三元叶轮优化
卢金铃,席 光,祁大同(西安交通大学能源与动力工程学院,710049,西安)5Xq/P#dl2TT$x
摘要:利用反问题计算间接对叶片进行参数化并结合神经网络技术,建立了一种基于三维黏性流动分析的
叶轮优化设计方法.该方法将设计变量分为叶片的设计变量和子午流道的设计变量,并分别作为子系统和中`i z v*a(_ Z;Q;Z'~
心系统.首先,在子系统内保持子午流道形状不变,利用遗传算法对叶片进行优化.然后,在中心系统内利用
试验设计理论安排训练样本,采用神经网络建立叶轮性能与子午流道设计参数之间的响应关系,同时计及叶#r i"\ Ze Yi
片形状对叶轮性能的影响,对子午流道及叶片进行优化.该方法设计变量个数较少,通过将设计变量进行分 L j.k&e2| y&j/S6V;z-W
类并分步优化,有效地减少了计算量,对三元叶轮的优化问题具有良好的应用前景.通过对一台混流泵叶轮-V1n%t DJ;|I:U
进行优化,扬程和效率分别提高了7.43%和3.37%,从而验证了该方法的实用性.z"]ka _ UQx
关键词:子午流道;叶片;优化;神经网络
中图分类号:O357 文献标识码:A 文章编号:0253!987X(2005)09!1021!05{8s E}w2{$U"ph
OptimizationMethodofMeridionalChannelandBladein3!DImpeller"Yah's$C
LuJinling,XiGuang,QiDatong9b(R Gi2y
(SchoolofEnergyandPowerEngineering,Xi'anJiaotongUniversity,Xi'an710049,China)yJ-t b0K'R{$d
Abstract:Animpelleroptimizationmethodbasedonthe3!Dviscousflowanalysiswasproposed,where
thebladewasparameterizedusinginversedesignmethodandneuralnetworkswasadopted.Thedesign,H)C;W(E0NQB \i
variableswereassortedintotwosets:themeridionalchanneldesignvariablesandthebladedesignvariablesthatweretreatedasmainsystemandsubsystem,
respectively.First,inthesubsystem,thebladewas-Z'g:t|RK \
optimizedaccordingtogeneticalgorithmwiththemeridionalchannelremainedconstant.Then,inthemaing)@ q1a!k
system,neuralnetworkwasadoptedtoconstructtheresponserelationbetweenimpellerperformanceand)E9}R/|HH+IA
meridionalchanneldesignvariables,wherethetrainingsampledatawereschemedaccordingtothedesign&^"`3Z;u8y i.z(a
ofexperimentalmethod,andtheeffectofbladeshapeontheimpellerperformancewastakenintoaccount*d Y&y.~'Z$d"V8_
andthemeridionalchannelwasoptimized.Fewerdesignvariablesandlesscalculationeffortwererequired
inthismethodthatmaybewidelyusedintheoptimizationofthree-dimensionimpellers.Anoptimizedimpellerinamixed-
flowpump,wheretheheadandtheefficiencyincreasedby7.43%and3.37%respectively,
confirmedthevalidityofthisnewlyproposedmethod.
Keywords:meridionalchannel;blade;optimization;neuralnetwork 看看那 谢谢了 好东西 楼主辛苦! 我正要找这方面的东西呢。楼主好人啊 刚才怎么没有回复成功
页:
[1]