Learning Forum

[hbahrein]

I am currently working on a critical project involving the modeling of human brain biomechanics. Initially, I successfully developed and debugged my model using the SMP solver. However, I recently discovered that my model encounters significant issues when using the MPP solver.

After seeking guidance from the support through a forum post, it became clear that there are fundamental differences between the MPP and SMP solvers, particularly in how they handle contact formulation. Consequently, I have embarked on the task of debugging my model with the MPP solver to overcome these challenges.

However, I have encountered a perplexing problem during this debugging process. A small portion of my model, comprising less than 10 elements within a model of over a million elements, behaves abnormally when subjected to various loading conditions. Strangely, there is no discernible distinction between this problematic section and the rest of the model. The right side of the pia mater, which is a shell surrounding the brain, exhibits erratic behavior in the MPP solver, even though it shares identical material properties, element formulation, and contact formulation with the left side. It's worth noting that this exact model works when simulated with the SMP solver.

I am urgently seeking your expertise and assistance in resolving this critical issue. Your guidance and insights into this matter would be immensely valuable in helping me ensure the accuracy and reliability of my brain biomechanics model.

Your prompt attention to this matter would be greatly appreciated. If you require any specific details or additional information about my model, please do not hesitate to ask. I eagerly await your response and assistance in addressing this challenging problem.

Thank you for your time and support.

[Jim Day]

The fact that MPP has trouble, and SMP does not, indicates you should consider a tweak to the contact definition and/or the way the model is decomposed for MPP. Well, let me add one more thing ... a reduction in time step scale factor (TSSFAC in *CONTROL_TIMESTEP) together with use of a double precision executable may, in themselves, resolve the instability. Decomposition: Results of MPP can vary (slightly) depending on how the model is decomposed (how it's split among the processors). Use *CONTROL_MPP_DECOMPOSITION_SHOW to view the decomposition of the model. Does one of the decomposition boundaries coincide with the problematic area? There are variable *CONTROL_MPP_DECOMPOSITION_option commands to alter the decomposition. Contact: Provide an image of your *CONTACT input for the contact in question and provide a narrative of what's on the SURFA and SURFB sides of that contact. In that narrative, specify what element formulations are employed.