3 Replies Latest reply on Dec 4, 2017 10:41 AM by Bill McEachern

    A question about RMS results

    Swme Eng

      Hello all,

       

          I just wanted to clarify my understanding of RMS results when running random vibration analysis. From what I have read, I understand the RMS results to be a statistical value that means the displacement values for a given input PSD represent one sigma of certainty within the normal distribution. Is this the correct breakdown of the meaning of these values? Can I also then say that if I multiply by 3x I would capture 6 sigma and thus know with roughly 99.7% confidence that the actual displacement values would be those computed or less?

       

      Any help getting a good lame-man's explanation or technical resource to read through would be great. I have read through Vibartion Analsys with SolidWorks Simulation 2015 by Kurowski and I have skimmed the COSMOS Companion but never really felt like either provided great explanations.

        • Re: A question about RMS results
          Bill McEachern

          that's pretty much it.

            • Re: A question about RMS results
              Swme Eng

              Bill,

               

                  One other questions, when setting up a random vibration study, I know you have to define a min/max frequency that corresponds to your input PSD, and then determine the number of points along that PSD to test by setting the number of frequency points. Is there some way to make sure that you aren't missing some local maximum value from the PSD? It seems especially applicable to isolated PSDs where you seem amplification on some lower frequency, but scalar reduction everywhere else.

                • Re: A question about RMS results
                  Bill McEachern

                  All the maximums happen at the natural frequencies. I would use something with low numbers of DOF so it runs quick and run it with various numbers of points and bias' and see how it changes your output quantities of interest. Best way to get a feel for it. To get a good response estimate between resonant peaks you need more points but too many generates a large data set and the generation of the output can be burdensome on the performance. I would test 5, 10 and 20 points and run one of them or more again with various bias parameter settings. The bias parameter squeezes the points around the resonant peaks and spreads them out in the valleys.