¿Cómo actualiza y calibra su modelo de vibración FEA en función de los comentarios y cambios?

If the mass properties of your FE model do not match closely (a few %) of those those measured or predicted (CAD) you physical article no need to go further! I’ve seen where this first step was ignored which resulted in severe issues with FE results!

In regards to structural dynamics, one of the main objectives I've always sought with model correlation is that the primary modes and large mass participation modes which are responsible for the majority of the stress in random vibration environments are correlated within an acceptable percentage in frequency, usually within 5%, so long as the mode shapes are also well correlated between the FEM and actual vibe test data.

In my opinion, the very first step consists of defining what are the design criteria the design at hand needs to satisfy. Do you need to predict global bending and torsional modes of a high-rise building or the vibration of a specific floor on which a machine will be mounted? The response to this question will affect the extent of the model, the element types and the mesh refinement. It is only once you understand what you need to verify in your design that you can design the proper FEA model and then pick the correct objectives and criteria to validate your model.

Analyze the energy transfer paths and dissipation mechanisms between the model and the real system. This involves understanding the pathways through which energy moves and the processes through which it is dissipated or dispersed. This will provide system-level insights into the behavior of the model, helping better understand and predict system interactions.

In the case of a free body check. First 6 modes should be <10e-3 Hz. Next 7 to 10th modes should be > 10Hz. If not the case you may want to check deflections and see there is a mechanism in your model or something is not properly connected.

When comparing modeled to measured data it’s always important to revisit assumptions. How well are loss mechanisms in your system known. Is the problem coupled to another domain such that allows for additional energy storage or loss? How valid were the initial boundary conditions. It’s always useful to try to bound those variables and perform quick checks of lower level assemblies to build confidence in more complicated measurements. The more little answers that are known to be “correct” the less the final answer can deviate.

It is not always possible to build prototypes and collect experimental data. In these cases it is possible to compare with similar designs that have been thoroughly studied and validated. The differences between the models can be compared. The Analyst can compare if the changes between the known model and the custom model make sense. There will likely be behavior changes, but these changes can be predictable. Using engineering fundamentals an analyst can begin to make observations. A simple example would be to determine which design should have a larger fundamental period. Observations like this can be compared to the response spectra between the two different models for validation when test data is not available.

Usually you record deflections from experimental data. It should correlate closely with deflections from non linear FEA, if the deflections are small then it should also correlate with linear FEA.

I began my career in dynamic analyses using NASTRAN in 1974 working on the Space Shuttle. I often wondered if the results were worth the paper the results were printed on. Through the tutelage of people such as Mark Richardson, David Brown, Randy Allimang, John O’Callahan and PeterAvitable,I acquired a strong background in modal testing and FE model verification. After 42 years in the defense industry the lack of implementation of FE model verification is my biggest career disappointment. Acceptance of computer results as gospel is disheartening!! I’m retired so I feel free to make this statement!!