Simulation of Dynamic Charateristics

[chinmay]

I have an assembly (a cuboid shaped structure). The vibration test data of this structure is available with me.

I want to study the effect of vibration generated by a system mounted on top panel on another system mounted on same top panel. In order to complete this, I have prepared the FE model of only top panel with lumped mass (to match the total mass of systems mounted on this panel) and fixed/simply supported boundary conditions (actual structure is having bolted joints).
In this configuration, the 1st mode (out of plane) frequency of top panel in actual structure and FE model is very closely matching. (The test data contains details about only first mode)

So, Can we say that the FE model of top panel incorporates / matches the dynamic characteristics of the actual cuboid structure ?
If not, what other parameters should be included/matched to simulate dynamic characteristics of the actual structure; so that any study carried on this FE model is valid on total structure too

 

[RogueOne]

Are you able to analyze the mode shapes that would've been measureable during the testing? If your model and your tests are showing similar mode shapes, amplitudes, and frequencies given the same input, then you could say that the model matches.

However, if the top panel would only be marginally effected by systems affixed to it, then it is possible that your model could be accurate without incorporating the effects that the affixed systems have on the top panel's structure.

 

[Dr.D]

In general, you cannot adequately replace a flexible substructure with a rigid mass and expect to preserve the dynamic characteristics. Dynamic characteristics, such as natural frequencies and mode shapes, are system properties, not properties of a localized part of the system.

 

[chinmay]

In general, you cannot adequately replace a flexible substructure with a rigid mass and expect to preserve the dynamic characteristics. Dynamic characteristics, such as natural frequencies and mode shapes, are system properties, not properties of a localized part of the system.

Sir, the complete cuboid is ~ 1000 kg, whereas top panel assembly is only ~62 kg. The systems mounted on top deck are highly stiff (frequency >1Khz). So my idea is to make FE model of only top deck (instead of complete structure) for analysis, but this top deck model must have same dynamic characteristics as that of when it is with cuboid structure. So, I have considered 2 point mass (29 kg each + 4 kg panel mass) located symmetrically near CG of panel, and fixed boundary condition. In this configuration the 1st mode frequency is matching with the experimental result. I want to know, what other parameters should be verified , so that the FE model of panel alone can be used to study response on system mounted on top deck (source of vibration is also on top deck).

 

[Dr.D]

To be fully equivalent, you must match all natural frequencies (not just the first) and all mode shapes. This is more than pretty difficult to accomplish!

 

[chinmay]

To be fully equivalent, you must match all natural frequencies (not just the first) and all mode shapes. This is more than pretty difficult to accomplish!

ok, I have experimental data upto 120 Hz only, in which only first mode is captured.
I will try to model the complete structure.
Thanks for the reply

 

[RogueOne]

I re-read what I posted and I want to clarify what I meant when I said

"..However, if the top panel would only be marginally effected by systems affixed to it, then it is possible that your model could be accurate without incorporating the effects that the affixed systems have on the top panel's structure."

What I was saying was that although your model might be accurate (as far as the limited amount of testing you've done will show), it is not precise until you incorporate the effects that the affixed systems have on the top panel's structure.

 

[chinmay]

I re-read what I posted and I want to clarify what I meant when I said

"..However, if the top panel would only be marginally effected by systems affixed to it, then it is possible that your model could be accurate without incorporating the effects that the affixed systems have on the top panel's structure."

What I was saying was that although your model might be accurate (as far as the limited amount of testing you've done will show), it is not precise until you incorporate the effects that the affixed systems have on the top panel's structure.

Thanks for your response.
The sides of the panel is approximately 800 mm. During testing accelerometer was put at 4 locations on the panel. In my FE model, I gave the same input force as that in testing and checked the response at the location of accelerometer, (exact location was not known, so I took the average of response of surrounding 8-10 nodes). Also measured damping value for 1st mode (from test) was used in FE model. In this configuration, difference in test & FE response is within 5%. So, I assumed that this model is correct and can be used for further analysis.