12. You may try this method which was posted at simwe forum, but not my original idea.
Anyway, I will feel very happy even if it probably just helps you just a little bit.
11.关于数据的输入输出
1.输出数据到 dat 文件:
*NODE PRINT, NSET=nset_name,FREQ=1
COORD
得到的是变形前的坐标还是变形变形后的坐标??
偶在前面已发文问过,偶有时得到的是变形前的,有时得到是变形后的,一直没有弄明白
这个问题。偶想得到的是变形后的坐标
2.
其实 abaqus 自己就带有相关的功能:
abaqus job=job-1 suspend.可以将计算挂起.如果你需要重新进行运算
输入 abaqus job=job-1 resume.
3.
如何输出大量节点的时间历程曲线?
在环境文件 abaqus_v6.env 中添加一句
max_history_requests=0 即可。
12.后处理积分
CAE 自带此功能,比如对速度时程积分得到位移时程,可以这样操作:
利用 XY Data---->ODB History Output 将速度时程存为 V ,然后 XY Data---->Operate on XY
Data---->Operations
看到那个 integrate (X)
这个可能还是只是实现对时间的积分,如果对 dx,dy 的积分
13. 如何实现呢?
比方说,Q=v*A
已经知道流速的矢量分布,怎么得到流量的分布了!
这也是可以的,关键在于你的 x-y data 数据的两轴数据是什么?其积分就是 y 数据值对 x
值的积分!
看看:CAE user manual:33.4.4Overview of X–Y data operations
13.接触分析激活杀死
Usage: Use both of the following options:
*MODEL CHANGE, TYPE=CONTACT PAIR, ADD
surface_1, surface_2
*CONTACT INTERFERENCE
surface_1, surface_2,
Usage: *MODEL CHANGE, TYPE=CONTACT PAIR, REMOVE
14. 利 用 queue 的 功 能 由 本 地 机 器 向 远 程 unix 机 器 提 交
abaqus 作业的方法 [精华 ]
利用 queue 的功能由本地机器向远程 unix 机器提交 abaqus 作业的方法
假设:
14. 1. 远程 unix 机器的安装目录为/abaqus,
2. 远程计算机 IP 地址为 192.168.0.166,主机名为 ibmlinux
3. 本地计算机 IP 地址为 192.168.0.18,主机名为 training
下面的讨论涉及的内容相应改变
Step 1 分别设置本地计算机和远程计算机之间的主机名和 IP 地址对应。
1. 修改本地 hosts 文件, windows 上一般位于 c:windowssystem32driversetc 下, unix 上位
于/etc 下,加入远程计算机主机名和 IP 地址对应行,如:
192.168.0.166 ibmlinux
2. 修改远程 hosts 文件,位于/etc 下,加入本地计算机主机名和 IP 地址对应行,如:
192.168.0.18 training
Step 2 设置本地计算机对远程计算机运行 rsh 和 rcp 命令正确。
1. 两台计算机登陆名和密码一致。举例而言,若远程计算机登陆名和密码为 root/root,则设
置本地计算机也用同样的用户名和密码登陆。
2. 修改远程计算机.rhosts 文件,该文件位于对应用户名登陆后的主目录下,如/home/root,
加入本地计算机的 IP 地址使得本地计算机可以访问到远程计算机。
3. 在本地计算机的提示符下键入
rsh ibmlinux ls
测试 rsh 是否能够正常使用,如果可以列出远程计算机主目录下文件,代表 rsh 和 rcp 工作
正常。
Step 3 设置远程计算机该用户的默认登陆程序为 csh,修改 passwd 文件,位于/etc 下,如:
root:Ty91eFGzybEE2:0:3::/:/usr/bin/csh
25. BTW:请问你的是 abaqus 的那个版本?6.2 or 6.3?6.3 的有盗版了?
: (1),请问如何得到 M,C,K 矩阵?用什么命令?
试一试:*element matrix output
: (2),想要用 Newmark 方法求解.求解问题,在 ABAQUS 中如何实现?或者
: 那里能找到类似的例子?
看看:standard manuals:analysis 的 dynamics analysis 和 explicit manuals
: (3),ABAQUS 中能实现这样的东西吗?就是我需要平滑 ABAQUS 产生的位移场
: 还是这个平滑只能在其他环境中完成?
试试这个输出选项:
*El print,Position=average at Nodes
: abaqus/CAE 好像不能生成节点集,只能定义某个实体,后处理用这个实体上的节点。
: 看到 CAE 的例题都是这样做的。
NODE SET 是 abaqus 的基本功能,也是其方便使用之处,
在 CAE 中是支持的。
这个例子我做过,没有问题。
: : 我用 tool》set》create 居然不能选择由 CAE 生成的节点(不是顶点)
这是因为你没有正确操作 Partition Edge 这一步。
【 在 sunhaolan 的大作中提到: 】
: abaqus/CAE 好像不能生成节点集,只能定义某个实体,后处理用这个实体上的节点。
ABAQUS divides all of its analysis procedures into
two main groups: linear perturbation and general.
ABAQUS makes a very clear distinction between
perturbation and general analysis procedures because
loading conditions and "time" are defined differently
for the two cases. Furthermore, the results from each
type of procedure should be interpreted differently.
26. The response of the model during a general analysis
procedure, known as a general step, may be either
nonlinear or linear. In a step that uses a perturbation
procedure, which is called a perturbation step,
the response can only be linear. ABAQUS treats
such steps as a linear perturbation about the preloaded,
predeformed state (known as the base state) created by
any previous general steps; therefore, its capability
for doing linear simulations is rather more general
than that of a purely linear
analysis program.
各位大侠,我在定义 step 的时候,前三个用的是 standard 中的相关命令.在
step 4 的时候,我想用 explicit 来求解,并且定义了 amplitude,来模拟一个
位移随时间的简单变化,命令如下:
*step name="~~~",nlgeom,inc=1000(好像 explicit 里没有 inc 这个参数)
*dynamic,explicit
,2
*amplitude name=ramp
0,0,1,1,2,1
*boundary amplitude=ramp
refpunch(刚体参照点)2,2,-0.03
*end step
有问题吗?
另外 standard 的命令和 explicit 的命令能出现在同一个.inp 文件中吗?
要转换的.
ABAQUS/Standard and ABAQUS/Explicit are separate program modules with
different data structures; hence, the
explicit dynamics procedure cannot be used in the same analysis as any of
the procedures in ABAQUS/Standard.
However, ABAQUS provides a capability to import a deformed mesh and associated
material state from
ABAQUS/Explicit into ABAQUS/Standard and vice versa. This procedure is
described in ``Transferring results between
ABAQUS/Explicit and ABAQUS/Standard,'' Section 7.6.2 of the ABAQUS/Standard
User's Manual.
31. : 加载步,请问这个 initial 步,主要作用是什么?能不能去掉?
不能去掉,所有的分析都有,是默认的步
: 2、A solid extrusion base feature 这句话是什么意思?
: extrusion、revolution、等是什么意思?
这两的是三维建模时候,在画完二维图形,如何来生成三维图形,
extrusion 意思是你给定一个厚度,然后二维图形第三个方向上面伸展这么多形成三维图形
revolution 意思是你给定一个旋转轴,二维图形绕其旋转后形成三维轴对称图形
: 谢谢!
发信人: fangxj (fxj), 信区: FEA
标 题: Re: abaqus 中接触问题请教
发信站: BBS 水木清华站 (Fri Jun 14 15:51:20 2002)
接触问题不收敛有两个方面不妨试试:
一、在*CONTACT PAIR 里调试 ADJUST 参数;
二、调一些模型参数,比如 FRICTION 等。
【 在 octopuszy (猴哥) 的大作中提到: 】
: 偶的模型中存在两个物体的接触,计算过程中报错:
: ***WARNING: OVERCLOSURE OF CONTACT SURFACES FOUNDATIONOUTSIDE and
: BUCKETOUTSIDE IS TOO SEVERE -- CUTBACK WILL RESULT. YOU MAY WANT
: TO CHANGE THE VALUE OF HCRIT ON THE *CONTACT PAIR OPTION.
: 。。。。。。。。。。。。。
。。。。。。。。。。。。
: ***WARNING: CONVERGENCE JUDGED UNLIKELY. INCREMENT WILL BE
ATTEMPTED AGAIN
: WITH A TIME INCREMENT OF 9.76563E-05
: ***ERROR: TOO MANY ATTEMPTS MADE FOR THIS INCREMENT: ANALYSIS
TERMINATED
: 不知该如何解决?
standard 使用隐式作动力分析,explicit 用显式作动力分析,你该用 explicit。
【 在 uptonow (到目前为止) 的大作中提到: 】
: 请问这两本书有什么不同,
: 我做的是圆形杯子的深冲的模拟过程,我这里只有 standard 一书
: 那位有 explicit 书的大虾请留个言,我想复印,谢谢:)
一般说来,你那样是正确的,如果不行,你可以在
dat,log,msg 中找到中断的增量步,假如为 60,不妨可以
在 res_pulse 中,
38. Since the redefinition of field variables in USDFLD is local to the current
increment (field variables are restored to the values
interpolated from the nodal values at the start of each increment),
any history dependence required to update material properties by
using this subroutine must be introduced with user-defined state variables.
The state variables can be updated in USDFLD and then passed into other user
subroutines that can be called at this material point,
such as those listed above. The number of such state variables is
specified with the *DEPVAR option...
要引入材料的非线性,需要 usdfld;而要引入历史相关,就得要靠 statev 帮忙了!
(我的理解不全面,可以自己参考 manual 中的这一段.版主 g 上吧,我以后忘了看文摘区
就行了
感谢!
现在弄明白了,原来我的文件里多了几行空行,没想到 abaqus 的 inp 文件不允许有空行的。
感谢 wjytj 的帮助!刚开始用 abaqus,太弱了。
【 在 wjytj 的大作中提到: 】
: :*NGEN
: :1,111,11
: :11,121,11
: **此时,12,22,23……110 节点并未生成,所以生成 12,22 以后的节点是没有意义?..
: 看看这样行否?
: *NODE
: 1,0,10
: 11,10,10
: 111,0,0
: 121,10,0
: ...................
: solution-dependent variable
40. 17.请问怎么实现双曲线
subroutine fric(lm,tau,ddtddg,ddtddp,dslip,sed,spd,
1 ddtddt,pnewdt,statev,dgam,taulm,press,dpress,ddpddh,
2 slip,kstep,kinc,time,dtime,noel,ciname,slname,
3 msname,npt,node,npatch,coords,rcoord,drot,temp,
4 predef,nfdir,mcrd,npred,nstatv,chrlngth,props,nprops)
c
include 'aba_param.inc'
c
character*80 ciname,slname,msname
dimension tau(nfdir),ddtddg(nfdir,nfdir),ddtddp(nfdir),
1 dslip(nfdir),ddtddt(nfdir,2),statev(*),
2 dgam(nfdir),taulm(nfdir),slip(nfdir),time(2),
3 coords(mcrd),rcoord(mcrd),drot(2,2),temp(2),
4 predef(2,*),props(nprops)
c
parameter(zero=0.0D0,two=2.0D0)
c
if (lm .eq. 2) return
c
a=props(1)
b=props(2)
c
lm=0
tau(1)=dslip(1)/(a+b*dslip(1))
tau(2)=dslip(2)/(a+b*dslip(2))
ddtddg(1,1)=tau(1)/slip(1)
ddtddg(2,2)=tau(2)/slip(2)
ddtddg(1,2)=zero
ddtddg(2,1)=zero
dslip(1)=dgam(1)
dslip(2)=dgam(2)
return
end
41. 18.取消坐标系等的显示
CAE 进行后处理画的等值线图挺不错,只是如果通过 Print to File 得到的图形质量很差,远
远不如屏幕截图的效果。但通过屏幕截图也有个问题,就是坐标系——特别是有关 ODB 中
当前 step,increment 等不需要的信息也一起给截来了。请问如何关闭这些信息?
Module Visualization-Canvas-Viewport Annotation Options
19.如何在计算中修改材料特性
If you like to use USDFLD, in the subroutine, you can get step number and just use a if to
determine the field variable value due to the step number. Say, step 1 is 1 and step 2 field
variable become 2. In the material definition, just define material properties as a function of
your field variables. Here I use elastic as example. I'm not sure about the grammar since I do
not have manual at hand. You need to check them. something like this:
*elastic, dependencies
2e5,0.3,,1
1e5,0.3,,2
Here I assume your material does not have any temperature dependency and you need to make
sure field variable is the 4th value. Then, in step 1, you have 2e5 as Young's modulus and in step
you have 1e5.
yogayoga,我还是有问题,我修改材料特性后,计算不收敛。请您帮我看看。
在 inp 文件中,按你讲的使用了下面语句定义材料特性:
*elastic, dependencies=1
15e7,0.3,,1
2e10,0.2,,2
然后采用 usdfld 子程序,现将子程序附在下面请您帮我检查检查,并提出宝贵意见,谢谢。
SUBROUTINE USDFLD(FIELD,STATEV,PNEWDT,DIRECT,T,CELENT,
1 TIME,DTIME,CMNAME,ORNAME,NFIELD,NSTATV,NOEL,NPT,LAYER,
2 KSPT,KSTEP,KINC,NDI,NSHR,COORD,JMAC,JMATYP,MATLAYO,LACCFLA)
42. C
INCLUDE 'ABA_PARAM.INC'
C
CHARACTER*80 CMNAME,ORNAME
CHARACTER*3 FLGRAY(15)
DIMENSION FIELD(NFIELD),STATEV(NSTATV),DIRECT(3,3),
1 T(3,3),TIME(2)
DIMENSION ARRAY(15),JARRAY(15),JMAC(*),JMATYP(*),COORD(*)
c
c
FIELD(1)=time(2)
c print*,field(1)
c
c
RETURN
END
First, if you want a sudden change, you can not use total time. Abaqus will linearly interpolate
all the value. If that's what you want, OK. If you want a sudden change at step two, use kstep
instead.
Second, consider the real physical situation: you got equilibrium before you changed material
properties. After you changed material properties, you naturally lose it. Stress need to be
redistributed and changed to get force balance. This is the reason for non convergence.
You have several way to deal with it: if you have static problem and rate effect is minimal, you
can keep the bc and let the model balance itself. The other way is to decrease the increment
length. You can also try damping (you can just apply stabilize directly) or adding some
viscoplastic
or plastic effect. Dynamic analysis will also help to do the trick. Basicly, the method normally you
deal with unstable problem.
改变材料的参数甚至模型,可以通过以下的办法:
(1)在不同的分析之间传递结果
细节请见:ABAQUS Analysis User's Manual /7.7.1 Transferring results between ABAQUS
analysis products: overview
该方法的主要缺点就是建立模型时不能使用多个 parts 或 assembly。
但在某些情况下非常方便,如想在第一步变形的基础上做后续分析。
(2)使用场变量,然后定义材料时指明参数依赖于该变量
如果场变量(不一定要有物理意义)很明确,比如目前讨论的问题,可以定义一个变量,
在
step1 的时候为 1.,step2 为 2.(可以随便定义了):
*STEP,name=step1
43. *STATIC......
*FIELD, VARIABLE=1
NALL, 1.
*END STEP
*STEP,name=step2
*STATIC......
*FIELD, VARIABLE=1
NALL, 2.
*END STEP
NALL 是包含所有节点的集合,可以根据需要改变成欲改变参数的节点集。
然后再定义材料时,指明某个参数依赖于该场变量:
*MATERIAL
*ELASTIC, DEPENDENCIES=1
200.E9, 0.3, , 1.
180.E9, 0.3, , 2.
这样计算到 step1 时,所有节点的场变量 1 为 1.0,abaqus 由此查表确定模量为 200.E9,到
step2,
因为场变量变为 2.,所以模量为 180.E9。
如果场变量连续变化或依赖于其他计算结果,如温度,应变等,可以使用 USDFLD 子程序
(25.2.38
USDFLD
User subroutine to redefine field variables at a material point.
)来连续改变该变量。在定义材料时同样只需要指定一个参数随该变量变化的表 ,
ABAQUS 在计算
时自动插值计算材料参数。
对你的情况,不需要使用该子程序。
If you like to use USDFLD, in the subroutine, you can get step number and just use a if
to determine the field variable value due to the step number. Say, step 1 is 1 and step
2 field variable become 2. In the material definition, just define material properties
as a function of your field variables. Here I use elastic as example. I'm not sure about
the grammar since I do not have manual at hand. You need to check them. something like
this:
*elastic, dependencies
2e5,0.3,,1
1e5,0.3,,2
Here I assume your material does not have any temperature dependency and you need to make
sure field variable is the 4th value. Then, in step 1, you have 2e5 as Young's modulus
44. and in step you have 1e5.
20.输出计算过程中的总质量和总刚度矩阵
element matrix output,elset,file name=user defined,MASS=YES,OUTPUT FILE=USER
DEFINED
21.先张预应力:
先张预应力:
INITIAL CONDITIONS, TYPE=STRESS, REBAR
22.用户子程序的使用
假设你的输入文件为
a.inp
b.for
那么在 ABAQUS command 中的命令应该是这样的:
abaqus job=a user=b
对于 abaqus64pr11,command 中输入:
abq64pr11 job=a user=b
就可以了。
当然首先你要用 cd 命令进入 输入文件 所在的当前文件目录。
强烈建议使用 command 来操作。
子程序文件名后缀应为 .for,而不是 .f
45. 23.怎样设定用双 CPU 机器进行 ABAQUS 计算
try ABAQUS j=your-job CPUS=2
你是什么操作系统系统。我只试过 2000 和 xp。
在上述系统下 abaqus 仅支持 domain 并行计算。CMD 为
abaqus job=? parallel=domains domains=2 cpus=2
直接用 CPUS=2 不行的说。不过 6.41 我没试过。
如果直接用 abaqus job=?
此时看任务管理器,应该 2 个 cpu 都是 50%在跑:explicite。我没试过 stantard 下会如何。
如果在 xp 下,单击进程名 explicite,然后右键。可找到选项分配 cpu。
可以指定进程在那一个下跑或同时在两个上跑。还可以指定优先级。
如果同时算两道题,可选一个进城为低于标准。这样还能干别的事。
如果只是上网、下载就用默认就行了。
我是 2 X xp 1800+的机器。
24.中途停止正在运算的 JOB
If you run Abaqus with interactive mode, you can use Ctrl+C to kill it.
If you run in backgroud, you can use:
abaqus suspend job=jobname to suspend your job;
abaqus resume job=jobname to resume your suspended job;
abaqus terminate job=jobname to terminate your job.
46. 25.自适应网格技术
一个模板:
*HEADING
...
*ELEMENT, TYPE=..., ELSET=ACOUSTIC
Data lines to define acoustic elements
*ELEMENT, TYPE=..., ELSET=SOLID
Data lines to define structural elements
*SURFACE, NAME=TIE_ACOUSTIC
Data lines to define the acoustic surface interface with the structural mesh
*SURFACE, NAME=TIE_SOLID
Data lines to define the solid surface interface with the acoustic mesh
*TIE, NAME=COUPLING
TIE_ACOUSTIC, TIE_SOLID
...
*STEP
*STATIC
*ADAPTIVE MESH, ELSET=ACOUSTIC, MESH SWEEP=10
...
*END STEP
**
*STEP
*STEADY STATE DYNAMICS, DIRECT
...
*END STEP
26.abaqus 计算与内存
求解器会估算内存使用量,ABAQUS 计算时估算的是物理内存加上硬盘共享内存,如果所
需超出二者之和,就停止计算。 如果一次物理内存装不下,就先放在硬盘内存区上,然后逐
步读入物理内存,这就是为啥物理内存扩大后,解算精度可以提高的原因了。 同时隐式不能
超过 3G,显式不能超过 2G ,就像增量步不能超过 1000000 步一样。(硬软件决定的,不可
能无限的)而且,显示、 隐式对 CPU 和内存的要求不同,这要看你算的问题的类型了。 对于
同一个算题,隐式计算,CPU 的作用没有内存大,就是说 CPU 的数量和快慢没有内存的改
变对计算时间的影响大。而显式则不同, CPU 的个数和速度对计算时间有很大的影响。
53. 对于累积变形产生一定数量的小单元的分析过程,采用变比例的方式对这些单元进行质量
缩放,从而限制稳态时间增量的下降。
Input File Usage: Use either of the following options:
*FIXED MASS SCALING, TYPE=BELOW MIN, DT=dt
*VARIABLE MASS SCALING, TYPE=BELOW MIN, DT=dt
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: At beginning of step or
Throughout step, Scale to target time increment of: dt, Scale element
mass: If below minimum target
通过质量缩放使所有单元具有相等的单元稳态时间增量
对所有单元进行质量缩放,致使它们具有相同的稳态时间增量,有效地影响到模型的特征
频谱。由于会引起质量属性的剧烈变化,所以这种方法只适用于准静态分析。并且它意味着
某些单元的比例缩放因子可能会小于 1。
Input File Usage: Use either of the following options:
*FIXED MASS SCALING, TYPE=SET EQUAL DT, DT=dt
*VARIABLE MASS SCALING, TYPE=SET EQUAL DT, DT=dt
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: At beginning of step or
Throughout step, Scale to target time increment of: dt, Scale element
mass: Nonuniformly to equal target
全局质量缩放和局部质量缩放
对单元组指定定比例或变比例的质量缩放,用于对模型的局部区域进行质量缩放。对于指定
的单元组,重复定义质量缩放方法时,局部定义将覆盖全局定义。
Input File Usage: Use either of the following options:
*FIXED MASS SCALING, ELSET=elset
*VARIABLE MASS SCALING, ELSET=elset
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: At beginning of step or
Throughout step, Region: Set: elset
分析步开始时进行质量缩放
55. 将针对所有单元进行变比例缩放。多个变比例质量缩放所涵盖的单元组不能重叠。局部质量
缩放覆盖全局质量缩放。
Input File Usage: *VARIABLE MASS SCALING, DT=dt, TYPE=type, ELSET=elset
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: Throughout step, Scale
to target time increment of: dt
等增量步间隔进行质量缩放
用户可以指定两次质量缩放之间的增量步数。例如,指定频率为 5,表示执行质量缩放的时
刻分别为分析步开始、 5、 10、 15 个增量步。
第 第 第 值得注意的是,选择过小的频率会会增加
计算时间。
Input File Usage: *VARIABLE MASS SCALING, TYPE=type, DT=dt, FREQUENCY=n
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: Throughout step, Scale
to target time increment of: dt, Scale: Every n increments
等时间间隔进行质量缩放
另外,可以指定执行质量缩放的时间间隔。例如,对历时 1.0 秒的分析步指定时间间隔为
5,表示执行质量缩放的时刻分别为分析步开始、0.2 秒、0.4 秒……、1.0 秒。
Input File Usage: *VARIABLE MASS SCALING, TYPE=type, DT=dt, NUMBER
INTERVAL=n
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: Throughout step, Scale
to target time increment of: dt, Scale: At n equal intervals
分析步的开始和过程中采用不同的质量缩放
对于有些情况,理想的做法是在分析步之初采用一种质量缩放方法,而分析过程中进行修
改。
56. Input File Usage: 采用以下两个选项:
*FIXED MASS SCALING, FACTOR=factor, TYPE=type, DT=dt_init
*VARIABLE MASS SCALING, TYPE=type, DT=dt_min,
FREQUENCY=n
or NUMBER INTERVAL=n
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: At beginning of step
Semi-automatic mass scaling, Scale: Throughout step
举例
动态冲击分析中,网格中存在少数尺寸极小或形状极差的单元,这些单元控制了稳态时间
增量。分析步之初,可对这些单元进行质量缩放。此外,冲击导致部分区域内的网格发生严
重扭曲。稳态时间增量可能受冲击区内的单元控制。
实质上,冲击区内的单元相对于刚性表面是稳态的,选择性地使用质量缩放方法可以保证
整个动态响应不受影响。用指定的时间增量对这些单元进行质量缩放,可以有效地地减少计
算时间。
例如,采用定比例质量缩放,指定模型中所有单元的稳态时间增量下限值为 1.0E-6。此外,
采用变比例质量缩放,指定冲击区单元(elset1)的稳态时间下限值为 0.5E-6。这样,分析步开
始时,检查所有单元,如果单元的稳态时间增量小于 1.0E-6,分别对这些单元进行质量缩
放,使之达到 1.0E-6。除单元组 elset1 之外,所有单元在随后的分析过程中保持该质量缩放。
在分析步中,变比例缩放影响单元组 elset1,使其稳态时间增量不小于 0.5E-6。由于分析过
程中只对单元组 elset1 进行了质量缩放,可能会出现整个模型的稳态时间增量小于 0.5E-6
的情况。
多分析步中的质量缩放
从一个分析步转到另一个分析步,质量缩放可以保留,也可以删除,已经缩放过的单元质
量也可以重新初始化。跨分析步应用质量缩放方法,应遵循以下规则:
57. 如果新分析步中没有重新定义变比例质量缩放方法,前一步定义的变比例质量缩放
自动保留。
如果新分析步中没有重新定义定比例质量缩放方法,前一步分析结束时,无论单元
的质量是否经过缩放,都将保留下来。
多步分析中,除了第一步之外,分析步开始时由于使用了质量缩放,单元质量变化较大,
可能会影响到质量的计算精度。当单元质量变化较大时,建议新分析步开始时,先用定比例
缩放的方法把单元质量重新初始化(使之回到原始值),然后再定义必要的质量缩放方法。
删除质量缩放
在当前分析步中定义变比例缩放方法,将删除前一步中所有的变比例缩放。因此,为了将保
留前一步中的变比例缩放,当前分析步中应重新对其进行定义。
Input File Usage: *VARIABLE MASS SCALING
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Semi-automatic mass scaling, Scale: Throughout step
举例
假定在准静态分析的第一步,单元经历变形导致稳态时间增量急剧下降。此外,假定第二步
的变形对稳态时间没有明显影响。
*VARIABLE MASS SCALING, TYPE=BELOW MIN, DT=1.E-5, FREQUENCY=10
*VARIABLE MASS SCALING
第一步,定义每 10 个增量步对模型中所有单元进行一次质量缩放,单元-单元稳态时间增
量为 1 × 10–5。第二步,没有进一步采用质量缩放,沿用第一步经过缩放后的单元质量。
重新初始化
对于第一个分析步之外的其他分析步,默认采用定比例缩放重新初始化单元质量,使之回
到原始值。这样,定比例质量缩放可用于防止缩放后的质量用于新的分析步。这种方法适用
于从准静态分析步(需要进行质量缩放)转到动态分析步(无需进行质量缩放)。
当缩放新分析步中的质量时,可以指定合适的比例因子,或者指定合适的单元-单元稳态
时间增量和缩放类型。当仅对部分单元进行质量缩放时,需定义单元组。
58. Input File Usage: Use the following option to reinitialize the mass of the entire model to its
original value for a new step:
*FIXED MASS SCALING
ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Reinitialize mass
举例
假定某个分析过程依次包括准静态分析和动态分析两步。在准静态分析步中执行质量缩放,
在动态分析步中关闭该功能。
*FIXED MASS SCALING, TYPE=BELOW MIN, DT=1.E-5
*FIXED MASS SCALING
第一个分析步中,定义分析步开始时整个模型的单元-单元稳态时间增量为 1.0 × 10–5.第二
步中采用定比例缩放的方法将质量矩阵重新初始化。
质量缩放适用场合
质量缩放不会影响下列项目:
热-力耦合中的热响应结果
重力载荷,粘滞压力载荷
绝热分析
材料的状态方程
流体单元和流体连接器单元
弹簧和阻尼器单元
与以上项目相关的密度保持不变。质量元、旋转惯量元、无限体和刚性单元也可以进行质量缩
放。然而,由于没有这些单元没有稳态时间增量,所以它们只能通过两种方式进行质量缩放,
一种是用户指定比例缩放因子,另一种是施加非均匀的单元-单元稳态时间增量。如果采用
后者,则至少有一个单元是有稳态时间增量的。壳和梁的旋转惯性元就是基于质量缩放。
60. ABAQUS/CAE Usage: Step module: Create Step: General, Dynamic, Explicit or Dynamic,
Temp-disp, Explicit: Mass scaling: Use scaling definitions below:
Create: Automatic mass scaling, Feed rate: V, Extruded element
length: , Nodes in cross section: n
28.ABAQUS 多处理器进行并行计算的效果研究
环境:
ABAQUS6.3+8 IntelXeon 700MHz CPU+4G Ram+Win2k AdvServer SP3
在命令行模式下,abaqus 命令的下面三个参数进行并行计算的控制:
[cpus=number-of-cpus]
[parallel={loop | domain | supernode | tree}]
[domains=number-of-domains]
ABAQUS/Explicit:
parallel 参数可选 domain,loop
domain 进行拓扑域并行,loop 进行循环级并行(默认)
但在 NT 系统下,不支持 loop 参数
cpus 数要可以整除 domains 数,也就是一个 cpu 可以进行多个 domain 的计算
以 Getting Started with ABAQUS/Explicit
6.5 Example: circuit board drop test
circuit.inp 分析为例(standard_memory = "256 Mb")
1cpu:
abaqus job=circuit int
cup 利用率 100%,运行时间 506s
2cpu:
61. abaqus job=circuit parallel=domain domains=4 cpus=2 int
每个 cup 利用率接近 100%,运行时间 402s
4cpu:
abaqus job=circuit parallel=domain domains=4 cpus=4 int
每个 cup 利用率为 80%左右,运行时间 297s
8cpu:
abaqus job=circuit parallel=domain domains=8 cpus=8 int
每个 cup 利用率为 40%左右,运行时间 364s
ABAQUS/Standard:
parallel 参数可选 supernode,tree
supernode(默认)对单波前进行并行处理,tree 对多波前同时进行并行处理
domains 参数无效
对于线性方程并有稀疏刚度矩阵的模型并行计算有效
以 ABAQUS Release Notes
2.7 Parallel sparse solver
t1-std.inp 分析为例(standard_memory = "1000 Mb")
1cpu:
abaqus j=t1-std int
cup 利用率为 70%左右,运行时间 390s
4cpu (supernode):
abaqus j=t1-std parallel=supernode cpus=4 int
所有系统 cpu 均被使用,每个 cpu 利用率运行过程中不断大幅变化,运行时间 454s
4cpu (tree):
abaqus j=t1-std parallel=tree cpus=4 int
所有系统 cpu 均被使用,每个 cpu 利用率运行过程中不断大幅变化,运行时间 352s
8cpu (supernode):
abaqus j=t1-std parallel=supernode cpus=8 int
每个 cpu 利用率接近 100%,运行 40 多分钟后还无法结束
8cpu (tree):
abaqus j=t1-std parallel=tree cpus=8 int
每个 cpu 利用率接近 90%,运行时间 545s
结论:
多 CPU 并行处理对 Explicit 效果显著,对 Standard 在很多情况下效果不明显,甚至可能
使运算更慢,只使用 Standard 的同学基本可以不用考虑买多 cpu 的机器
62. 29.yahoo 讨论组摘录 --contact+overclosure
8776 Urgent help: overclosure of contact problems
I have a contact problem which the mesh of the master Surface and the slave surface are of the
same size. Acturally the Master nodes and the slave nodes are sharing the same coordinates. They
have the same location but different node numbers. But when i run the abaqus. The warning
message of " the Overclosure of the master surface and slave surface is too much Severe.".
In abaqus manual, one way to slove it is to use " Adjust=0.0 " in the "contact pair" to establish
contact at the Initial configuration?
Or to use " contact interference" ? How to manange contact when master surface and slave
Surface are the same location, but will seperate or penetrate when Loading??
Re
Liang, I had solved a similar problem by increasing the mesh density of the Slave surface by about
2 times and increasing the distance between master and Slave surface so that there is a small gap
between them. I am not sure if you Can do that in your problem but this is an option.
These are the warnings you are seeing. What are the error
messages when it aborts as seen from the message file?
What exactly asre you simulating? Usually your slave
surface (more deformable?) Must be more finely meshed than
the master surface.
5996 Exponential contact pressure-overclosure relationship & separation
I am dealing with an interface where I have defined a
softened interface following an exponential relationship.
It appears that this configuration does not allow the
separation of the two surfaces. Would you know how to
disable this behavior? I have found in the doc how to not
allow separation, but I have here the oposite case, I need
the separation.
Re
You can just toggle off the box for "allow for
separation" :)
Yes, just erase NO SEPARATION on the *CONTACT PAIR
card !! Softened contact DO allow separation, the
particularity is that you will have a contact pressure
even if the surfaces are opened.
5176frction and contact
I have a co-axial cylindrical shell structure in which a
polymeric cylinder has been placed between two steel
cylinders and then pressure is applied on the external
surface of the structure. I am looking for the collapse
pressure of the structure when de-bonding between
63. polymeric layer and steel layers is allowed. So far I have
a model which works with low and moderate friction
coefficient (0.1 to 0.6).
However, to make sure that my model is behaving well, I
decided to run a fully bonded model (layers sharing node
with each other) and then compare the result with a de-
bonded model with high friction coefficient (e.g.
0.9-1.0). I expect that the buckling pressure should be
close.
Now the problem is as friction coefficient is going up
(>0.7), analysis fails without a single increment progress
with messages regarding overclosure and contact opening,
etc. Looking at the ABAQUS manual, it seems that there are
two different methods to solve the problem. One is using
NO SEPARATION contact behaviour and the other one is to
use UNSYMM=NO. Both of them are working in my case when
applied but I do not know which one is the correct one or
the most appropriate one.
My questions are: 1- Is my understanding of the high
friction and the way to avoid convergence problem is
right? 2- How can I decide what should I choose (NO
SEPARATION or UNSYMM=NO)? Any help would be greatly
appreciated.
4473 how to model the two plate contact together?
I model the two steel plates connected, the sides of them
are modeled by connector element, and the contact surface,
I use contact command. But ABAUQUS, give me the warning
below:
***WARNING: THE SYSTEM MATRIX HAS 239 NEGATIVE
EIGENVALUES.
***WARNING: OVERCLOSURE OF CONTACT SURFACES ANGLEPLANE
and GUSSETPLANE IS TOO SEVERE -- CUTBACK WILL RESULT. YOU
MAY WANT TO CHANGE THE VALUE OF HCRIT ON THE *CONTACT PAIR
OPTION. SLAVE NODE 3513 OF CONTACT PAIR
(ANGLEPLANE,GUSSETPLANE) HAS JUST CLOSED. A NEW CONTACT
PATCH NEEDS TO BE DEFINED FOR THE SLAVE NODE.
And I use this commond *Surface Behavior, no separation,
pressure-overclosure=HARD and connenctor element is 3D
Connector element : beam to located at the edge nodes of
two plates I only want to model the two contact surface
deform with each other without friction,They can't
penetrate into each other's inside. Would you give me some
suggestion! Thanks in advance
Re
I am not all that experienced with Abaqus, but I think
64. what the error means is that there are 239 negative
volumes / elemenst in ur model. It might be due to meshing
very close surfaces together that your elements between
the plates have turned inside out. Ur model has to be free
of all the negative elements before you can do any
analysis on it. If u are modeling two simple plates, u
can try meshing them on another meshing program and then
make sure that there is no negative elements in the mesh
before u import it in Abaqus to do any analysis. Guys,
feel free to correct me if I have mistaken.
May be ur initial penetration is too much (modeling) or
ur initial time step is too high. ABAQUS is trying to
resolve overclosure there. Either adjust model so that it
will be "just contact" or "just before contact" distance,
or reduce ur initial time step so that u can avoid
overclosure thingy. But it may be because of normal
issue ? If it it plate contact, I think it will be working
fine with few adjustment...
By the way, how are u dealing with edge line contact ???
I get the same messages when running a contact problem.
Is it realted to ABAQUS CAE, did anyone tried to use the
"Contact Pair" option in ABAQUS Explicit?
If it worked some other way please advice, thanks
**4400contact
I am currently doing some analyses including contacts for
a composite shell. To have a better understanding about
contact, I started by modelling a composite beam
consisting of three layers (Steel/Polymer/Steel). The
steel plates were master surface and polymeric layer was
slave. I was expecting that when I am pressing the steel
plates, the polymeric layer should slide and squeeze out
(the Poisson 抯 effect).
I was using solid elements (brick element). On the contact
surfaces, the slave and mater nodes had the same
coordinate but different node numbers. When I was using
reduced integration point formulation for 1st or 2nd order
elements, the solution did not converge and I had warning
messages regarding numerical singularity and overclosure
suggesting to change HCRIT, etc. Even when I refined the
mesh for slave surface (twice as master surface), the
problem did not disappear until mesh refinement in slave
surface was in a way that the nodes in slave and master
were not 揷 oincide? Furthermore, I had to use full
integration elements rather than reduced integration
elements -even in 2nd order elements- to allow the
analysis converges, which is strange to me. First of all,
my question is why this is a case (mesh refinement in
65. slave surface and avoiding reduced integration point
formulation for the 2nd order elements)? Having succeeded
in analysis of composite beam (although I have problem to
understand why this is a case), I modelled a long shell
(L/D=10) with three layers (Steel/Polymer/Steel), allowing
the same sort of contact describing above. I ended up with
memory shortage (I have modified the pre-memory and
standard memory as high as I could). A simple solution to
memory is to add RAM, but I am hopefully trying to solve
the problem in other ways. Now my second question is, how
one can 搊 ptimise?The size of the model when you are
dealing with contact. Any help will be greatly
appreciated.
Re
It's very hard to define "optimal contact definition" in
general. It depends on many variables...but usually... 1.
Define contact where only contact is occuring. Usually I
run without contact (disp loading) to see where it
penetrates, then grow contact from that region... 2. If it
really is not serious contact (large sliding etc), try to
use simplified version (small sliding, etc).
So, I think the less contact I have it's better
situation... I am not sure about full-integration, reduced
integration shell...I thought ABAQUS is using kinematic
relation type contact (penalty in explicit), surface is
generated based on nodes/surface (master), then keep
calculating relative distance between master/slave
(nodes) etc...not sure where integration point thing comes
in.
More node (refined) will generate smoother
surface..depends on what contact u are using (finite
sliding, small sliding, etc), the number of nodes to check
distance will increase.....
Solution will converge if u use small step, displacement
type loading until it reaches problem points, apply very
small initial increment and check in what instance ABAQUS
is having trouble...(assuming no serious overclosure)..
4337 modeling an arch dam
I'm trying to model an arch dam by CAE scripting. As far
as i've understood, the solid modeling is based on
extrusion operation for 3D solids. But this may not be
applicable for generation of the arch dam blocks. Any
idea on this?
Re
This may be create in others cad/cam/cae & then should be
convert to *.inp for abaqus.
Thanks for your reply. But I think ABAQUS should consider
adding some more features for creating base solid models
66. in next versions. Feature based modeling is quite a strong
tool and I beleive adding further capabilities to it, CAE
can be used as a very strong mesh generator.
Do U have Abaqus 6.3.3?
Yes, truly, all depends on terms. Up to an output of the
new version it is possible to make, probably, only so with
required quality.
4141 Possible contact problem
Hi group, I am running an impact simulation using ABAQUS/
EXPLICIT between an analytical rigid body and a deformable
element based solid. At the start of the step the rigid
body is in contact (no overclosure) with the node based
surface of the element based solid. A force is then
applied to the reference node of the rigid body. I am
interested in the displacement of the surface nodes on
the deformable body, but they are consistently too low in
comparison to the experimental results I have.
I have read of lots of people having contact problems and
was wondering if it is possible that I am having a contact
related problem. I have checked the total contact force
between the surfaces and it seems to be about right. Is
it possible that the contact is not working properly if I
am getting the right contact force between the surfaces?
Re
You haven't said anything about the nature of your
deformable body. Is the real object undergoing plastic
deformation or creep that you are not modeling?
In rereading your original question , I see that you are
doing impact. Creep is probably not an issue and I'm not
sure about plasticity or viscoelasticity. How great is the
discrepancy between FEA and experiment?
Hi Danny, thanks for your reply. Impact occurs over a
short time (~1ms) so as you say creep, viscoelasticity
shouldn't be a factor. I expect a displacement of 2.8mm
based on experiment, and FEA yields a value of 1.5mm.
There are other possible sources of error in the
simulation that I am investigating, but I wanted to
eliminate the possibility that the contact wasn't working
properly.
*4004 simulate micro indentation
I am using ABAQUS 6.2/Standard to simulate micro
indentation procedure. (ABAQUS/Explicit is not available)
At first, I wanna to do a very simple simulation. A
sphere indenter is put into aluminum block material. I set
the indenter as rigid and constrain it fully. Then the
67. Step is static and distributed load on the bottom surface
of the block. Contace pair is the top surface of the block
(slave) and the rigid sphere surface (master).
Unfortunately, I always got the warning information:
overclosure of contact surfaces is too severe. What has
gone wrong? What shall I do to get rid of it and reach the
right solution?
Re
(1) check, if the surfaces do not touch each other
(2) use a small increment for the calculation start
(3) check if your rigid surface definition is o.k.,
sometimes the surface normal points to the wrong
direction, see manuals for right definition
--- > the program computes the distance of the slave
nodes from the master surface and takes into account this
surface normal - so you will get always a overclosure for
separated surfaces if you choose the wrong direction
I have noticed that you are modelling micro-indentation
in ABAQUS. I wonder if you know of any references that can
help me to get started with a micro-indentation model? I
have no idea on how to do it, since all the literature
that I have found says that classical plasticity does not
apply but that I should use strain gradient plasticity
theory, which is not available in ABAQUS. Any help will be
greatly appreciated.
Strain gradient plasticity is just a new kind of theory
to consider the size effect of material. It means the
material properties in small size scale (nano or micro) is
different from large scale. The theory is called to be a
bridge between continuum mechanics and dislocation
theory. It is mainly proposed by Gao Huajian and Nix in
Stanford univeristy and Huang in UIUC. They have done
indentation simulation based on the theory. (check their
papers). The constitutive equation should be changed in
the simulation. I believe it is a quite complicated job. I
am using classical plasticity theory because currently my
scale is much larger. (Dozens of microns)
**3666 CONTACT PROBLEMS
I am facing problem defining the Contact. I must define
whats the Geometry, Its 2D problem, Axis Symmetric. 2
Rigids body are holding Material Sheet, In the CONTACT
Region of Rigid bodies and material, The material is not
allowed to flow when the Deformation takes place, The rest
of material part is facing Fluid Pressure, I hope you
understand its Forming Process.
The problem is overclosure, The Nodes of material body
with in Rigid bodies get the problem and Convergence is
not reached so it aborts. Some body please guide me what
68. options i should use to overcome this problem
Re
Are you applying load or displacement to your rigid
bodies?
*Does your problem start out with an overclosure? If it
is your intention to have the rigid bodies contact the
deformable solid with an initial line-to-line contact,
you may want to use the ADJUST parameter on the *CONTACT
PAIR command.
I am not applying ne force on rigids bodies, they are
just holding the material, the nodes of material that are
under rigid bodies are fixed,
*Good, ABAQUS has trouble resolving contact with load
applied to rigid bodies; applying a displacement to the
rigid bodies rather than load is much easier for ABAQUS to
resolve. As Danny suggested, ADJUST might help. Also, it
will make your message file large, but you can try using
*PRINT, CONTACT=YES to give you a better idea of what is
going on at the contact interface (this will tell you
which nodes are overclosing and by how much).
This is puzzling. Do you mean that:
a) The rigid bodies are controlled by displacements (zero
or non-zero) and
b) The nodes on the surface of the deformable body which
are in contact with the rigid body are fixed????
Then nothing will move and no deformation will occur.
This geometry is almost the same as an example in "ABAQUS
Example Problems Manual - Volume 1", the Example is
CYLINDRICAL CUP DEEP DRAWING.
Now I want a little change in this.
1-- Instead of Punch I want to use the FLUID PRESSURE.
2-- Material held between the HOLDER and DIE should not
be moved, when the deformation takes place meaning only
part of material under FLUID PRESSURE is deformed. I hope
this will clear what i want to do. I will be greatly
looking for the suggestion.
3535 composite structure and contact
I would like to model 搒 liding?Behaviour of middle layer
of a three-layer composite structure. For simplicity,
imaging there is a solid rectangular beam (C3D8RH),
consisting of three layers. The middle layer (core) is
much softer than the exposing (face) layers. Then
structure is going under external pressure from one of the
face layers. In reality, the nodes of core have the same
coordinate than faces (i.e. Sharing the same nodes). But
when you want to study the sliding behaviour of core, you
need to have, I suppose, different series of node sets
(for face and core respectively), but with same
69. coordinates and then allow the contact between them. My
problem is I will get the following warning message in the
.msg file, which it is understandable:
***WARNING: OVERCLOSURE OF CONTACT SURFACES 揂 NN-I?Br>
and 揊 LINE?IS TOO SEVERE -- CUTBACK WILL RESULT. YOU
MAY WANT TO CHANGE THE VALUE OF HCRIT ON THE *CONTACT PAIR
OPTION.
Now if I want to change the HCRIT, it will not be a case
and in the case of 搒 ame coordinates? I do not know how
much it should be. Have anybody crossed such problems? Any
help would be greatly appreciated.
Re
***I have faced similar problems earlier and the
following worked for me:
Use ADJUST parameter to specify the depth of adjustment
zone. This might allow you to avoid changing the HCRIT
value.In most contact problems, overclosure, due to
smaller than required HCRIT, occurs right at the
beginning of the simulation, which means that the slave
nodes are penetrating the master surface in the initial
configuration itself. ADJUST parameter will reposition
the nodes and enable modeling of the sliding behvior with
good accuracy.
Thank you very much for your email. Actually, I have tried
it and still could not perform the analysis successfully.
I also had problem with memory shortage which means that I
should make my model smaller. If it does not make any
problem, could you please forward me the contact part of
input file you have written for your own model to my email
address
3517 Contact Simulation...Severe Cut Back Problem
I am having problem in my simulation my message file is
giving this warning
***WARNING: OVERCLOSURE OF CONTACT SURFACES SA000005 and
SP000009 IS TOO SEVERE -- CUTBACK WILL RESULT. YOU MAY
WANT TO CHANGE THE VALUE OF HCRIT ON THE *CONTACT PAIR
OPTION.
I have reduced the value of HCRIT till 0.0001 m. But
still having the same message. Is it due to the high load?
And how can I get rid of this message as it sometimes stop
after a no. Of iteration step due to this problem.
Waiting...
Re
Check your surface normals. If they are in the wrong
direction, ABAQUS thinks your surfaces have already
penetrated each other.
Thanks Rich for the reply...
My *.dat file shows that the normals have been assumed by
70. ABAQUS. The message is
***WARNING: NODE 4 ON SURFACE SP000009 HAS FACETS WITH
NORMAL VECTORS DIFFERING BY MORE THAN 30 DEGREES.
CONVERGENCE DIFFICULTIES MAY OCCUR AT THIS NODE WITH
FINITE-SLIDING CONTACT. THE NORMAL CONTACT DIRECTION AT
THIS NODE WILL BE (0.0000,0.70622,0.70799).
Any remedial action to avoid this.
I think this topic of Contact Penetration has been
discussed in the previous mails. Try a word search with
contact.
*3504 Values for softened contact
I am considering using the pressure-overclosure=linear
option on the *SURFACE BEHAVIOUR card. I am interested in
using this in conjuntion with the NO SEPARATION parameter
and hopefully to have tension between the contacting
bodies also (I don't think hard contact with no separation
does this?).
My problem is what values to pick for the pressure-
overclosure relationship. There is no particular data I
need to use so is there numbers which people tend to use
in general to get the advantages of softened contact (I
am having problems with hourglassing) and to have tension
available?
Re
I believe that NO SEPARATION parameter can't be used in
conjuntion with the modified surface behavior such as
pressure-overclosure=linear at least in ABAQUS/standard.
My understanding to this is that once contact was
established the slave nodes will be constrained fully to
the master surface.
***3200 Contact overclosure
I have a model with 3D-solid elements (quad. Tetrahedral)
and a static step in abaqus/standard. I defined a contact
between two round surfaces, initially separated. During
the second step the surfaces get in contact with each
other. I allways have an overclosure of the surfaces and
the increments are getting smaller and smaller untill the
analysis stops. I changed master/slave and tried every
value of hcrit, but it didn't work. It's allways the same.
What could I do else?
Re
Perhaps you should use C3D10M instead of C3D10. C3D10M are
designed for contact problems.
Are you using ABAQUS/Standard? And is it a high enough
speed impact? If so, like I had to, Explicit may be a
better option. Just my thoughts,
71. Thanks for your answer, but I use C3D10M elements.
*Yes, I am using abaqus/standart. Can I do in explicit a
static analysis?
--Not really I think, it is only for dynamic analysis.
Sorry, I thought from your previous mail you were running
a dynamic analysis. Didn't read properly!
A couple of suggestions. Try adding command *CONTACT
CONTROLS, AUTOMATIC TOLERANCES in your steps. This helps
to reduce 'chattering' at the contact face (see the manual
for detail. Alternatively change your element type to 4
nodes (especially on the slave surface side) maybe C3D4 ?
*The element choice plays a crucial part in contact
element. Many of the problems like overclosure,
Hourglassing atc can be reduced to a great extent by the
right choice of elements. I would suggest you try to
change the element type and run it once and the try
observing at what point of the second time step is the
overclosure starting. Keep checking for these in between
the solution. Also if the geometry is not too complex try
the brick elements.
**Try to impose a small displacement instead of a load in
order to initialized the contact.
First step: small displacement
Second step: you replace the displacement by the load
Becarfull: do not forget your real boundaries.
Second method: use the parameter APPROACH in *Contact
Controls command
Do you have SMALL SLIDING or finite sliding ? How do you
move your second body? Just a force load ? If this is
true, try a step with a small prescribed motion to close
the contact, then run a second step, remove the boundary
and add your force. This will run. Use *CONTACT CONTROLS,
ANALYSIS=DISCONTINUOUS.
1378 Explicit and Contact
I'm trying to do a simulation of a axisymmetric bolt
thread being formed. I followed an example in the examples
manual as a guide, but I'm still having problems with the
contact.
1) The die is made of a rigid surface.
2) The bolt material starts out as a rectangular slab.
3) The rigid surface is the master surface.
4) Friction is used (mu=0.2).
5) Material of bolt is steel
6) The bolt mat'l is fixed along the line of symmetry.
7) The die moves towards the bolt at a velocity of 1600
mm/s.
72. The problem I encounter is that the forming result is
totally bizarre.
First, I get an initial warning that my overclosure is
1.33 (mm?) And the ABAQUS will compensate for it in the
first step. I think my problem is here. What I end up with
is that the material "moves" to contact the die without the
die moving in the first increment. At least that is what
is looks like to me. I checked it in ABAQUS/POST, and it
does look like the elements first form around the die
(which doesn't move in that increment), and then the when
the die does move, it doesn't deform anything. Any ideas
or suggestions???
Re
What I end up with is that the material "moves" to contact
the die without the die moving
Abaqus post and viewer rescale the graphis to keep things
centered on each increment. I'm guessing this is causing
the illusion of motion.
Have you checked if the outward normal at the contact
surface is in the right sense? Perhaps it is the reason of
having an initial overclosure.
Yep, that was the problem. The model works just fine now.
Thanks for all help. Since i've never used rigid
analytical surfaces, I didn't realize this was an issue.
It is.
*1294 contact overclosure
We have a problem involving too many contacts. Is there
any other option other than ADJUST (*CONTACT PAIR -
keyword) parameter to curb penetration / overclosure. We
understand this ADJUST is to take care of geometry
modelling / meshing errors. While we donot give this the
solution terminates at the beginning itself. Giving ADJUST
= 0.0 makes the solution progress. But we have been told
not to use this parameter. Is there any other way in
place. Thanks in advance.
Re
Between what type of bodies are you modeling contact? The
problem depends to some extent on this. If you are
modeling contact between deformable bodies, you might want
to try the WEIGHT parameter. Also, try adjusting the mesh
size of the elements belonging to master and slave
surfaces.
You might try *CONTACT INTERFERENCE,SHRINK
**593 Couple of Questions
I'm wondering nobody couldn't give me guide(s) about my
questions? Could you please reply me if you can? Soheil
(below is my previous message) Dear Group I'm a phd
73. student at Heriot-Watt University, UK.
I want to model the bending of pipe on rigid circular
surface. I faced with problem in "contact" of pipe which
has been modelled by "S4" element and the rigid surface
which modelled by "R3D4" elemnt. I would like to know:
1- What is the best way for modelling of a 3-D circular
rigid surface (e.g. An exterior surface of a quarter of
cylinder)?
2-What is the command name in ABAQUS for "merging" of
nodes? In other softwares like LUSAS or NISA, the "merge"
command is used. I guess it is MPC in ABAQUS, isn't it?
3-Which parameters should be taken into account when you
want to choose a magnitude for increment, i.e. INC=....?
4- I faced the following WARNING message in .msg file.
What should I do to overcome it? What does it mean "VALUE
OF HCRIT"?
***WARNING: OVERCLOSURE OF CONTACT SURFACES SHELSUR and
RGDSUR IS TOO SEVERE -- CUTBACK WILL RESULT. YOU MAY WANT
TO CHANGE THE VALUE OF HCRIT ON THE *CONTACT PAIR OPTION.
(SHELSURF is positive side of pipe and RGDSUR is rigid
surface).
5-Finally, about print of .odb files, what is the
procedure? I couldn't find any rough described path in
ABAQUS/Viewer manual. I want to know what should I type in
"Print Command" box when you open "print" window. I seek
help of everybody and I do appreciate your help in
advance.
Re
1.Best way of modeling rigid surface is building up the
surface with analytical surface. Analytical one is always
better than numerically discretized one in view of
analysis costs, and accuracy. It's like mises is better
than tresca in terms of easiness of numerical
implementation.
2.It's MPC TIED. It does not merge the nodes as you want,
it simply eliminates dof of specified nodes in stiffness
matrix and makes this node inactive during current
analysis.
3.Im not sure what is the keypoint of this question. If
you want to use user defined time increment , use DIRECT
keyword in *STATIC.
4. This message generally arise when there is large
amount of penetration during current sdi. HCRIT means ...
If there developed contact penetration larger than this
value, there'll be cut-back rather than more iteration.
74. 305 Problems of Overclosure
I am trying to model a machining process, where the
workpiece is clamped onto the machining platform.
Different sections of the workpiece are machined at
different stages and this requires new set of clamps to be
introduced into the model for each stage. The clamps are
idealised by using contact analysis. This is because I do
not want some sections of the workpiece to be rigidly
constrained.
The problem is that I have severe overclosure in the
model for some contact pairs when I activate new contact
pairs in the restart analysis, to introduce the new set of
clamps for machining the next section of the workpiece.
I am not very familiar with contact analysis, but I have
read that we can use commands such as *contact
interference, shrink or *contact pair,adjust to solve such
problems (I was not aware of these command initially).
However, as these comands are defined at the first step of
the analysis, would the overclosure problems be rectified
for the reactivated contact pairs that were previously
inactive in the restart analyses? I would very much
appreciate any other suggestions to solving the problem
too.
Re
The clamps are idealised by using contact analysis. This
is because I do not want some sections of the workpiece to
be rigidly constrained. I'd say dont use contact pairs at
all. You are adding a complicated nonlinearity where all
you want is a stiff-but-not-rigid boundary condition.
Imagine your clamp is jsut a stiff spring and use
*EQUATION.
If you must model the clamping, start with the clamp
slightly away from the part and push it into contact as a
step. Soft contact helps quite alot with convergance as
well.
*5994 assembly parts instance options in ABAQUS
I hope everybody is doing fine. I am using the assembly,
parts instance options in ABAQUS/Explicit to build my
model. I am facing problems at the common interface of two
parts of a model. Does anybody know--how abaqus joins the
two parts or how they define the commmon nodes for two
parts? When I am checking my data file I found that
globally abaqus is definig two nodes at the same location.
And if it is doing that then how the nodes are coneected
between parts? Do I need to define the connection at the
common interface or assembly option is supposed to do it
automatically?
If any body has any idea or any experience with the
75. assembly parts instance options in ABAQUS please, reply,
ASAP.
Re
Abaqus will not join the two instances together. You will
need to define some interaction between the two instances
using interaction module. If u want to merge two parts
then you can merge using assembly module itself. If u want
to merge the nodes of the two instances then you can do
the following, if your part is a geometrical
representation then, mesh it and create input file for
that. Open it using abaqus, which will open it with only
nodes and elements (orphan mesh)...then you can opt for
merging the two orphan mesh part instances in assembly
module (but make sure to you dont merge the nodes)...then
that creates third part with the nodes at the interface
unmerged. Then in part module you can click the node on
the interface you want to merge and then merge it,
Thanks a lot for your reply. But actually i am not using
ABAQUS/CAE. We don't have the license for that. So I am
building my model trough text editior. I am definig the
parts first and meshing them. Each part contains tie
constraints is some regions. The tie constarints are
defined with in each part. Then I am definig the assembly
by positioning the parts using *parts instance option.
When the whole assembly is built I am defing the boundary
conditions and contact regions. I have contact between
steel and concrete at one face. And i am definig that (in
the step data) afetr definig the assembly.
My model is running and there is no error message in any
file. But the only thing is that each part is behaving
individually.
As u have have mentioned about the merging option, how can
I do that by text editor?
The solution is to define a *CONTACT,TIED card between
the two parts. But the stresses in this region won't be
accurate...
584 Couple of Questions about "CONTACT"
I'm a phd student at Heriot-Watt University, UK.
I want to model the bending of pipe on rigid circular surface. I faced with problem in "contact" of
pipe which has been modelled by "S4" element and the rigid surface which modelled by "R3D4"
elemnt.
I would like to know:
1- What is the best way for modelling of a 3-D circular rigid surface (e.g. An exterior surface of
76. a quarter of cylinder)?
2-What is the command name in ABAQUS for "merging" of nodes? In other softwares like
LUSAS or NISA, the "merge" command is used. I guess it is MPC in ABAQUS, isn't it?
3-Which parameters should be taken into account when you want to choose a magnitude for
increment, i.e. INC=....?
4- I faced the following WARNING message in .msg file. What should I do to overcome it? What
does it mean "VALUE OF HCRIT"?
***WARNING: OVERCLOSURE OF CONTACT SURFACES SHELSUR and RGDSUR IS
TOO SEVERE -- CUTBACK WILL RESULT. YOU MAY WANT TO CHANGE THE VALUE
OF HCRIT ON THE *CONTACT PAIR OPTION.
(SHELSURF is positive side of pipe and RGDSUR is rigid surface).
5-Finally, about print of .odb files, what is the procedure? I couldn't find any rough described
path in ABAQUS/Viewer manual.
I seek help of everybody and I do appreciate your help in advance.
30.原创 :无限 元建立 方法, 希望得 到加
分
loveestboy1981
这是我探索出来的关于无限元单元建立的过程,其实无限元建立最难的不是 INP 文件怎么
通过*element 去改变单元类型,而是如何把单元节点转换为与无限元单元节点一致的情况
也 就 是 一 个 单 元 的 最 后 几 个 节 点 要 在 同 一 侧 ( 外 侧 ) 。 www.simwe.com f
以下是我建立的方法,如果能得到各位的同意的话,希望能得到加分肯定。
以一个简单例子作为探讨对象,建立模型,划分网格,通过 mesh--creat mesh part
建 立 网 格 单 元 PART : Part-1-mesh-1, 然 后 定 义 装 配 。 这 些 都 是 基 础 的 步 骤 。
89. Q: 我在计算时 MSG 文件出现如下错误,是不是由于节
点 数 太 多 啦 ? 该 如 何 处 理 这 个 问 题 ?
ERROR: SPECIFIED STANDARD_MEMORY VALUE OF 8000000 IS TOO SMALL TO RUN
THE
NALYSIS. STANDARD_MEMORY MUST BE INCREASED. MINIMUM POSSIBLE VALUE
IS
23477555. LOOK AT MEMORY ESTIMATES SECTION OF .dat FILE FOR FURTHER
INFORMATION ?
A: 对 abaqus_v6.env 文 件 中 的 STANDARD_MEMORY 的 值 进 行 修 改 。
修 改 Site 文 件 夹 下 的 abaqus_v6.env 中 的 配 置 , 如 下 :
#
# System-Wide ABAQUS Environment File
# -------------------------------------
pre_memory = 33554432
standard_memory = 33554432 ##********* 修 改 这 里 ******************
#
# NT specific settings 。
Q: WARNING: THE SYSTEM MATRIX HAS 148
NEGATIVE EIGENVALUES
1 ABAQUS VERSION 6.3-1 DATE 27-NOV-2002 TIME 22:08:00 PAGE..
For use by None user license from HKS Inc. ..
STEP 1 INCREMENT 1 STEP TIME 0.00
STEP 1 S T E A D Y S T A T E S ..
AUTOMATIC TIME CONTROL WITH -
A SUGGESTED INITIAL TIME INCREMENT OF 0.300?
A: 将你的 INITIAL TIME INCREMENT 改小些试试看,不过,一般出现此类问题,多半是
你
的 模 型 有 问 题 , 欠 约 束 或 者 其 它 什 么 的 。
Q: ZERO PIVOT 是 什 么 意 思 ?
A: zero pivot 可 以 理 解 为 刚 度 矩 阵 出 了 问 题 , 例 如 奇 异 。
可能有不同的原因,如: 你所模拟的是软化性质材料,该点因破坏等原因而软化至不
能吃劲; 模型有问题,如约束不够,或者是单元拓扑出错等;或者是你的 UMAT 中写的
[dds
dde] 有 错 ; 。
94. 那要如何操作?首先建立一个圆柱 PART,然后建立 2 种材料,那如何将 2 种材料分配给一
个 P
ART INSTANCE 的不同部分?或者如何将 2 个不同直径的 PART 组装为一个整体?请指教。
would like to answer your question, although I have never played around
Abaqus/cae. Since the geometry of the model is simple (a cylinder), my
solution is you do not have to use /cae, and you manually write your input
file or rewrite the input file generated by abaqus/cae. I guess I am kind of
cheating in answering your question. However, I think some hints could be
helpful.
1) Again, take advantage of the symmetry if possible. If this is an
axisymmetrical problem, you need to model only a plane. Maybe a quarter model?
2) When you define the elements, use "elset" to group two different elements:
such as:
elset, elset=material1, generate
1,100
200,300
elset, elset=material2, generate
101,199
which means the elements from #1 to #100, and from #200 to #300 are one
material, while elements from #101 to #199 are the other.
Then, you could define material properties by
material, name=material1
......
material, name=material2
......
32.[转帖 ]abaqus6.4 导入外来模型的几点小经验!
abaqus6.4 的 CAE 建模功能仍不是太强大,许多情况下,还得借助于第二、第三方软件完成
模型的建立,但在导入过程中,总会或多或少的存在问题,近来做了几个这方面的试验,
有几点发现,解释如下:
1.对于三维模型,pro/e 等 CAD 软件可以建出很好的模型,存成 iges、sat、step 等格式导入
abaqus 可以直接用,个人感觉能用 sat 或 step 格式出现的模型最好不用 iges 文件导入!
2.复杂模型,当导入 abaqus 时,模型可能已经枝离破碎了,无法通过几何修补将模型改好,
而且分网也相当困难,但通过第三方前处理软件 hypermesh,就可以很好的实现模型在导
入时出错的问题,hypermesh 可导入的几何模型格式相当多,一般的 CAD 软件的格式都可
103. 重启动不能改变你的原始分析中的任何参数,也就是说,你的启动点的模型必须和原始分析中
的模型完全一致的,所以不要企图采用 restart 的方法来改变边界条件,材料参数或者网格的密
度等等。这些需要另外的技巧来实现。
37. ABAQUS 的单位心得
Quantity SI SI (mm) US Unit (ft) US Unit (Lxcad)cm
(inch)
Length m mm ft in cm
Force N N lbf lbf N
2
Mass kg 3
tonne (10 kg)吨 slug lbf s /in (102 kg)
Time s s s s s
2 2 2 2
Stress Pa (N/m ) MPa (N/mm ) lbf/ft psi (lbf/in ) N/cm2
Energy J mJ (10–3 J) ft lbf in lbf
Density kg/m3 tonne/mm3(1012) slug/ft3 lbf s2/in4 102 kg/cm3(108)
Ace m/s2 mm/s2 cm/s2
单位:建议采用国际单位制
采用 m、kg、N、s 国际单位制时,重力加速度 9.8m/s2,质量为 kg 密度为 7850 kg/m3
MP,EX,1,210e9
MP,DENS,1,7850
ACEL,9.8
采用 mm、Ton 、N、s 时,重力加速度 9800mm/s2, 密度为 7850e-12 Ton/mm3,应力 MPa
MPDATA,EX,1,,206e3
MPDATA,PRXY,1,,0.3
MPDATA,DENS,1,,7.85e-9
ACEL,0,0,9800,
力=质量*加速度 关系为:N=kg.m/s2=0.001t.1000mm/s2=t.mm/s2
采用 cm、 Mcm、N、s 时,重力加速度 980cm/s2,力为 N 密度为 7850e-8 Mcm/cm3
MPDATA,EX,1,,206e5
MPDATA,PRXY,1,,0.3
MPDATA,DENS,1,,7850e-8