Frequency response load setup

[Anonymous]

Hello,

I am new to Nastran and quite confused. 

First, could you give me some advice if Nastran is the correct way for my analysis? My aim is to simulate the vibrations of a shale shaker which basically has 2 centrifugal motors that apply a uni directional force with a sinusoidal function, the displacement follows the same sinusoidal curve.

I am trying to approach this problem with the modal frequency response analysis as I got the idea that could be right.

Given that, I am quite confused (I am a mechanical engineer but never worked on vibrations) about the enforced motion setup and what Nastran is doing with that.

If, for example I apply an acceleration (as enforced motion) in the shaker vibration direction, what is Nastran doing with this acceleration value? What is the relation between the acceleration value and direction and whatever Nastran is applying for the simulation? Is Nastran applying a force that gives that acceleration and in a sinusoidal form?

Then, I did not understand the load scale factor and probably for the same reason as above.

I hope this makes sense. I have no example file as I am still too confused to approach the simulation (I did run some simple tests on very basic models).

I am running Inventor Nastran 2022

Thank you

Luca

[John_Holtz]

Hi @Anonymous . 

(edit 2023 Apr 10. Corrected "first displacement" and "second displacement" to "first derivative" and "second derivative".)

You asked " is the correct way for my analysis?".

Answer: Yes, Frequency Response gives the result at the maximum extreme due to a sinusoidal load.

You asked "I apply an acceleration (as enforced motion) in the shaker vibration direction, what is Nastran doing with this acceleration value? What is the relation between the acceleration value and direction and whatever Nastran is applying for the simulation? Is Nastran applying a force that gives that acceleration and in a sinusoidal form?"

Answer: I'm not sure of the actual calculation, but I'm sure that it is all related. The displacement and acceleration are certainly related. For a simple mass on a spring harmonic motion:

• displacement=max mag*sin(forcing frequency*t+phase angle)
• first derivative: velocity = forcing frequency*max mag*cos(forcing frequency*t+phase angle)
• second derivative: acceleration = forcing frequency^2*max mag*cos(forcing frequency*t+phase angle)

The input (for whatever type of load you are applying) is the "maximum magnitude". But real loads are often a function of the forcing frequency. This is where the load factor comes into use. The load at some frequency Fi = max mag*load scale factor, interpolated from the load table at the frequency Fi.

This article may be helpful for the Frequency Response analysis. See Analyze a rotating component with an imbalance load.

Let us know what questions you have.

[Anonymous]

Hello John,

thank you very much. Yesterday, after "thousands" of searches (mainly in this forum), I simply typed in the word "shaker". And I found out that, guess what, there was a post about a shaker! This is the post:

https://forums.autodesk.com/t5/inventor-nastran-forum/shaker-for-grain/m-p/8544307

This way I found out that page you linked while replying to me and other useful information. 

Then, I found out other useful information, in particular in this post about spring connectors and the need for rigid body connectors where to fix the springs:

https://forums.autodesk.com/t5/inventor-nastran-forum/spring-connectors-how-to-use/m-p/7483436

I think I have enough to setup the analysis and run some simulations. I hope they'll work, I will post the files soon (I am sure I will have other things to ask).

Thanks!

[Anonymous]

Hello,

I have a question about the results. I am looking at the displacement related to the above simulation, a frequency response analysis of a shaker, attached is a picture. 

Setup: 1 force on each motor (2 motors), 4 springs connectors between the shaker and a fixed base.

The direction of the displacement is correct (orthogonal to the motors' axis and just along this direction). I am confused about the magnitude and how the displacement is shown. The applied load is a sinusoidal function, the amplitude of which is the force magnitude in the setup, therefore the displacement has to follow a sinusoidal function. What does the displacement value in the results represent? The amplitude of this function? Is this the sum of positive and negative amplitude? I mean, if you look at the horrible sketch, is A or B? Or something else?

  

Thank you!

[John_Holtz]

The results are the peak at each of the forcing frequencies, so they show the amplitude A as the model displaces back and force.

When you plot the point "A" for each frequency that you shake the model at, you get a graph like this: (random figure found on the internet). This shows that at each natural frequency (10 Hz, 75 Hz, 210, 400), the result spikes due to resonance.

[Anonymous]

Thank you, that makes sense.

I've got an issue with a spring on a more complex model that replicates the same simulation (a shaker on 4 springs), here is a picture of the model and the related spring (first on the left):

The analysis setup is the same as the other but when I run the analysis on this model, I get this message:

"Could not find a node to apply a connector, please check FE mesh and connector"

I assume is that spring as when I delete it I don't get any message and if I run the simulation with it, in the results the spring is shown not deformed.

Any idea about what is going wrong here? The mesh is fine and the rigid connectors for the spring are linked to nodes (the four spring are all the same with all the same rigid connectors). The spring has stiffness setup on x/y/z and rotations as well (exactly as the other springs).

Thank you

[Anonymous]

Hello,

I proceeded to refine the mesh and the message disappeared, but still not convinced why it used to appear

[maildodanillo]

I`m testing the resonant frequencies of a body but seems like the code only produces force in the positive axis not full ac -1 0 1.... the thing is...
This is the graphics

I have for the simulation

but I was hopping for something like this....

Appreciate any replies...

[John_Holtz]

Hi @maildodanillo 

You are incorrect when you write "seems like the code only produces force in the positive axis not full ac -1 0 1". The analysis is for a load that follows an infinite sine curve: load = load magnitude*sine(forcing frequency*time+phase angle) where

• load magnitude is entered in the model setup.
• forcing frequency is entered in the Dynamic Setup
• phase angle can be specified for the load.

Time is not entered because the analysis is not calculating results versus time. Only the results from the peak is output, and therefore it does not matter if the value is "positive" or "negative". The model vibrates equally in both directions. In other words, X*load magnitude*sine(90) = -X*load magnitude*sine(270), so why output both +X*load magnitude and -X*load magnitude?

It looks like you have copied the data from the Nastran results to Excel, so you should be able to create the graph that you want.

John

[maildodanillo]

Hi John thanks for the fast reply.

So I got to print this result in "-X*load magnitude*sine(270)" and add to this actual graph?

I`ll try that, but my main question is just for one second to consider that +x, 0, -x.

Could it be used than as a two degrees of freedom system?

As two +x, 0 equations, doesn't sound correct in the math point of view.
Something going back in the negative convergence to same value as inertia and stress.

I wonder that the graph should have some x^3 for answer or something like it.

changing energy and displacement in between +x and -x.

Am I that wrong?

[John_Holtz]

Sorry @maildodanillo 

Maybe I do not understand what you are trying to do.

• I thought you have results for X values of 0 to 10000 and Y values that you have plotted. (Where these results come from is technically not important, but since you responded to an existing post about the analysis type names Frequency Response, I assumed that you are running a Frequency Response Analysis.)
• Since the graph only has Y values that are positive and you want to add plots are are negative Y values, all you need to do is add another column (or 5 or 20 or how many plots you want to add) to your spreadsheet and enter the formula -1*Y. That would give the "mirror image" that you show in your mocked-up image.

Let us know if you are asking about something different and/or a different type of analysis.

John

[maildodanillo]

Hi John yes,

Actually I believe that the mechanics principle used will not give me the results I want....
I see what your saying about changing the phase to plot a counter movement graph.

In fact I`m studying this exact behaviour, for my master thesis.

Rayleigh`s equations only consider a single sided diaphragm movement, I`m trying to understand dynamic loads and the backlash effect from the center when its load reaches the stiffness and resonant point.

I don`t think the equations are wrong but they are not fully developed, seems like any round object will curve itself to a +x or-x depending on small differences, then the movement only occurs at a biased direction the same at which material is naturally conformed.

When the load or weight raises to the stiffness point the material migrates to the other side with a backlash effect.

look at the image below,  the ripple effect appears to have two modes one clockwise and the other counterclockwise...

I`m looking for ways to reproduce that at a solid disc.

Thank you again for you attention!

[maildodanillo]

Hi I just wanted to have if possible a suggestion why this is happening...
Below is my simulation with ANSYS, and it seems a bit diferrent.

This is almost the same simulation with NASTRAN, shapes are very close but frequency is very off around 10khz.

[John_Holtz]

Hi @maildodanillo 

Frequencies are related to the stiffness and the mass. If the frequencies are different, you should check the material properties (modulus of Elasticity for stiffness and mass density for mass) and constraints (for stiffness of the model). Until you resolve the differences in the frequencies, there is no need to run a Frequency Response analysis.

The first graph looks like a plot of random numbers. It does not look like the result of a frequency response analysis. The other two graphs look like a frequency response analysis where the peaks at resonance are clearly defined.

John