Using Large Mass to Input EQ Acceleration 4

[damstructural]

Is anyone familiar with the following technique used to input an earthquake acceleration time history into an ANSYS model (or other FEA code)?

A very large point mass is located at the base of the model (deep in the foundation part of the model). The point mass is several orders of magnitude larger than the total mass of the rest of the model.

The acceleration time history is applied as a force on the very large point mass. The model is not restrained except for the 3 degrees of rotation at the point mass.

If anyone has any experience with this method, I'd appreciate a response to ask some further technical questions.

 

[Drej]

The large mass method (LMM) is a standard method of using an acceleration time-history by applying a force to the large mass as you've described it. The mass itself is usually, say, 1E+05 (in each direction) bigger than any of the masses used in the model for this.

Cheers,

-- drej --

 

[GregLocock]

The reason you do it that way is to give a source of vibration that can't do the impossible, but is insensitive to your modelling assumptions. In mechanical terms a naive modeller might use an excitation technique that develops an acceleration regardless of the impedance of the structure. The structure will move even if it is unrealistically stiff, developing large forces.



Cheers

Greg Locock

 

[damstructural]

I don't think I'd call this method "standard" for civil/structural analyses, which is why I posted. The more "standard" method for buildings, bridges, and dams is to input the time history directly at the restraints, as an acceleration or displacement.

I'm trying to gauge how often this method is used, and to learn more about it. I'm reviewing a project where this method was used, and our consultants at Caltech and UC Berkeley are just barely familiar with it.

Dreg, How sensitive is the magnitude of the "Black Hole" mass that is used. You cite 1E+05, they're using 1E+04. When does a problem with roundoff error start to kick in, and how do you calculate mass participation ratios with this monster mass?

Greg, In a large earthquake, everything moves, and the acceleration history is carefully chosen to match a response spectrum for the best estimate of the anticipated ground motion. Where looking for, and expecting to see large forces at the modal frequencies. I don't see any advantage to this "LMM" method for the reasons you state. I might be missing something.

Thanks for your replies. It looks like this method is used fairly often for mechanical/structural engineering. Is that a fair assumption?

 

[Eap2n]

Hi,

As far as NASTRAN goes, the large mass method is no longer necessary as of version 2004 because direct enforced acceleration in dynamic solution sequences was implemented...actually officially it was included in v2001 (there is a good explanation in the v2001 release guide) but there were a lot of errors found so it is best to stick to the large mass or lagrange multiplier methods if you are using v2001 or earlier. The main advantage of the direct enforced method is convenience and numerical precision (which sometimes can be a problem with the LMM).

Erik

 

[tomirvine]

The large mass method is used in aerospace vehicle analysis.

For example, consider a circuit board in an avionics box. The box is mounted on a bulkhead that applies a base excitation. The original source of the excitation may be vibration from the engines or external aerodynamic flow.

Anyway the FE model of the circuit board and avionics box housing are attached via rigid links to the base mass, which is several orders of magnitude higher than the total box mass. The FE software then converts the base acceleration to an equivalent force which is applied to the mass.

The response of the circuit board to the base acceleration can then be calculated. This can be done as a frequency response function or as a modal transient solution.

In addition, a separate case can be run as a normal modes analysis without the base to obtain the participation factors and effective modal mass values.

The software I use is FEMAP with NE/Nastran.

This same method should also work for earthquake analysis.

Tom Irvine

http://www.vibrationdata.com

 

[damstructural]

I like what I'm hearing that this method has a successful history and its still being used. This is an eye-opener for me about how other engineering disciplines solve this problem differently. Any other ideas on ways to validate the ground motion input?

I like Tom's idea about making a run without the large mass, but I'm wondering if ANSYS needs an input excitation to provide a frequency analysis. That would seem to be a good check on the effect of the large mass.

I'm also going to be looking at the acceleration output history at the foundation-structure interface nodes to see how the force has been transmitted to the structure from the "Black Hole." By the way, the foundation is massless, which is typical for FE earthquake engineering.

Any other ideas out there on model validation?

Great responses. Very helpful.

 

[Drej]

"I don't think I'd call this method "standard" for civil/structural analyses, which is why I posted. The more "standard" method for buildings, bridges, and dams is to input the time history directly at the restraints, as an acceleration or displacement."

LMM is a STANDARD METHOD OF USING AN ACCELERATION TIME-HISTORY INDIRECTLY WHEN THE ACCELERATION CANNOT BE INPUT DIRECTLY [TO THE PROGRAM]. I have used this method successfully on many, many occasions for dynamic transient, as well as harmonic analyses for a wide range of industries (nuclear, automobile, aerospace...).

"How sensitive is the magnitude of the "Black Hole" mass that is used. You cite 1E+05, they're using 1E+04. When does a problem with roundoff error start to kick in, and how do you calculate mass participation ratios with this monster mass?"

This is difficult to be precise about, but 1E5 is a (relative) nominal figure I have used successfully in the past. Some people use less, others more, but of course, if you go higher you need to be aware of numerical problems caused by this; go lower and you're no longer inputting acceleration but force itself. Look carefully into the results to see this. Mass participation should be calculated and output by the software without a problem.

 

[GregLocock]

One way of looking at the size of the seismic mass (as we'd call it) is to think about the effect on the modal frequencies. The first mode will be softened by m^-.5 , ie if you used a mass ratio of 100 then you'd get a 10% error in your first mode's frequency. I can see why you'd want to use at least 10^3 times the mass of the structure.

Of course this large mass element does introduce numerical problems for the solver, so the accuracy of the frequencies needs to be considered against that.




Cheers

Greg Locock

 

[Eap2n]

There is a section in the Basic Dynamics book of the NASTRAN manual which describes the large mass method and it recommends that the "dummy" mass be 10^6 times the mass of the entire structure.
Erik

 

[136695]

I am trying to do a seismic analysis on a wall mount rack with the large mass method using Ansys. I am concerned about the location (X,Y,Z) position and magntude of the point mass. I am trying to excite it by 5G. Please help me

 

[136695]

HI,
I am trying to do a seismic analysis on a wall mount rack with the large mass method using Ansys. I am concerned about the location (X,Y,Z) position and magntude of the point mass with respect to the part. I am trying to excite it by 5G. Please help me

 

[tomirvine]

The location of the base point mass is rather arbitrary.

The key is that the base mass is connect to the base nodes of the structure via one or more rigid links. The degrees-of-freedom for the rigid links are specified by the user.

Tom Irvine

http://www.vibrationdata.com