zengxiaodong 发表于 2015-9-27 21:20:35

ANSYS Emag“现代技术”新单元Plane233线性摄动分析有问题!

本帖最后由 zengxiaodong 于 2017-9-1 13:02 编辑

从Ansys V12开始,增加了一个Plane233单元,这个单元是所谓的“现代技术”单元,而Plane13和Plane53则被称为“传统技术”单元!

传统技术单元:求解力和力矩时,需要施加力标志FLAG;求解微分电感时可以用Lmatrix宏命令(工作点的名义解要用Frontal求解器)

现代技术单元:求解力和力矩时,不需要施加力标志,仅需选中节点和单元,然后用EMFT命令即可;单元不支持Lmatrix宏命令,而用线性摄动分析来求微分电感和绝对通量。

问题是:

在V14.5版本中,并无电磁场线性摄动分析的内容,因此,涉及到微分电感的问题,仍然要使用传统技术的单元!

在V16.2版本中,有了电磁场线性摄动分析的内容,因此,即使涉及到微分电感的问题,也可以使用现代技术的单元!


zengxiaodong 发表于 2015-9-27 21:27:22

本帖最后由 zengxiaodong 于 2015-10-17 18:06 编辑

在V16.2版本中,理论上仍然支持传统技术单元Plane13和Plane53,并且在Lmatrix宏命令之前,不再强调一定要用Frontal求解器来求工作点的解,反而强调用Sparse求解器求解。

但是在用Plane53进行求解时,如果用Sparse求解器,则并不能随后调用Lmatrix宏命令,仍然出现要用Frontal求解器的错误并终止。而如果用Frontal求解器,却出现不收敛的情况,似乎强行要切换到现代技术单元了!



zengxiaodong 发表于 2015-9-27 21:38:55

本帖最后由 zengxiaodong 于 2015-9-28 20:12 编辑

其实,在Ansys V11中,EQSLV命令的手册说明还有Front选项,从V12以后,EQSLV命令手册已经没有了Front选项,但是如果真的输入EQSLV,Front这个命令却一直都还有反应,即使V16.2也是如此!

zengxiaodong 发表于 2015-9-27 21:52:55

本帖最后由 zengxiaodong 于 2015-9-27 21:54 编辑

关于线性摄动分析,参见如下手册









zengxiaodong 发表于 2015-9-28 20:38:14

Plane233单元是否支持速度效应呢?在手册中前后矛盾!

先是说不支持



然后说支持



后面解释也是支持




最后给出了一个实例,用Plane233单元求解了一个速度效应旋转物体

zengxiaodong 发表于 2015-10-17 10:43:02

本帖最后由 zengxiaodong 于 2015-10-17 10:47 编辑

zengxiaodong 发表于 2015-9-27 21:27
在V16.2版本中,理论上仍然支持传统技术单元Plane13和Plane53,并且在Lmatrix宏命令之前,()不再强调一定要用 ...

支持直接装配(direct assembly )的单元,可以用Sparse求解器,然后调用Lmatrix宏命令!

而不支持直接装配的单元,仍然要使用Frontal求解器,得到emat文件和tri文件,然后才能调用Lmatrix宏命令。

但是,Plane53单元不支持直接装配,因此应该使用Frontal求解器!

zengxiaodong 发表于 2015-10-18 08:18:58

本帖最后由 zengxiaodong 于 2015-11-18 23:14 编辑

zengxiaodong 发表于 2015-9-27 21:27
在V16.2版本中,理论上仍然支持传统技术单元Plane13和Plane53,并且在Lmatrix宏命令之前,不再强调一定要用 ...
在Ansys 16.2版本中,用Plane53单元,并且用Frontal求解器时,始终出现不收敛的问题,经反复试探,原来是收敛准则的问题!

过去版本中用CSG收敛准则很好,但是在16.2中就不行,会导致无法收敛错误(其实是已经收敛,迭代结果可用)!

因此,应先删除CSG收敛准则,改用磁位A收敛准则,例如用命令CNVTOL,A,,1e-08即可。

然后,再调用Lmatrix宏命令,就可以求线圈系统的微分电感矩阵和绝对磁链了。:)

zengxiaodong 发表于 2015-10-18 10:19:56

本帖最后由 zengxiaodong 于 2015-10-18 15:55 编辑

同一个永磁电机模型,先用14.5计算空载特性如下(plane53单元,Lmatrix法,CSG收敛准则)







zengxiaodong 发表于 2015-10-18 10:21:04

本帖最后由 zengxiaodong 于 2015-11-17 20:22 编辑

同一个永磁电机模型,再用16.2计算空载特性如下(plane53单元,Lmatrix法,矢量磁位A收敛准则)









可见,除了齿槽转矩波形外,其余自感、互感、磁链的波形几乎一模一样!而齿槽转矩波形的差异可能是迭代精度的不一致导致的,为此改用CNVTOL,A,,1e-08命令重新计算了一次转矩(原来用CNVTOL,A,,1e-05),但是结果是转矩波形无任何改变!{:soso_e114:}

zengxiaodong 发表于 2015-10-18 10:22:40

本帖最后由 zengxiaodong 于 2015-11-17 20:25 编辑

同一个永磁电机模型,最后用16.2计算空载特性如下(用Plane233单元,线性摄动法)



可见,除了磁链波形以外,电感波形几乎已经可以说是面目全非,完全没有人样子了!



zengxiaodong 发表于 2015-10-23 11:02:10

本帖最后由 zengxiaodong 于 2015-10-23 11:14 编辑

电脑中了病毒,暂时无法计算了!

中的是文件加密病毒cryptowall 3.0,每个文件夹下面都会多出来4个文件(文件名:HELP_DECRYPT.HTML、HELP_DECRYPT.PNG、HELP_DECRYPT.TXT、HELP_DECRYPT.URL),要到指定网站付费500美金才能解密,实在太厉害了!

zengxiaodong 发表于 2015-11-14 20:54:28

本帖最后由 zengxiaodong 于 2015-11-14 21:09 编辑

zengxiaodong 发表于 2015-10-23 11:02
电脑中了病毒,暂时无法计算了!

中的是文件加密病毒cryptowall 3.0,每个文件夹下面都会多出来4个文件( ...
电脑格式化重装了,坚决不支付500美金!



下面是用plane233计算的空载齿槽转矩


命令如下:
CMSEL,S,'rotor'
NSLE
ESLN
EMFT
*set,TQ,_tzsum*0.087*8

其中rotor是预先定义好的转子部件,0.087是轴向长度,8是因为模型是1/8

而以前用plane53时的命令:

TORQSUM,'Rotor'
*set,TQ,TORQVW*0.087*8




zengxiaodong 发表于 2015-11-17 20:34:27

上面的对比中,单元分网完全一样,仅仅是将原来模型的plane53单元改成了plane233单元,也就是节点数目和节点位置都是完全一样的!

用lmatrix命令的求解时间为30小时,而线性摄动求解时间为4.6小时,因此提速了6.5倍左右!但是,线性摄动的求解精度实在是太差了!

我可以确信,求解过程没有错误,理由如下:

1、自感、互感、磁链的数值范围均各自两两处于比较接近的程度;

2、自感均为正值,而互感均为负值;

3、磁链波形也属于完全合理的情形。

zengxiaodong 发表于 2015-11-20 18:52:55

本帖最后由 zengxiaodong 于 2015-11-25 22:14 编辑

为了搞清楚问题之所在,加密网格进行计算,原来节点数29587,加密后网格节点数54993,计算时间大幅度增加为12小时,计算的结果如下:




可见,加密网格并无实质性的改善,自感、互感的波形并不是很确定地三相完全一样,而且也不是在一个电周期中完美地变化2次!从这点来说,就可以确定plane233单元用摄动法求电感是存在很大的误差的。



zengxiaodong 发表于 2015-11-21 18:56:49

在有限元中,能量是最精确的物理量,因为整个理论体系就是建立在能量(或余能)最小化的基础之上的,因此,目前16.2版Ansys手册(4楼可下载)中用摄动法求电感的例子一定有错误,后面逐步讨论之!

news 发表于 2015-11-23 07:33:15

:victory::victory::victory::victory::victory:

zengxiaodong 发表于 2015-11-26 14:47:41

zengxiaodong 发表于 2015-10-18 08:18
在Ansys 16.2版本中,用Plane53单元,并且用Frontal求解器时,始终出现不收敛的问题,经反复试探,原来是 ...

又发现一个问题,在使用Plane233单元作摄动分析时,一定要有CSG收敛准则!

如果仅有磁位A收敛准则,在摄动分析后,增量磁能、增量磁共能都是零,因此计算出来的自感、互感、绝对磁链也就都是零,显然这是错误的结果!

zengxiaodong 发表于 2015-11-28 18:32:55

本帖最后由 zengxiaodong 于 2015-12-8 00:22 编辑

按照ansys公司的说明,“传统技术”单元,例如Plane13以及Plane53等等,将来是要逐步淘汰的,取而代之的就是“现代技术”单元,例如Plane233等,但是现在看来,由于Plane233单元实在不给力,一时还不会完全淘汰传统技术的单元!

As technology advances, ANSYS, Inc. continues to develop robust new element types using the most current technologies available. Current-technology elements are more advanced and feature-rich than legacy elements. For example, support for the following capabilities is available only in applicable current-technology elements:
[*]A vast array of material constitutive options (such as anisotropic hyperelasticity, cast iron plasticity, enhanced Drucker-Prager plasticity, Hill plasticity, hyperelasticity, shape memory alloy, plasticity, rate-dependent plasticity, viscoelasticity, and others).
[*]Association of a single element with several material constitutive options (such as a combination of hyperelasticity and viscoelasticity with Prony series expansion).
[*]A curve-fitting tool (TBFT) for calibrating material parameters through experimental data, which currently supports creep, hyperelasticity, and viscoelasticity.
[*]Fracture mechanics parameter calculation (CINT), which uses the domain-integration approach to calculate the J-Integrals for both linear and elastoplastic material behavior at a designated tip location (2-D) or at specific location along the crack front (3-D) through a structure component.
[*]Initial state conditions (INISTATE) that may exist in the structure.
[*]Custom user-defined material models created via the UserMat subroutine.
[*]3-D smeared reinforcing, provided by elements such as REINF265 when used with 3-D solid and shell elements (referred to as the base elements) to achieve the effect of extra reinforcement to those elements.
[*]Nonlinear stabilization, a viscous-based algorithm for improving convergence behavior when instabilities are expected.
[*]Rezoning for solid elements.
[*]Linear perturbation based on nonlinear analysis.
[*]Precise control of the element technologies (ETCONTROL) used in element formulation for applicable elements.
[*]A layered-section option (via SECDATA and other section commands) for shell and solid elements.

In Selecting Elements for Your Analysis, current-technology elements appear first. You can readily identify them, as they are not typically associated with specific material types, nor do they specify restrictions such as “linear.” Most of the elements listed are current-technology elements in the sense that they are the best technologies that ANSYS, Inc. is able to offer at the current product release.A few legacy elements are still supported to meet the needs of longtime users who have input files containing those elements. Legacy elements may eventually be replaced by new elements.
It is good practice to use current-technology elements rather than legacy elements in your analysis wherever possible.







[*]After considering element redundancy and consistency issues, ANSYS, Inc. may in future releases move legacy element documentation to the Feature Archive or undocument those elements altogether; however, the table does not imply that all legacy elements listed are immediate targets for such action.
[*]The table is not a definitive listing of legacy-to-current element equivalents in terms of either formulation or use of shape functions; for example, a suggested current element may require a more refined mesh in some cases, or may require adaptation via appropriate constraints for specific 2-D analyses.
[*]While a given KEYOPT setting can allow you to approximate the behavior of a legacy element, it may not be the most desirable KEYOPT for the current element. For structural-only analyses, try the ETCONTROL command for element and KEYOPT recommendations. For more information, see Automatic Selection of Element Technologies and Formulations.





zengxiaodong 发表于 2015-11-30 17:20:48

本帖最后由 zengxiaodong 于 2015-12-6 08:54 编辑

*set,II,1153873.3374162411843244072844507
*AFUN,Deg
*SET,AA,-35449955*1.16*0.0
*SET,AAA,50
*SET,IA,AA*sin(90+AAA+Ang*16)   
*SET,IB,AA*sin(90+AAA+Ang*16-120)   
*SET,IC,-1.0*(IA+IB)   

ALLSEL

/solu

antyp,static,new
RESCONTROL,define,last,last
BFA,A+,JS,, ,IA,0
BFA,A-,JS,, ,-IA,0
BFA,B+,JS,, ,IB,0
BFA,B-,JS,, ,-IB,0
BFA,C+,JS,, ,IC,0
BFA,C-,JS,, ,-IC,0

Cnvtol,csg,,1.0e-12
eqslv,sparse,,,,keep
SOLVE

finish

/post1
CMSEL,S,'rotor'
NSLE
ESLN
EMFT
*set,TQ,_tzsum*0.087*8

*CFOPEN, 'c:\TQres',txt, , Append
*VWRITE,Ang,TQ
(2e16.8)
*CFCLOS

finish

ALLSEL,ALL

/solu
antyp,static,restart,,,perturb
perturb,stat,TANGENT,current,allkeep
solve,elform

BFADELE,all,JS
BFA,A+,JS,, ,II,0
BFA,A-,JS,, ,-II,0
solve

BFADELE,all,JS
BFA,B+,JS,, ,II,0
BFA,B-,JS,, ,-II,0
solve

BFADELE,all,JS
BFA,C+,JS,, ,II,0
BFA,C-,JS,, ,-II,0
solve

BFADELE,all,JS
BFA,A+,JS,, ,II,0
BFA,A-,JS,, ,-II,0
BFA,B+,JS,, ,II,0
BFA,B-,JS,, ,-II,0
solve

BFADELE,all,JS
BFA,A+,JS,, ,II,0
BFA,A-,JS,, ,-II,0
BFA,C+,JS,, ,II,0
BFA,C-,JS,, ,-II,0
solve

BFADELE,all,JS
BFA,B+,JS,, ,II,0
BFA,B-,JS,, ,-II,0
BFA,C+,JS,, ,II,0
BFA,C-,JS,, ,-II,0
solve

finish

/post1
file,,rstp

set,1,last
etable,_iene,iene
etable,_coen,coen
ssum
*get,iene_1,ssum,,item,_iene
*get,coen_1,ssum,,item,_coen
L1=2*iene_1*8*0.087
PSIA=coen_1*8*0.087

set,2,last
etable,_iene,iene
etable,_coen,coen
ssum
*get,iene_2,ssum,,item,_iene
*get,coen_2,ssum,,item,_coen
L2=2*iene_2*8*0.087
PSIB=coen_2*8*0.087

set,3,last
etable,_iene,iene
etable,_coen,coen
ssum
*get,iene_3,ssum,,item,_iene
*get,coen_3,ssum,,item,_coen
L3=2*iene_3*8*0.087
PSIC=coen_3*8*0.087

set,4,last
etable,_iene,iene
ssum
*get,iene_12,ssum,,item,_iene
L12=(iene_12-iene_1-iene_2)*0.087*8

set,5,last
etable,_iene,iene
ssum
*get,iene_13,ssum,,item,_iene
L13=(iene_13-iene_1-iene_3)*0.087*8

set,6,last
etable,_iene,iene
ssum
*get,iene_23,ssum,,item,_iene
L23=(iene_23-iene_2-iene_3)*0.087*8

*CFOPEN, 'c:\Allres',txt, , Append
*VWRITE,Ang,L1,L2,L3,L12,L13,L23,PSIA,PSIB,PSIC
(10e16.8)
*CFCLOS
finish

zengxiaodong 发表于 2015-12-4 17:28:03

本帖最后由 zengxiaodong 于 2015-12-7 19:32 编辑

解释一下:

II是线圈1A电流时换算出来的电密,此时考虑到了匝数。



*SET,AA,-35449955*1.16*0.0
*SET,AAA,50
*SET,IA,AA*sin(90+AAA+Ang*16)   
*SET,IB,AA*sin(90+AAA+Ang*16-120)   
*SET,IC,-1.0*(IA+IB)
这一段是计算三相电流密度,其中的0.0代表空载,1.0代表额定负载,2.0代表超载1倍!

页: [1] 2 3 4 5 6 7 8
查看完整版本: ANSYS Emag“现代技术”新单元Plane233线性摄动分析有问题!