experts help please

[mhn10624]

hi
i am working on modeling tig arc welding of thin aluminum plate (simple but weld). i have done this :
1.writing fortran code -dflux- for moving double ellipsoid heat source
2.create part (a .5 m *.5 m square with thickness .005 m)
3. define material (temperature dependent k,density,Cp,...)
4.difine load,body heat flux
5.predifined condition for temperature (ambient=25)
6.inteaction : convection( -h- dependent on temperature) and radiation - constant
7.step1 : 75 sec(welding time) - heat transfer and step2: 325 sec(cooling time)-heat transfer
8.mesh : standard-heat transfer - linear
9.job : load dflux
10. submmit
heat transfer problem is solved but now what is the problem ?!!
conduction is not sensed by the elements as well as it should!
a node .12 m appart from the weld line should have a maximum temp (200 c) as an experimental work reported.but temperature variation is just 25 c ! also nodes dont sense the temperature of the heat source till it reaches them
a node has constant temp till the time the heat source pass it. if it is on the weld line temp varies good but from .05 m to the end of plate side we have no sensible temp change.is this a basical problem of code ABAQUS or ????
thanks alot

 

[mhn10624]

to mrgolthrope :
thanks for your suggestion. I have 3D model and all dimensions in m . 0.5*0.5*.005 m plate

to corus :
i have read many papers (steel and aluminium).you are right.some of them report conciderable temperature drop from weld line.but we are engineers.we cant tell lie to ourselves!!
we have a plate with k=220 w/m.c in contact with a heat source which melt the aluminium at the middle. now the temp. of a point at the side of plate should change from 25 to 50 ! unacceptable.
my friend who used abaqus told me that he had such a problem and he used t/tmax and it was solved.but here it doesnt work.i dont have time to learn Ansys and Abaqus doesnt answer.i dont know what to do?
i want to solve mechanical problem (for residual stress) with data achieved from thermal one(temp. distribution).wrong temp. dist. will result in wrong stresses.

 

[corus]

Check the results from your 'engineer' who made the measreuments. I think he/she has mistaken 12 mm for 0.12m. From your results plot your temperatures at that point (using XYdata/create/field output) and you should see the maximum temperature fit your measurements. The overall distribution with time still appears different though. That may be due to the maximum temperature at the weld not reaching melting point using your heat source, and latent heat effects not being seen. It is noted though that your measurements aren't particularly smooth and temperatures appear to go up and down for no apparent reason. I'd be wondering why they vary so much and how valid/reliable are your measurements.

 

[mhn10624]

thanks a lot corus for your responses
the dimension .12 m is correct ...
did you checked the dflux code? was any problems there?
if we make a,b,cr,cf so small,temp. on weldline goes up (700 or more). it is more than melting point.but just centerline senses this variation.temp. on weldline goes up but at sides no .
consider the results of experiment are wrong.
what should we expect?temp. rise at sides is expected.
forget my code and modeling.think about what should happen in such a case in reality and what does happen in abaqus modeling?!!!
believe me it is a basic question !
thanks again and waiting for any idea from anybody!

 

[mhn10624]

i used data for steel 304l from an article. temperature variation at points 3mm , 8mm , ... , 23 mm was checked in that article for plat 200mm*200mm*3mm .
i validated the method in this points by changing a,b,cr,cf of heat source (which are near weld line)
but when i check temp at points more than 40 mm from weldline, we dont see any sensible temp. change
maybe this method of simulating welding with abaqus is just effective in HAZ area?!!
any other methods available?does it mean that ansys is better?!

 

[corus]

In an earlier post I pasted a link to a paper on welding aluminimum plate where it describes a steady state heat transfer solution, relative to the movement of the arc weld. In that paper it describes reasons why there are dispcrepancies between measured and predicted temperatures, eg. the ambient temperature increases. The paper is remarkably similar to the work you are doing so it might be an idea to compare your model with their measurements, and their finite difference model results. You probably just have to change the values in dflux1 to match their results.

 

[mhn10624]

yes
i checked that
steady-state has no problem (effect of density and Cp is ignored and everything will be ok-heat is conducted through plate with no resistanse)
thanks a lot again
i will check this topic
if you found something, i will be glad (so glad! ) to know ...

 

[corus]

No, the solution in the paper is a quasi-staedy state solution which does still depend on the diffusivity (k/pCp) of the material and as such isn't pure conductivity. The equation is different to that used by finite element theory for heat flow, and you'd need to describe a new element stiffness matrix, I think, in order to solve it. As it stands no finite element program for heat flow would solve that equation. As I said it'd be worth checking their solution with your transient heat flow model and compare your solution, say when the weld is half way across the plate. Note that they describe a surface heat flux as opposed to your body heat flux condition.

 

[mhn10624]

is it possible to solve quasi-steady state problem with abaqus?
if i solve the problem in steady-state manner( in "step" , mark "steady state" instead of "transient"), density and Cp are ignored and it soves it as a complete steady problem([k]{T}=0)
how is it possible to solve this form : [alpha=k/pCp]{delT}+{Tx}=0 ??