What is random vibration?

First of all, I want to talk about what is random vibration. Random vibration has two characteristics:

The first feature is "random" just like its name, and there is no way to know how it vibrates in the next second.

The second feature is that although it is random, there is a statistical law within a certain period of time (Gaussian distribution or normal distribution).

Therefore, randomness is not a specific vibration that can be represented by a certain function. When it comes to randomness, it is not a vibration at a point in time. It must be a vibration within a period of time to be called random.

After we know what random vibration is, let's learn about the control of random vibration.

In the process of random control, the controller first needs to generate a Gaussian random time-domain vibration signal, and then amplify the vibration table to generate Gaussian random vibration through the power amplifier, and perform PSD equalization and re-correction through sensor feedback. Because randomness has no regularity at a certain point in time, and only has a statistical regularity within a period of time, so we must use PSD equalization to control it.

Therefore, your control spectrum must be rough at the beginning of the PSD equalization process. If your controller is very convergent and smooth at the beginning, don't hesitate to change it! This controller must be fake!

So, the longer the time, the more convergent and smoother it is? The answer is not necessarily!

After starting to balance for a certain period of time, his convergence and smoothness will not change. This is because the convergence and smoothness are also related to an important parameter of the balance, that is, the degree of freedom DOF! I will not go around purely theoretical things. For example, if the definition of freedom theory is not very clear, you can use Baidu. Here you just have to remember that the larger the DOF value, the more frames involved in the PSD equalization, that is, the longer the statistical time for random vibration, so there is a table for everyone to look at. This table shows the free DOF and random spectrum. Very bad relationship. For example, when the degree of freedom is 120, the point that exceeds the +1dB tolerance should be 2.892% or more.

You can compare this table with your own controller spectrum. Last time I met a customer’s request that he has the conditions to control within 0.7dB and not have over-tolerance. The degree of freedom is 120. I said I can’t control it. This is a violation of the random control theory, but he said that they have The controller can be done, this time I can only hehe.

There is another question. I don’t know if you have paid attention to it during the experiment. When you start the test with small-scale equalization and then gradually increase the large-scale level, does the PSD spectrum jump directly from the small-scale level to the large-scale level without any re-equalization, or does your PSD spectrum change from the rough spectrum again? Started a new equilibrium?

If there is no rebalancing, then your product has probably been tested. For those high-value products such as satellites, spacecraft, etc., it is fatal. In the event of an accident, you still have no way to find out where the problem is.

Some controllers directly linearly proportionally amplify the low-level feedback signal as the feedback signal that enters the full-level test at the beginning during the random level increase process, instead of re-measurement feedback equalization. They made a big mistake by assuming that the response at the low level is consistent with the response at the full level. What happens if the mode is excited at a large number of levels? That is, the spectrum is still very good, all within the tolerance band, and it is relatively smooth and convergent, but the actual product vibration may have exceeded the limit.