Simulation - Static Stress does not match reality

[Serip_Technical]

Hi everyone,

I’ve run into a discrepancy between Fusion 360 simulation results and real-life testing, and I’d really appreciate your thoughts.

I'm working on a ceiling luminaire component (canopy + crosspiece) and simulated a vertical load (Y-axis, same as gravity direction). In Fusion, the simulation predicted that the screws would fail first under load.

However, after doing a real traction test in a lab, the screws actually ripped through the aluminium canopy instead — so the material failed, not the screws.

I’ve double-checked my simulation setup and believe it’s configured correctly:

• Materials are properly assigned (aluminium for the canopy, steel for the screws)

• Contact surfaces are defined

• Mesh looks fine

• Load and constraints mimic the real setup (traction along Y)

Is there something Fusion doesn't take into account well in these cases?
Could it be a contact/mesh issue, or a limitation in how screw-to-plate interactions are simulated?
Any ideas on how to better match simulation results to reality?

Thanks in advance!

[bwalker145]

I wonder if the torque that the bolt is experiencing in the simulation, due to the constraints, is creating an artificial localized point of high-stress, causing the premature failure. This seems to be supported by the color mapping of the surrounding material. Could possibly try running the simulation again with simplified bolt geometry, not sure if that would help or not.

As a sanity check, I'd recommend doing some hand calcs for shear strength of the selected fasteners to compare against your results; this will give you a good idea if the results seem reasonable. Failing with a load of 1500 N (337 lbf) doesn't make sense to me with 4 bolts of this size.

Your lab results show the canopy material failing at 9000 N (2023 lbf). I'd have to do the calcs to verify, but with the load distributed across 4 connections & using alloy steel fasteners, that seems like a reasonable outcome [Edit: reasonable in the sense that the bolts didn't fail first at this load].

[henderh]

Hi @Serip_Technical,

First of all, thank you for providing the detailed background info. I agree the Fusion Sim results should match reality within ~10% (or better) as FEA software should when the inputs are properly defined, and known limitations are taken into consideration.

The NASTRAN solver is accurate, and passes numerous numerical result validation NAFEMS certification test cases. If we can presume the results are accurate, I wonder what else could be causing the discrepancy.

Disclaimer: This is my personal opinion, which is not PE licensed (i.e. I could be completely mistaken here).
TL;DR the casting in real-life wasn't acting perfectly isotropic as it is during simulation.

My current theory is there was a stress concentration that caused a brittle rupture effect, that may have manifested during the manufacturing process.

It looks cast, and if there was a crack formed it will be the root of the stress concentration and will cause the run to rupture and total failure, occurring below the operating load (Fracture Mechanics). The surfaces created from the break can often reveal the likely source of the failure, and if the crack formed quickly or over time (where a portion of the surface is rubbed smooth from multiple load cycles before the crack runs). I wonder if there was a crack or a cold weld formed when flow fronts met during casting. Another possibility might be the contacting surfaces weren't perfectly smooth, and a small protrusion that was smashed may have been enough of a local stress riser, that it caused a crack to form.

What a bummer it failed when it shouldn't have. If you can recreate the test, verify again the material is as perfectly isotropic as possible around the likely failure zone(s). In my past life, I was involved in the cast parts industry. In one project I worked on the CAD design of a rotisserie frame to inspect turbine castings (IGT, etc.) using x-rays. I've heard a cheaper option is Magnaflux (if it doesn't have to spin very fast).

Please let us know if you have any additional questions, comments or suggestions.

Best regards,

[TimelesslyTiredYouth]

Hello,

Really interesting thread. Just to add something — I’ve noticed that in Fusion 360, screw interactions can be tricky to model accurately, especially if you’re relying on bonded contacts or if the mesh around the holes isn’t tight enough. One thing I’ve tried in similar setups is using the Bolt Connector feature instead of modeling the full screw geometry — it gives better control over preload and shear, and can highlight whether the failure is material- or fastener-driven. Also, adding frictional contact instead of just bonded might shift the stress paths closer to real-world results. Just a thought — curious if anyone’s tried those in this kind of setup?

Sleepy Regards

Ricky

[Serip_Technical]

Thank you. I will experiment with the contact settings and provide feedback.

[Serip_Technical]

It is cast! 

I agree that the simulation assumes a perfectly isotropic material, which is rarely the case in real life.

Our cast parts do tend to have internal porosities, so the idea of pre-existing microcracks is plausible. However, upon closer inspection of the failed parts from the test, we noticed signs of gradual crack propagation, not a clean brittle break, which supports the fracture mechanics theory.

That said, there’s one aspect I find a bit contradictory: all four screw holes failed in a very similar way, almost simultaneously, which seems unusual if the failure originated from a single casting defect. I would have expected a more localized failure, with one hole giving out first and the rest following under redistributed load.

Thank you a lot for your feedback.

[Serip_Technical]

I'm going to re-run the simulation with simplified screw geometry and double-check the constraints around the screw holes. Thank you!

[bwalker145]

@Serip_Technical  Just curious, what was the main failure mode? The picture on the left looks like the thin upper web broke out of canopy, but the picture on the right almost looks like the fastener head got pulled through the clearance hole instead; not sure if it's a just a trick of the picture.