Random Vibration PSD vs Sinusoidal

[SBKroske]

1) Is it possible to convert a sinusoidal vibration profile into a random PSD curve?

2) For an automotive transmission application, does this PSD curve look excessive?

Hz g^2/Hz
4 0.1
20 0.1
100 3.2
160 4.0
220 1.1
800 0.1
2000 0.01

Thanks for any advice.

 

[Tobalcane]

SBKroske

Let me note first that I’m not an “expert” or a professor of vibration, just a mechanical engineer (electronic packaging field) who does a lot of vibration for my projects. Please if others think that I’m wrong, correct me.

<<<1) Is it possible to convert a sinusoidal vibration profile into a random PSD curve?>>>

Well let me ask, why do you want to convert a sinusoidal vibration profile into a random profile? I know that on my vibration table, it can run a sinusoidal profile (you just have to pick that option) or down load a random vibration profile. You don’t have to convert just pick the profile you want to run. Sinusoidal profiles are initially used to simulate rotating oscillating machines such as wheels, engines, gears…etc. Random profiles are usually used for acceptance test, screening out problems out of a unit during production (ESS), simulate transportation…etc. But, mostly it is a good simulator of the real world situations. In my field, I usually design my unit to with stand the random profile. If the unit does well during random vibration, it will do well in the field.

<<<2) For an automotive transmission application, does this PSD curve look excessive?

>>>

You have to find the “Grms in” (square root of the area under the curve) and determine if your design can handle the dynamic load. Since your big spike is at 160 Hz, you have to find out if any of your devices have a natural frequency close to 160 Hz. If so, you have to make sure that the device can withstand the G loads and have enough sway space so that it will not bump into other devices.

Well that’s my take.

Tobalcane

 

[marchino]

Hello to everybody,
I'm not an expert on vibration but I'm working on some specification requiring vibrational tests.
It's possible to compare a random profile with a sine one?
I want to find which is the more aggressive.

thanks,
Marco

 

[kaiserman]

SBKroske

(1)
As I'm sure you are aware there is no true random vibration equivalent for sine vibration, for multiple-degree of freedom systems. For a single degree of freedom system it is possible to generate equivalency based on a damage or damage potential.

A good article shows how to establish a random vibration profiles from swept sine tests. The article is a good place to start. Similarities in damage potential are considered in the article. (Machine Design/February 6, 1992). I do not have an electronic copy of this article, but if you would like I could scan it to a file and then e-mail it to you. Just let me know your e-mail.

Sine testing in our lab is a great analysis tool (estimate natural frequencies and transmissibilities) whereas random is a better durability tool. However, our company often uses "sine-on-random" testing for electronic components mounted to automotive engines and on transmissions. The road input vibration is random in nature, whereas the engine sine component is too strong to ignore. Therefore, "sine-on-random".

(2) In regards to your listed random vibration profile, the "excessive" question depends. Typically, in our lab we take field vibration data off of various products on several courses as well as nuetral run-up vibration. We then crunch the data and establish life profiles. However it typically becomes to cost prohibitive to test components about all profiles or for the full life duration (100,000 miles or 150,000 miles). In our lab we encorporate two test exaggeration and acceleration techniques. First we focus on the greatest damage potential courses, such as rough road - brick, gravel, etc., pot holes, RR Xings etc. Very low damage potential exists for city and highway courses so we typically do not subject the components to these environments. Right off the bat then, the test times are much shorter.

Next we then follow test acceleration techniques listed in MIL-STD-810. Basically we exchange stress for time. Thus, we increase our random vibration test profile levels and shorten the test times. This allows for very manage test times.

Some words of caution. Whenever tests are accelerated the fatigue material exponent must be known exactly. A DOE can be performed to establish this value. Also, random vibration testing profiles can only be increased so much. At some point, you will be introducing overstress conditions. Typically you want to focus on fatigue type failures not overstress failures. Therefore you need to review the S/N (stress vs. cycles) material curves. These curves provide a good estimation of stress vs. fatigue cycles for a particular material. They may aid is establishing the amount a profile can be increased. Another word of caution. The S/N material curves are based in test samples and are distributions. We have found that these distributions can be very broad distributions.

Therefore in answer to your "excessive" question, I have a few questions. Is the random vibration profile you have listed from a spec? Or, is the profile based on collected field data? Or is it dyno data? Also, how long will you subject the transmission to this profile? Is it an accelerated profile? All of these questions will help determine just how sever the profile may be.

I hope this helps in some small way.

Kaiserman

 

[frankandrews]

I think there are a couple reasons why you would not want to use a sine test as an alternate to random vibration:

1) Random vibration has amplitudes which vary with time. At each frequency, the amplitude varies based on a statistical distribution. If you were to determine a sine input at a given frequency, you would have to consider if you wanted to use a "one sigma" value or say, a "three sigma" value. A "one sigma" value has an amplitude 1/3 that of the "three sigma" value, but it occurs much more frequently. So, if you were to try a one sigma value, your equipment might pass, but you may have neglected the more significant amplitudes which could be more damaging. Alternatively, if you use a 3 sigma value, you may cause a premature failure or introduce unrealistic secondary effects, such as heat build-up.

2) In random vibration, all frequencies are excited at the same time. This means that if you have two or more structural resonances, they could all be excited at the same time, and you could have damaging interactions which would not be evident on sinusoidal testing. Sinusoidal testing excites one frequency at a time.


I hope this helps. If your interested, I have an article that explains much of this in more detail. Contact me if you are.




Frank Andrews, PE
SVS-Engineering