How to know if you have a good mesh in Nastran In-CAD

[@shantanukajale]

Is there any specific reason why Nastran doesnt provide an option for hex meshing ?

[@ReachShaunNow]

Yes. Otherwise, it would be an option in at least one of the many simulation/analysis software that is already available. However, let's change that to platonic meshing - as in platonic solids.

[@ReachShaunNow]

Wouldn't it make sense to make the 'element size' equal to either: a) the model's smallest feature's dimension and equal to that smallest feature's decimal precision or b) equal to the precision and/or tolerance of the device or being who builds the physical model? (e.g. If a 3D printer with a precision of 0.005 inches and a tolerance of +/- 0.005 inches is being used to manufacture said model - then, we would set the mesh's element size to 0.005 inches and set the mesh's tolerance to +/-0.005 inches.) I'm using Inventor 2018 with Nastran In-CAD (now has a tolerance setting on the 'settings' dialog box) -- just FYI. My question is, why don't any of the other options in the settings matter?

[@jmkubli]

Hi +Reach Shaun, Mesh sizing is relative to the size and complexity of the model. b) For example, if we set the mesh size to the precision of 0.005 (the smallest feature that the 3D printer can create) and we have a simple block that is 2 x 2 x 2 inches, we would end up with a model that has 64 million elements. This would take a couple days to solve. And if it did solve, could be wrong. Computers use a finite number system, and can track only so many significant digits. This can lead to rounding errors. a) Now the above block is smooth and doesn't have any features as small as the tolerance. But lets say we have a tiny embossed feature, like text that is 0.001 inches thick. If we mesh that and run the analysis. That feature would have almost no effect on the results. And due to the finite number system, be so small relatively to the over results, that there is no change in the displayed results. It is all a balancing act, my two examples are on the extremes to help understand the issue. Now as regard to the other settings, they mostly control number of elements on curved faces and are relative to the size of the model, so changing them doesn't always have a noticeable change in the mesh as other factors are more dominating. That is why it is easier to manually add mesh controls to areas that you are interested. It is good to know or have some expectation of what will happen in the model. Not knowing and not apply some engineering logic to it can lead to failure in getting the correct results.

[@ReachShaunNow]

I understand this. Let's think in terms of 'perfect accuracy' and 'every aspect matters from the smallest to largest'. Let's get in the anal-retentive mindset; granted, software is currently incapable of running such a data-loaded test in a desirable time with desirable performance. Features in one location produce no change in test result but, perhaps coupled with features opposite of aforementioned location do. Maybe; just maybe, products are designed to a standard that 'everything' (even the embossed/engraved words on products) - serve a purpose beyond aesthetics/cosmetics. What I'm understanding is the only reason we need to "refine our mesh with mesh control on particular features" is because the amount of 'faces' and 'points' produced by generating a highly accurate mesh would be overwhelming for the program's capabilities. And, the solution is to simply omit "unnecessary" elements/features, cut models in half/quarter sections and run more individual tests (at individual areas of concern) rather than running one test (at all areas of concern) using the complete model with accurate meshing/results simultaneously. Am I dreaming? Did I just have an epiphany? Maybe some programmer somewhere will take this challenge and begin manifesting what would be the greatest evolution in FEA analysis software to date.

[@jmkubli]

Why? You may get a more accurate result for a Simulation, but then you would have to manufacture exactly the same way you simulated it. Down to the atomic level in creating a perfect crystalline structure. Simulation is a educated guess on what is a conservative approximation of the material properties, which is a averaged guess of the manufacturing process effects on materials and dimensions.