化工进展 ›› 2021, Vol. 40 ›› Issue (2): 671-677.DOI: 10.16085/j.issn.1000-6613.2020-0640

• 化工过程与装备 • 上一篇    下一篇

基于BP神经网络的旋风分离器分割粒径模化与预测

刘金鹏1,2(), 赵兵涛1,2(), 钱魏锋1,2, 李会梅1,2   

  1. 1.上海理工大学能源与动力工程学院,上海 200093
    2.上海市动力工程多相流动与传热重点实验室,上海 200093
  • 收稿日期:2020-04-22 修回日期:2020-05-29 出版日期:2021-02-05 发布日期:2021-02-09
  • 通讯作者: 赵兵涛
  • 作者简介:刘金鹏(1996—),男,硕士研究生,研究方向为旋流流动与过程强化。E-mail:381845366@qq.com
  • 基金资助:
    上海市自然科学基金(17ZR1419300)

Modeling and prediction of particle cutoff size of cyclone separator based on BP neural network

Jinpeng LIU1,2(), Bingtao ZHAO1,2(), Weifeng QIAN1,2, Huimei LI1,2   

  1. 1.School of Energy and Power Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
    2.Shanghai Key Laboratory of Multiphase Flow and Heat Transfer in Power Engineering, Shanghai 200093, China
  • Received:2020-04-22 Revised:2020-05-29 Online:2021-02-05 Published:2021-02-09
  • Contact: Bingtao ZHAO

摘要:

为精确建立分割粒径与旋风分离器结构参数和操作参数之间的复杂映射关系,发展了基于数据驱动的BP神经网络(BPNN)的分割粒径模型。使用全局量纲分析,提出环形空间雷诺数、表征旋风分离器本体尺寸影响的量纲为1数和排气芯管插入深度尺寸比作为网络输入参数,表征空气动力等效分割粒径大小的量纲为1尺寸作为网络输出参数,分别确定了训练算法和隐含层神经元个数对BPNN分割粒径模型预测精度的影响。结果表明:贝叶斯正则化算法优于L-M算法和拟牛顿算法,并在隐含层神经元个数为7时达到最优预测性能。与理论模型、半经验模型和多元回归模型进行比较,结果表明,贝叶斯正则化BPNN分割粒径模型展现出了较好的预测能力和泛化性能,模型预测的均方误差为0.136、决定系数为0.975。

关键词: 离心分离, 分割粒径, 神经网络, 贝叶斯正则化

Abstract:

In order to establish the complex relationship between the particle cutoff size and the full dimensions and operation parameters of cyclone separator accurately, a BP neural network (BPNN) model was develop based on data-driven for particle cutoff size. The global dimensional analysis was used, where the annular Reynolds number, the dimensionless number characterizing the impact of cyclone body dimensions, and vortex finder dimension ratio were proposed as network inputs, while the dimensionless dimension which characterized the aerodynamic equivalent particle cutoff size was adopted as network output. The influences on the training algorithm and the number of hidden neurons on model accuracy were determined, respectively. The results showed that the Bayesian regularization algorithm was superior to the L-M algorithm and the quasi-newton algorithm. The optimal prediction performance was achieved when the number of hidden layer neurons is 7. Compared with theoretical model, semi-empirical model and multiple regression model, the Bayesian regularization BPNN particle cutoff size model showed better prediction ability and generalization performance, with the mean square error of the model 0.136 and the coefficient of determination 0.975.

Key words: centrifugation, particle cutoff size, neural network, Bayesian regularization

中图分类号: 

京ICP备12046843号-2;京公网安备 11010102001994号
版权所有 © 《化工进展》编辑部
地址:北京市东城区青年湖南街13号 邮编:100011
电子信箱:hgjz@cip.com.cn
本系统由北京玛格泰克科技发展有限公司设计开发 技术支持:support@magtech.com.cn