Hi Pooria,

as a fellow Abaqus/Explicit user who has endured many failed simulations, may I suggest the following easy-to-implement steps:

1.) Check for consistency of units. If you decide to use SI units for your modelling, then stick strictly to using SI units throughout your model property definition. There is a chance that the main problem with your computational modelling is with the defined material properties, specifically the thermophysical properties. Nobody likes to be told that they might have provided the wrong properties to Abaqus. Still, it is a real problem for the Abaqus/Explicit solver.

2.) Do dimensional analysis on your governing equations, for example, your mechanical equilibrium and heat diffusion equations. Sometimes I use hand calculation to make sure that the results make sense to me. You would be surprised at contrasting results between hand calculation and mindless FEM 'plug and play' with Abaqus. If you are using absolute zero temperature and Stefan-Boltzmann constant, ensure that the units are dimensionally consistent. At least, let them make sense to you before asking Abaqus to use them.

3.) Check for machine precision. This is usually where SI units might not serve your purpose. So, you need to use another consistent system of units that requires fewer zeros (10 MPa instead of 10000000 Pa). When it is inconvenient for me to use SI units, I usually refer to one useful example in the Abaqus Example Problem Manual, e.g. [

[1.3.18 Inertia welding simulation using Abaqus/Standard and Abaqus/CAE] http://130.149.89.49:2080/v6.8/books/exa/default.htm?startat=ch01s03aex48.html

4.) Reduce the complexity of your DFLUX subroutine and ensure that the subprogram is verifiable, only increasing in complexity when fewer solution procedures are accurate enough. If you're using a 'borrowed' DFLUX, you might struggle because the original author may have put things in place to achieve a specific result. Not all subroutines are generic; a hard lesson to learn when utilising the Abaqus/Explicit solver.

5.) If you are using the mass-scaling algorithm, suspend it temporarily so that your numerical procedure employs the default mass-scaling of Abaqus/Explicit solver. You will have fewer problems to contend with although such an approach could be computationally demanding. There is a temporary workaround; see the next suggestion.

6.) Try to use a much smaller analysis time step, e.g. 0.01 s instead of 1.0 s or 10 s, depending on your thermal-mechanical analysis. It will allow you to test very quickly and many times if your small and simple adjustments are profitable for progressing your research work.

7.) Try to use a much smaller sized model, with bigger elemental length or smaller total number of elements/nodes. I find this to be useful especially if the intent is to quickly see if my ODB results make sense. It is painful to run a simulation job for so many wall-clock hours only to realise a minor error in the model definition.

I hope that these suggestions will help even a little. I will follow your question here and learn from other researchers in this space.

Thank you Saviour Okeke for your attention and thorough descriptions.

Your answer is useful to me but I still can't solve my problem.

Regards

Go through Abaqus Explicit user manual and youtube videos on welding. It will help.

Dear Pooria,

If you are using model change (element activation), try to change the length of each segment for added filler material and see if it helps. The length of those segments might be unsuitable with regard to your welding speed. Perhaps you will reduce the length of those segments and element size in longitudinal direction will accordingly reduce.

Your time period in step module will change based on that.

Regards,