Written by: Shaun Bentley, Application Engineer
1: Start simple, but smart.
If you are newer to Simulation, you may be tempted to take whatever model you have and throw it directly into the software. Experienced users may roll their eyes at this general approach, since frequently the result it leads to is a quagmire of irrelevant details and errors.
Some preparation is frequently required and should typically be focused around modeling only the absolutely relevant features as the starting point. Examples of preparations that can be made:
- Suppress unnecessary features (outside fillets, irrelevant holes, etc.).
- Consider running the analysis on only the part (rather than the assembly).
- Consider completely reconstructing the model from surface bodies (for shell elements).
- Treat all contacts as totally bonded.
The primary purpose of this step is to be a kick-off point to allow you to BUILD an understanding in easy to interpret steps rather than to get caught in the weeds and hope you have the right answer when it finally runs.
2: Add complexity in small steps
So you already ran the simple model, but the results do not match the physical prototype. Next you should add small bits of detail in the relevant regions. Were some of those fillets you suppressed earlier actually drastically improving the stiffness? Re-include them and see if you observe a different result (frequently you won’t). Perform these same tests by adding on other relevant details in order of priority. This procedure will also allow you to build an intuition over what is important and what is not.
3: Apply appropriate changes to the mesh
A “fine” mesh isn’t necessarily an accurate mesh. Typically you should test at least two or more mesh sizes to see if your results change. If I use a “fine” mesh and get 10ksi stresses, and then I change it to a “finer” mesh and get 15ksi, I don’t trust either of these results. We call this process of changing the mesh “mesh convergence” since in many cases (not all cases; see “stress singularity”) the results will stabilize around some number.
4: Simulation fixtures and simulation loads
Are you sure you have them right? You are? No, you’re not. None of the simulation fixtures and loads exists in reality. Only contacts really exist (per se; I don’t want to get in a quantum physical debate here). What I’m getting at is that we use fixtures and loads to approximate the behavior of these contacts (i.e., bolted contacts, welded connections, etc). Sometimes, problems warrant modeling more precisely how your geometry is fixed. Maybe generating a 3D model in a portion of the fixture would help. Make sure you add portions of the fixturing in steps to see the incremental improvements to your results (like I said in step 2 above).
5: Give DASI a call!
We have engineers on staff and would love to show you how to apply the above principles in practice.