雾化效应,导致其雾化效果比单喷嘴和直混式雾化均差。
关键词:  神经网络  预膜雾化  直混式雾化  雾化模型  燃烧室
DOI:10.13675/j.cnki.tjjs.210001
分类号:V231.2
基金项目:
Prefilming Atomization Model Based on Neural Network
WANG Zhi-kai1,2, ZHAO Hong-hua1, LIU Yi-bo1
1.AECC Hunan Aviation Powerplant Research Institute,Zhuzhou 412002,China;2.School of Power and Energy,Northnestern Polytechnical University,Xi’an 710129,China
Abstract:
In order to realize the rapid prediction of the prefilming atomization quality, a prefilming atomization prediction model based on neural network is established with structural design parameters and operation parameters as input parameters, SMD as output parameter. The experimental data are utilized to train and test the network, the results indicate that the model has a high precision and good ability of generalization in wide operating ranges. The variation of predicted SMD with input parameters conforms to the experimental phenomenon, which can be used as the basis of structure optimization of the prefilming atomization device under given operation parameters. The neural network can learn the implicit rules of the atomization process well. Prefilmer effect of resistance atomization appears gradually as the fuel pressure differential increases, which would lead to a poor atomization performance than that of the single injector and direct-mixing atomization.
Key words:  Neural network  Prefilming atomization  Direct-mixing atomization  Atomization model  Combustor