|
楼主 |
发表于 2013-8-28 14:29:11
|
显示全部楼层
硬下心来看了看英文帮助 发现产生这个问题的原因还是有很多 . W5 ]- u9 c* U; c2 n1 m
看来还要继续研究 参数设置要有依据 尽量避免这种问题
d' ~9 c8 B8 e5 O谢谢大家' s& X. c# q( B5 ~
3.9.6. Solution does not converge
$ N4 n4 }; S r# H+ ZThere are several common reasons for a solution not converging. - u. Z( d% r5 U; } T; x. e* x
( Q4 {5 y0 x1 o6 i1.
9 U: \! r; O% J0 A* G4 Z- R; o7 D. n( u- t
The material has a large rigid body motion. Much of the deforming body has a very low strain rate or is rigid.
/ D1 N1 J- i% p+ ~% t8 w9 R& c+ S/ r! ]' N7 m. g( e+ \: E. B) o
2. 2 n) k8 L) r) c! R0 m
# S0 u; v H& p2 r
The material is not strain rate sensitive.
. j5 i* ]! [- R/ e# _/ G' }1 C. F8 P P( @* }* u* r
3. 4 C' q: ]* I. r
9 A i* P* e: ?: NElasto-plastic material is undergoing large deformation or has an inappropriate initial guess.
3 G& {/ k0 R9 J+ ~
' @9 [+ i8 Z9 x1 Y8 j0 j" aIn cases where a problem will not converge, the following checklist should help with the troubleshooting process. This will help with the most common cases. 3 s7 j: F9 x- [9 L
- T; k) [! }7 [
Ÿ Increase your Force Norm or Velocity Norm up to one order of magnitude. The Force Norm may in fact be raised as high as .1 or even eliminated for a few steps. This should not lead to significant error, but could result in reduced accuracy of load calculations. If convergence is improved, allow the simulation to run for 3 or 4 steps, then try reducing the settings to their original values. 4 b' K$ l: D3 q( O+ S3 ]0 @
+ {' o6 n7 ^1 w9 ?. N
Ÿ If the simulation is being run with principal die movement under load or energy control, run a couple of steps under speed control to allow the solution to stabilize before continuing under the original mode.
' E- f2 ~9 a; w7 g; S( G
8 F% V e, p# S. L6 W, dŸ Increase your limiting strain rate to 1/50 or 1/100 of the average strain rate. This should not cause any significant effect on solution accuracy. If you have an extremely difficult case to converge, this value may be lowered to 1/10 of the average strain rate for a few steps, then reset to a more normal value. Over the years, we have recommended that the limiting strain rate should be 1/100 to 1/1000 of the average strain rate. If this value is set too low, it will result in an artificially lower load calculation.
% e; K! Y Y2 X0 f
" l; n `7 t- A' W# K+ ]Ÿ Check your material data versus your process conditions to insure that no "strange" material properties are being passed to the FEM engine. Be particularly aware of extrapolation issues. For example, if your process conditions are in an area that is outside of the defined flow stress region, this "reverse strain rate sensitivity" create a problem that is almost impossible to allow DEFORM to converge on an accurate solution. This may be handled by re-evaluating your raw data and adjusting it as required. Since it is highly unlikely that a material has a lower flow stress at a higher strain rate, the common cause for this type of data is the lack of an adiabatic heating correction. In other words, adiabatic heating at the higher strain rates artificially heated and softened the material causing an apparently lower flow stress. If no clear cause can be determined, find data that does not exhibit this reverse strain rate sensitivity.
: ~! I6 ~7 v9 k. y! x# C9 N4 n
; f* W# Q% _1 a3 n7 BŸ Lower your penalty constant of plastic objects to 250,000 to 500,000 using a constant value (PENVOL). This may lead to volume loss if the value is much lower than 100,000 for typical engineering materials.
4 L' u7 f3 S3 [9 k$ }, s
% u- {3 G. H. W$ z! c! tŸ Reduce your time step. This advice applies particularly for elastic-plastic materials. A very small time step can frequently allow the DEFORM system to get through a tough region of convergence. After a large number of nodes are in contact with dies and the simulation is in progress, a larger time step can be resumed. This may be accomplished through either controlling the time step or a control modifier that will lead to substepping such as DEMAX. 4 E! {/ }( _) w$ D# ~- B7 s
9 t: A; }: @: n# l, f
Ÿ Change the initial guess calculation method. Refer to Object Properties for a discussion of EP Initial Guess. ! H+ B7 b0 k6 c S
5 H l( s' `$ z4 E5 Y* c& }
Ÿ In the case of cold materials that exhibit little to no strain rate sensitivity, this is one of the hardest cases to gain convergence. In fact there should be a very slight strain rate sensitivity even in the cold forging world. A user may help with convergence by creating an artificial strain rate sensitivity. This is not far from reality and may be done by adding a data set of flow stress at a higher strain rate with slightly higher flow stress data. See Figure 11.2 for an idea of how to handle this type of issue.
% k$ g. _2 f( w9 X F' p+ i$ `
& U; n7 M: l, |8 ?. UŸ In a few cases, convergence problems can be caused by a course mesh in an area with high local deformation, such as under the corner of a punch during a piercing operation. In these cases, generate a finer mesh and set the remeshing criteria to have a higher bias towards boundary curvature and strain rate |
|