Time stepping schemes in transient thermal analysis

[harshn]

Hi All,

Whenever I try to do a transient analysis on Plane55 element using following code , I see a popup during the iteration on which "Nonlinear Solution" is written. I am assuming my system to be linear, can someone advise me why this "nonlinear Solution" is being shown and is there some way to do time stepping using Crank-Nicholson or backward Euler method. OR how can I know which time stepping scheme is being followed

/PREP7

length=1.0
height=1.0
blc4,0,0,length,

ET,1, PLANE55
MP,Dens,1,2400
mp,c,1,1.94e6/2400
mp,kxx,1,1.8
ESIZE,0.05
AMESH,ALL

FINISH

/SOLU

ANTYPE,TRANS
IC,ALL,TEMP,0
TIMINT,ON,THERM
time,3600
lumpm,0
nsubst,5
neqit,100
autots,off
lnsrch,on
outres,all,all
kbc,1

temperature = 10
NSEL,S,LOC,Y,height
D,ALL,TEMP,temperature
NSEL,S,LOC,Y,0
D,ALL,TEMP,temperature
ALLSEL
Solve

Finish

Regards
Harsh

 

[cbrn]

Hi,
the transient analyses are assumed to be non-linear by definition (otherwise, why perform a transient analysis if all the timesteps are a linear combination of each other?), that's why the appropriate solution schemes are called.
If you know for SURE that the phenomenon is linear, then solve it only for the final state and eventually animate the results (it will give you a "pseudo time-variant" animation...).

Regards

 

[harshn]

Hi,

So, how do I know which time stepping scheme has been follwoed as I need to document it?

Regards

 

[cbrn]

Hi,
following your input log, it's a Newton-Raphson with Line-Search option turned on.

Regards

 

[harshn]

Hi @cbrn,

So what should I do if I want to user Backwar Euler method or Crank-Nicholson method?

Regards
Harsh Nandan

 

[cbrn]

Hi,
precisely, the time integration schemas and the iterative solution methods vary following:
- the type of solution matricial equation (second- or first-order)
- the SOLCONTROL status
- the TINTP settings

Assuming the following:
- your problem is thermal (first-order)
- SOLCONTROL is left to its default (ON)
- TINTP is left to the default,

then the solution procedure is the generalized trapezoidal Huges rule, the transient integration parameter is 1.0 so the time integration scheme is Backward Euler, and the iteration algorithm is Newton-Raphson (for which the options "line search" or "adaptive descent" can be activated as enhancements on the basic NR algorithm).

If you want Crank-Nicholson's unconditional stability, you have either to set SOLCONTROL,OFF or explicitly set TINTP,0.5

See Theory Reference, Chap. 17.2 "Transient Analysis".

Regards

 

[harshn]

Thanks @cbrn I was able to figure out