摘要
调控型气体密封作为一种非接触密封,可以提高系统服役过程中运行的稳定性,但是目前密封状态参数的计算方法存在计算模型建立过程繁杂以及迭代计算耗时较长等缺点。本文采用了一种符合智能调控系统输出精度及时效性要求的粒子群算法(PSO)优化下的BP(Back Propagation)神经网络方法,开发了基于PSO-BP的调控型气体密封状态参数的智能调控程序。并且对神经网络模型初始阀值与权值进行取值优化,讨论了粒子群种群数量,隐含层数,神经元数等参数对智能计算程序的影响。搭建了基于PSO-BP的调控型气体密封试验验证系统,验证了密封状态参数智能计算程序的精确度。实现了调控型气体密封的智能调控,提高了调控型气体密封抗干扰能力,促进了大型离心压缩机向宽工况、高参数、高效率和智能化方向发展。
As a non-contact seal,using regulatable gas seal can improve the running stability in the process of service by the introduction of intelligent control system.But the existing calculation method has many shortcomings,such as cumbersome process of establishing the calculation model or consuming time of iterative calculation.This paper proposes a BP(Back Propagation)neural network method optimized by particle swarm algorithm(PSO),which conforms to the requirements of output accuracy and timeliness for the intelligent control system.This paper developed intelligent control procedure based on PSO-BP to control type sealing state parameter.Then optimized thresholds and weight value of neural network,and discussed impact on parameters of intelligent computing procedure such as the particle swarm population,hidden layer,and neurons several.Finally,set up the test system.The test results verified the accuracy of intelligent computing procedure of the regulatable gas seal parameters.The research realizes the intelligent control of RGS,improves anti-interference ability of RGS and contributes the large centrifugal compressor to the direction of wide working condition,high parameters,high efficiency and intelligence.
作者
王磊
李双喜
朱乔峰
李欢
WANG Lei;LI Shuang-xi;ZHU Qiao-feng;LI Huan(Fluid Seal Laboratory of Beijing University of Chemical Technology, Beijing 100029, China)
出处
《流体机械》
CSCD
北大核心
2017年第11期10-16,4,共8页
Fluid Machinery
基金
国家重点基础研究发展计划(973)项目(2012CB026000)
关键词
调控型气体密封
神经网络
智能调控
性能参数
regulatable gas seal
neural network
intelligent control
state parameter
