What is the best criteria in calculating the frecuency cutoff value?

I know that for each of the natural frequencies (eigenvalues) corresponds a normal mode or natural state of vibration (eigenvectors). Any N-DOF system will have a total of 6xp degree of freedom (N = 6 x numer of node points).

For piping systems, I'm aware of the rule of thumb that the first 15% of a model's available modes are usually quite accurate in descrpting the system responce and the remaining 85% are poorer mathematical approximation.

[ecuation 1]= Total available modes = # of nodes or mass points x translational DOF - rigidly restrained translational DOF.

In others words, we can set the "Max No of Eigenvalues Calculated" as ecuation 1 x 15% and set 0 for "Frequency Cutoff" in the Control Parameters tab for dynamic analysis dialog box.

But, I'm looking for a good criteria in calculating the "Frequency Cutoff". For example, I've found that for dynamic water hammer analysis fcutoff= (SQRT(E/pipe mat den))/L, where L is the lenght of a single pipe element (elbow-elbow pair) in the primary run that is to have accurate stresses computed due to the passing of the water hammer.[From AutoPipe Advanced Training Outline]

Another criteria is the benchmark frecuency of 33 Hz as recomended in earthquake design.

I've found three more formulas for frequency cutoff estimation, based on pipe system total length, total lenght on water filled steel pipes and for empty pipes.

I'm setting a MathCAD file for Mass Lumped Spacing where I'm using the equation 13.3 of L Peng, 2009, Pipe Stress Engineering, ASME, New York, NY which is funtion of fcutoff. My objective is to generate a consolidated Lumped Mass Spacing Table.

If anybody know a good criteria in determinig fcutoff I'll apreciate the contribution.

Regards,
Edgar Perez