Stuck? 
support@mecway.com
support@mecway.com
Mecway
FEA
convergence tolerances
I am interested to know if any one has experimented with convergence tolerances, I see articles like the one below about abaqus and wonder if the same is true of Calculix. By introducing some changes in the *CONTROLS, PARAMETERS=FIELD parameters I have have some success making complex models converge that have other wise have not done so.
However I am way out of my depth and its pretty hit and miss. I totally understand that messing in this area could seriously affect the solution and its clear any results generated with changes in these parameters need to be treated with even more care than norml. I normally would see this as a last ditch effort to get a solution but I can see other uses. The potential benefits are making solutions converge where they have not done without modifying these parameters and speeding up runs where lots of expletory runs are being used.
https://www.eng-tips.com/viewthread.cfm?qid=107671
"By activating some of the *CONTROL facilities discussed below I have noticed reductions in solution times of more than 70% in a weld modelling problem with little or no change in the precision of results.
Firstly, in both the thermal and mechanical analyses it can save computer time by specifying more iterations per increment to prevent increments being cut-back, as follows:
*CONTROLS, PARAMETERS=TIME INCREMENTATION
** Relax checks on rate of convergence:
** IO IR IP IC IL IG
16 , 18 , 20 , 40 , 30 , 6
Secondly, a considerable amount of computer time can be saved by relaxing the convergence tolerances (which incidentally the ABAQUS manual admits are rather tight) as follows:
*CONTROLS, PARAMETERS=FIELD, FIELD=DISPLACEMENT
** ... or *CONTROLS, PARAMETERS=FIELD, FIELD=TEMPERATURE
** Relax solution tolerances:
** R_n^alpha C_n^alpha q_0^alpha q_n^aplha R_p^alpha eps^alpha
0.02 , 0.05 , , , 0.10 ,
Slacker tolerances give only very small differences in the solution that are insignificant, given the approximations inherent in weld modelling.
Finally, activating the (normally switched off) line search facility can lead to far fewer cut-backs, particularly those arising from:
'***WARNING: THE STRAIN INCREMENT HAS EXCEEDED FIFTY TIMES
THE STRAIN TO CAUSE FIRST YIELD AT - POINTS'
Activate the line search facility as follows:
*CONTROLS, PARAMETERS=LINE SEARCH
** Nls Smaxls Sminls fsls etals
4 , , , ,
However I am way out of my depth and its pretty hit and miss. I totally understand that messing in this area could seriously affect the solution and its clear any results generated with changes in these parameters need to be treated with even more care than norml. I normally would see this as a last ditch effort to get a solution but I can see other uses. The potential benefits are making solutions converge where they have not done without modifying these parameters and speeding up runs where lots of expletory runs are being used.
https://www.eng-tips.com/viewthread.cfm?qid=107671
"By activating some of the *CONTROL facilities discussed below I have noticed reductions in solution times of more than 70% in a weld modelling problem with little or no change in the precision of results.
Firstly, in both the thermal and mechanical analyses it can save computer time by specifying more iterations per increment to prevent increments being cut-back, as follows:
*CONTROLS, PARAMETERS=TIME INCREMENTATION
** Relax checks on rate of convergence:
** IO IR IP IC IL IG
16 , 18 , 20 , 40 , 30 , 6
Secondly, a considerable amount of computer time can be saved by relaxing the convergence tolerances (which incidentally the ABAQUS manual admits are rather tight) as follows:
*CONTROLS, PARAMETERS=FIELD, FIELD=DISPLACEMENT
** ... or *CONTROLS, PARAMETERS=FIELD, FIELD=TEMPERATURE
** Relax solution tolerances:
** R_n^alpha C_n^alpha q_0^alpha q_n^aplha R_p^alpha eps^alpha
0.02 , 0.05 , , , 0.10 ,
Slacker tolerances give only very small differences in the solution that are insignificant, given the approximations inherent in weld modelling.
Finally, activating the (normally switched off) line search facility can lead to far fewer cut-backs, particularly those arising from:
'***WARNING: THE STRAIN INCREMENT HAS EXCEEDED FIFTY TIMES
THE STRAIN TO CAUSE FIRST YIELD AT - POINTS'
Activate the line search facility as follows:
*CONTROLS, PARAMETERS=LINE SEARCH
** Nls Smaxls Sminls fsls etals
4 , , , ,
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!