Hello,
I am doing an 3D indendation simulation modeld on Vickers in the moment. I am using Abaqus Standard with a step time of 1.5, increment size initial: 0.001, minimum: 0.0000005, maximum: 1 and max. number of increments 100000. The Indender is formulated as a rigid body where a concentrated force of 1 Newton acts on the reference point. Furthermore gerneral contact is implemented. The mesh elements unter the tip of the indender (0.1x0.1ym) are 0.05x0.05x0.05ym. The mesh size increases towards the outside uniformly. For job execution I use the high performance computer 'Elwetritsch' at the TU Kaiserslautern.
The run time of the job is displayed as one month. Does someone have an idea why this takes so long or where my mistake is ?
Thanks everyone!
Consider these recommendations:
1. Don't use more elements than necessary
2. Ensure the scratch data fits in your system's RAM
3. Don't allow a single element to ruin your simulation speed in Abaqus/Explicit
4. Use scaling for quasi-static analyses in Abaqus/Explicit
5. Use parallelization
6. Pay attention to constraints
7. Regularly restart your system
Thanks a lot for your answer. I optimized the model according to your ideas with some success, but the run time still takes quite long. In the following i depicted an interation log:
ASYNCHRONOUS TRACKING ELEMENT LOOP STARTED
28 SEVERE DISCONTINUITIES OCCURRED DURING THIS ITERATION. 16 POINTS CHANGED FROM OPEN TO CLOSED 12 POINTS CHANGED FROM STICKING TO SLIPPING
CONVERGENCE CHECKS FOR SEVERE DISCONTINUITY ITERATION 1
MAX. PENETRATION ERROR 8.96646E-03 AT NODE PRUEFKOERPER-1.1361 OF CONTACT PAIR (ASSEMBLY_SURF-PRUEF_CONT,ASSEMBLY_SURF-INT_CONT) MAX. CONTACT FORCE ERROR 46.6995E-06 AT NODE PRUEFKOERPER-1.1669 OF CONTACT PAIR (ASSEMBLY_SURF-PRUEF_CONT,ASSEMBLY_SURF-INT_CONT) PENETRATION ERROR TOO LARGE COMPARED TO DISPLACEMENT INCREMENT.
AVERAGE FORCE 9.456E-09 TIME AVG. FORCE 1.743E-02 (FROM MOMENT FIELD) LARGEST RESIDUAL FORCE 9.754E-06 AT NODE 820 DOF 3 INSTANCE: PRUEFKOERPER-1 LARGEST INCREMENT OF DISP. 2.451E-02 AT NODE 16162 DOF 3 INSTANCE: PRUEFKOERPER-1 LARGEST CORRECTION TO DISP. 1.835E-02 AT NODE 16162 DOF 3 INSTANCE: PRUEFKOERPER-1 THE FORCE IS ZERO EVERYWHERE BUT THE FORCE RESIDUAL OR THE DISP. CORRECTION IS NON-ZERO
AVERAGE MOMENT 0.00 TIME AVG. MOMENT 1.000E-02 ALL MOMENT RESIDUALS ARE ZERO LARGEST INCREMENT OF ROTATION 1.904E-45 AT NODE 421 DOF 5 INSTANCE: INTENDER-1 LARGEST CORRECTION TO ROTATION 1.100E-45 AT NODE 421 DOF 5 INSTANCE: INTENDER-1 THERE IS ZERO MOMENT EVERYWHERE
COMPATIBILITY ERRORS: TYPE NUMBER EXCEEDING TOLERANCE MAXIMUM ERROR IN ELEMENT VOLUMETRIC 0 -5.758E-10 241391 INSTANCE: PRUEFKOERPER-1
NUMBER OF EQUATIONS = 1365629 NUMBER OF RHS = 1 NUMBER OF FLOATING PT. OPERATIONS = 2.84E+13
USING THE DIRECT SOLVER WITH 16 PROCESSORS
CHECK POINT START OF SOLVER
CHECK POINT END OF SOLVER
ELAPSED USER TIME (SEC) = 1944.5 ELAPSED SYSTEM TIME (SEC) = 36.830 ELAPSED TOTAL CPU TIME (SEC) = 1981.3 ELAPSED WALLCLOCK TIME (SEC) = 125
***WARNING: THE SYSTEM MATRIX HAS 19 NEGATIVE EIGENVALUES.
NEGATIVE EIGENVALUES MEAN THAT THE SYSTEM MATRIX IS NOT POSITIVE DEFINITE: FOR EXAMPLE, A BIFURCATION (BUCKLING) LOAD MAY HAVE BEEN EXCEEDED.
After all the process works and the increment size increases but considering the step time of 1.5 one iteration seems to take too long.
Thanks a lot for help!
Why do you start with such a small initial increment?
Even if you do, Abaqus Standard will increase the time increment if possible, so they may become much larger later on and allow the simulation to run faster.
I would suggest to first start with a coarser mesh, see how that works and then refine, in that way you will get an idea where possible convergence problems may occur.
- Badges
- Science topic