Modal mass and kinetic energy in FEM modal analysis

AI Thread Summary
The discussion focuses on the use of Ansys for modal analysis of a clamped beam using 20-node brick elements. The user successfully computes effective mass and kinetic energy but is initially confused about the modal mass and kinetic energy calculations, which are derived from unit normalized modes. The correct formulas for modal mass and kinetic energy are clarified, indicating that the modal mass is calculated using the eigenvector and mass matrix, while kinetic energy is derived from modal mass and frequency. Additionally, the user inquires about calculating participation factors in a locally defined coordinate system, as Ansys outputs results in a global coordinate system. The conversation highlights the complexities of modal analysis in FEM and the importance of understanding normalization in calculations.
Arjan82
Messages
624
Reaction score
618
TL;DR Summary
What is the formula for modal mass and kinetic energy of a modal analysis FEM computation? I do *not* mean effective modal mass.
So, I use Ansys (well known FEM software) and get the next output for a modal analysis toy problem (If you happen to know Ansys that's a pre, but I promise it shouldn't matter). The problem is a simple beam, clamped at one end. I used 160 20-node brick elements to solve it (so no Timoshenko beams or something like that).

[CODE title="Ansys output"]
The modes requested are mass normalized (Nrmkey on MODOPT). However,
the modal masses and kinetic energies below are calculated with unit
normalized modes.

***** MODAL MASSES, KINETIC ENERGIES, AND TRANSLATIONAL EFFECTIVE MASSES SUMMARY *****

EFFECTIVE MASS
MODE FREQUENCY MODAL MASS KENE | X-DIR RATIO% Y-DIR RATIO% Z-DIR RATIO%
1 81.73 39.42 0.5199E+07 | 0.000 0.00 95.85 61.05 0.000 0.00
2 159.3 40.53 0.2030E+08 | 0.000 0.00 0.000 0.00 96.07 61.19
3 490.2 41.77 0.1981E+09 | 0.000 0.00 30.22 19.25 0.000 0.00
4 593.4 31.73 0.2206E+09 | 0.000 0.00 0.000 0.00 0.000 0.00
5 859.8 48.92 0.7138E+09 | 0.000 0.00 0.000 0.00 31.92 20.33
6 1268. 77.93 0.2472E+10 | 126.6 80.61 0.000 0.00 0.000 0.00
[/CODE]

So I know exactly how to get the effective mass, which is dependent on direction. And by exactly I mean exactly. I extract the mass and stiffness matrix from Ansys, compute the eigenvectors of this problem (with Matlab):

$$
\left[ K \right] \left\{ d \right\} = w \left[ M \right] \left\{ d \right\}
$$

with ##\left[ K \right] ## the stiffness matrix, ##\left[ M \right]## the mass matrix, ##\left\{ d \right\}## an eigenvector and ##w = \omega^2## the eigenvalue. All is 'mass normalized' such that ##\left\{ d \right\}^T \left[ M \right] \left\{ d \right\} = 1##. And now we can compute the participation factor for the x-direction (assuming mass normalization):

$$
L_x = \left\{ d \right\}^T \left[ M \right] \left\{ r_x \right\}
$$

with ##\left\{ r_x \right\}## the influence vector, or just a vector with 1's at all degrees of freedom of the x-direction and 0 everywhere else. And now the effective modal mass in the x-direction is simply ##L_x^2##. If I do this with e.g. Matlab I get exactly the same results as Ansys does (all digits are the same, except maybe the last). So, no problem there.

But the modal mass and kinetic energy (KENE) on the left side of this table are a mystery to me. I need the formula for that but cannot seem to find it, not in the documentation of Ansys, not on the internet (for which the results get swamped by explanations of effective mass) and not in any book about the subject I own. Who knows how to compute those numbers?
 
Engineering news on Phys.org
Allright, apparently the documentation of Ansys is also a mystery for me, the answer is just right in there 😆. Also, I'm overthinking things, as usual 🤔

The modal mass is simply

$$
m = \left\{ d \right\}^T \left[ M \right] \left\{ d \right\}
$$

And the kinetic energy

$$
KE = \frac{1}{2} m \omega^2
$$
 
Oh, and crucially, the ##\left\{d\right\}## vector is not mass normalized, but unit normalized (i.e. the max absolute value of ##\left\{d\right\}## is equal to 1), otherwise ##m## would just be 1.
 
How to get participations factors in a different co-ordinate system altogether?...The ANSYS output is always in global co-ordinate system. And, then you can get PF in X,Y,Z,Rx,Ry,Rz directions each. But, what if what I want to calculate PF in a locally defined co-ordinate system?
 
Posted June 2024 - 15 years after starting this class. I have learned a whole lot. To get to the short course on making your stock car, late model, hobby stock E-mod handle, look at the index below. Read all posts on Roll Center, Jacking effect and Why does car drive straight to the wall when I gas it? Also read You really have two race cars. This will cover 90% of problems you have. Simply put, the car pushes going in and is loose coming out. You do not have enuff downforce on the right...
I'm trying to decide what size and type of galvanized steel I need for 2 cantilever extensions. The cantilever is 5 ft. The space between the two cantilever arms is a 17 ft Gap the center 7 ft of the 17 ft Gap we'll need to Bear approximately 17,000 lb spread evenly from the front of the cantilever to the back of the cantilever over 5 ft. I will put support beams across these cantilever arms to support the load evenly
Thread 'What's the most likely cause for this carbon seal crack?'
We have a molded carbon graphite seal that is used in an inline axial piston, variable displacement hydraulic pump. One of our customers reported that, when using the “A” parts in the past, they only needed to replace them due to normal wear. However, after switching to our parts, the replacement cycle seems to be much shorter due to “broken” or “cracked” failures. This issue was identified after hydraulic fluid leakage was observed. According to their records, the same problem has occurred...
Back
Top