How much effective and indicative is a response spectrum in a mechanical component ? 1

[sftvbr]

Hi All

As you most probably know, there are two analysis system types in Ansys software: response spectrum and random vibration. ( these could have other names in other softwares but these are the basis of my question)

reponse spectrum is normally used for siesmic analysis but I have not seen its usage in mechanical components under random load. Also the same discussion about random vibration: how indicative are they and how could they be used in an analysis before acquisition test that under random load will give random displacements ?

 

[Erik Panos Kostson]

These are all what is called a spectral response analysis, or response spectrum analysis (both names - an other- are used within FEA).

Now the input can be a response spectrum curve (used often is seismic analysis, and this spectrum is not random), or a PSD spectrum curve used typically in testing of components, this (PSD) is perhaps called random vibration in ansys, since this response, unlike the response spectrum in seismic analysis, is random.

Typically I would imagine,(I have not done much random vibrations, just taught this in FEA courses using some examples), one might do some tests, and find out that there is a problem with a component (), then of course since you have the PSD (tests, or from real life measurements), it is hopefully much easier and hopefully faster to redesign and reanalyse it in a FEA software/system (that is virtually, rather then doing several redesigned prototypes and testing), and then when a final design is obtained through the analysis, then do another (final_) physical test to validate the FEA.

In the random vibration, the most interesting (since typically displacements are used in civil say for Servic. Limit. State analysis where they ca not be say larger then Length/300 depending on the different codes) I believe are the stresses which are in this case statistical thus the stresses are typically given within 1 sigma stdv. (say 10 MPa as an example), thus 68.3% of the time they will be as high (e.g., here 10 MPa) per the analysis or lower (one can of course increase it to 2 sigma stdv, and then multiply the stresses by 2, so the stresses will be 20 MPa or lower for 95 % of the time).

 

[sftvbr]

Thank you so much