Author |
Message |
fmpgomes New member Username: fmpgomes
Post Number: 2 Registered: 04-2010
| Posted on Monday, July 05, 2010 - 01:05 pm: | |
Hello, Mr. Nelson. I found a problem in thermo-elasticity problem. When Young Modulus is contant, it's ok. When is a function of time, I have numerical problem, I think. Please, see this example. Where I'm wrong? The example is: SeparateTens-R0.pde Thank you for your attention. Flávio Mamede
|
mgnelson Moderator Username: mgnelson
Post Number: 197 Registered: 07-2007
| Posted on Monday, July 05, 2010 - 03:12 pm: | |
The trouble is likely due to the fact that your time dependent Young Modulus starts at 0 and is very small for many timesteps. This means that your material is liquid or gelatinous like toothpaste... Then gets stiffer over time. We arbitrarily added 1000 to your time expression and the script runs okay. If you really want to model something this soft, you may need to increase the accuracy, using ERRLIM in the SELECT section. i.e. SELECT ERRLIM = 1e-5
|
fmpgomes Junior Member Username: fmpgomes
Post Number: 3 Registered: 04-2010
| Posted on Tuesday, July 06, 2010 - 09:14 am: | |
Thank you for immediatly answer. I've tried like you said and runs just like before. Until four or five hours is ok, after the stress degenerate. Maybe for I'm using student version? Another question, I would like simulate construction in 'layers'. The domain of problem is increased belong the time. I tried using "if" comand for thermical conductivity and other properties, equal zero before a time 'tz1'. Is it 'legal'? Thank you. |
mgnelson Moderator Username: mgnelson
Post Number: 198 Registered: 07-2007
| Posted on Tuesday, July 06, 2010 - 04:19 pm: | |
1) When I run this on the Student version it completes in less than a minute. 2) Simulating construction in layers can be difficult. Moving mesh and staged geometry are two possibilities, but each pose significant hurdles. 3) It is 'legal' to set properties to 0 before some set time, but you have to evaluate the equations and your system to determine if doing so makes sense or not. |
fmpgomes Member Username: fmpgomes
Post Number: 4 Registered: 04-2010
| Posted on Wednesday, July 07, 2010 - 02:13 pm: | |
Sr. Marek, Muito bom mesmo (very good !!!). Thank you so much (BUT I DON'T KNOW WHAT YOU DID !!! Well, this is my work now.) Thank you again
|
mgnelson Moderator Username: mgnelson
Post Number: 199 Registered: 07-2007
| Posted on Wednesday, July 07, 2010 - 06:12 pm: | |
I took your time dependent expression (that was in your comments) and added 1000 : E = 1000*(t-tz2)^0.80/(.080 + .024*(t-tz2)^0.80) +1000
|