To whom it may concern,

I am developing a fully coupled temperature displacement analysis in which I use TTS-shifted viscoelasticity (Time-Temperature-Superpositioning) in combination with "inelastic heat fraction" to make use of the materials property to heat up due to cyclic deformation. In my simulation, multiple instances are in contact using surface to surface contact, which add additional heating due to coulomb friction. 

Usually my simulation runs smoothly and does what it's supposed to do, but after long calculation times, e.g. after 5.7 Million steps in my lastest simulation, suddenly one element, that is both involved in contact and that has the viscoelastic material, suddenly collapses and the simulation is stopped due to excessive element distortion .

I've validated all my materials properties and tested the model at different temperatures, while at short simulation times, the model doesn't crash.

Screenshot showing acceleration of the crashing element (Element inverts / collapses)

Screenshots showing Sharp temperature gradient at the failing element (probably a singularity or something like that).

Some things I use in the model:

1. Parts:
   1. Nothing special, some partitions to enable proper meshing
2. Property
   1. Two different materials:
   2. Material 1:
      1. Orthotropic conductivity (Temp dependent)
      2. Density
      3. Orthotropic *Elastic (Instantaneous Modulus for *viscoelastic)
      4. Inelastic Heat Fraction = 1
      5. Specific heat (Temp dependent)
      6. Viscoelastic (Time Domain, Prony) 
         1. TRS (VUTRS)
   3. Material 2: Basically same material but isotropic elasticity
3. Assembly
   1. Nothing special, no penetrations
4. Step
   1. Automatic Incrementation, Global Estimator, Improved DT-Method, Unlimited Max time increment, Time Scaling 1, No Mass Scaling, Linear bulk viscosity 0.06 and quadratic bulk viscosity 1.2
5. Interaction
   1. Kinematic Contact, weighting factor of 1, leading to the coarse mesh (which is also the stiffer material) always being master surface
   2. Contact Properties:
      1.  Penalty tangential behavior with temperature tependent friction coefficient, dropping to 0 at a given temperature
      2. Normal behavior Hard Contact with allowing separation after contact
      3. thermal conductance
      4. heat generation
   3.  Contact Controls with disabled fast local tracking and more frequent (10) global search frequency, other settings on default
6. Load
   1. A material is excited with a harmonic oscillation using displacement boundary condition (no sharp steps). 
   2. The material that crashed is fixed in most DOF at a spot outside ouf the crashing region
7. Mesh
   1. Mesh-Study was done, no bad quality mesh in the parts
   2. Fully integrated elements, using element second order accuracy and element distortion control on .2
8. Optimization
   1. None
9. Job
   1. Double precision in analysis and package, Parralelization

Some things I tried in the past, that didnt seem to help:

• Higher linear or quadratic bulk viscosity damping
• Using either the integrated WLF approach or a self-written VUTRS routine
• I checked all my thermal and mechanical (instantaneous elastcity and viscoelasticity) material properties and validated them
• I both tried kinematic contact and penalty contact and also balanced contact or weighted contact
• With special contact controls and without
• Different meshes, both more coarse and more fine meshes and other element shapes
• I tried disabling several functionalites one by one, which also didn't help to identify the problem, as sometimes other errors occured due

Some things I don't do:

• I dont use any form of scaling (mass or time scaling)
• As my model is already crazy computationally demanding I cannot use general contact

My Conclusion:  The elements crashing are always at Edges and mostly on corners of parts that are in contact with other parts, which makes me think that there is some sort of material instability. As the models runs well for milltions of increments and then suddenly crashes, it has to do with the temperature locally changing over the course of the simulation. If I artificially heat up the material homogenous to the temperatures where the material tends to crash, this doesn't lead to any sort of crashing (on short simulations). I assume that crashing in my case may also have to do with a sharp gradient of material properties... If the temperature dependent gradient in viscoelastic (TTS) material properties seems to be the problem for the material instability, do you have any recommendation on how to fix this? I am working on this for many months now, and I can't find a solution to my problem. As usually a single element is collapsing to zero volume, I assumed that quadratic bulk viscosity damping may be the solution, as it smears shock fronts, but it also doesnt seem to help, as elements crash, whatsoever.

If you got any ideas on what to try to fix the crashing, have a similar problem or any recommendations for me, this would be highly appreciated!

If this is the wrong community, do you have any recommendation on where I could get support?

Best regards,

Jan Dietrich

Abaqus