This week I participated in an NESC conference in Los Angeles. One of my colleagues brought up the following point.

Finite element models are created to represent physical models for analysis purposes. The finite element model natural frequencies can then be compared to data from a modal test of the actual hardware. At this point, the finite element model can then be “calibrated” so that its fundamental frequency agrees with the test results. This can be done, for example, by scaling the elastic modulus in the analytical model.

There are two problems with this approach:

1. The analytical model no longer “matches the drawing” of the hardware, (although this is an inherent flaw in all models to some extent).

2. Changing the elastic modulus will throw off any dynamic stress calculations.

This dilemma deserves further thought…

### Like this:

Like Loading...

Depending on requirements I build several models: A loads, heat transfer, dynamic, stress. The model mesh fidelity and attributes may be somewhat different depending on the objective but they all represent the outline drawing. For dynamics, I usually create a quite coarse model. I am looking to, for example, generate Grms data. I then use the Grms data as the input to the ‘finer’ mesh stress model to evaluate stress. I too dial the dynamic model stiffness matrix, not the stress model, to match vibe test data.

The effect of change in E on the stress at time instance t’ is negligible under the following assumptions:

1. Homogeneous, Istoropic, Linearly Elastic Material

2. Time instance t’ is much larger than time taken for sound wave to travel the solid

It would be interesting to write an algorithm that matches FEM models to actual modal results. This should be something optimization-based. Say you have the first couple modal frequencies and you’d be able to automatically adjust the elastic modulus (possibly other parameters like density) until it gives the best estimate, based on a cost function…