Your job relies on accurate fault detection
There is no doubt that the primary focus for most vibration analysts is the detection of rolling element bearing fault conditions. When a bearing fails unexpectedly it costs a great deal of money (downtime, secondary damage, etc.) and it is a black mark on your name and your department. For all the successes you may have achieved, missing just one bearing failure can set your reputation back months.
Then again, the opposite is also true. If you report that a bearing has a defect and must be replaced, yet it is found to be in good condition, you also don't look good. People lose confidence in your skills and in the technology.
So what is the solution? I guess that's obvious; don't make mistakes! If only it were that easy . . .
It's actually not that hard to detect faults . . .
Detecting rolling element bearing defects is not as difficult as it may seem-I bet you did not expect me to say that! With a good screw driver and frequent trips around the machines, most people would be able to detect that a bearing needs to be replaced. There are, of course, a few issues with this approach. First, it is hardly a safe practice. Second, the greatest benefits are achieved when the maintenance and production group have more than a few days warning that a machine needs to be stopped to replace a bearing.
The earlier the better
The challenge is to correctly assess the nature and severity of the defect and the life of the bearing. If you could confidently detect a bearing fault weeks or months before the bearing needs to be replaced, then the work can be planned to minimize the impact of the bearing change. You may even be able to extend the life of the bearing by correcting the root cause of the fault condition (unbalance, misalignment, poor lubrication, etc.).
There is some good news
The good news is that the design of rolling element bearings makes it much easier to detect fault conditions at an early stage. Thanks to the geometry of the bearing (and their unique "defect frequencies") it is easy to distinguish bearing vibration from other vibration generated by the machine. And thanks to the high frequencies generated in the early stage of wear, again it is easy to distinguish from other fault conditions. Armed with this information we just need to measure the vibration correctly and analyze the data correctly, and we can be very successful.
OK, I may have made it sound too easy. In truth there are a number of challenges to overcome. But understanding the challenges and their solutions is the key to success.
Key #1: Measure the vibration correctly
Assuming the goal is to detect the bearing fault as early as possible, the first thing you have to do is recognize that the way you measure the vibration is absolutely key to success. In the earliest stage of bearing wear, the frequency of the vibration is very high and the amplitude is very low. I don't care who you are; you just can't hear it. And if you use "conventional" sensor mounting techniques, you cannot capture these high frequencies.
Techniques such as ultrasound, Shock Pulse, enveloping (demodulation), Spike Energy and PeakVue are designed to detect high frequency, low amplitude vibration. Without getting into the details, these techniques work by first removing the high amplitude, low frequency vibration, then listening carefully to the high frequency vibration for the telltale signs of bearing wear, then transforming that vibration into a form that is easy to analyze. We'll examine these techniques more closely in the next article.
A little bit of background . . .
When a defect is first initiated, the surface of the bearing may not actually be damaged; the damage may be subsurface. Even when the damage does extend to the surface, the vibration generated is still weak. As the balls or rollers move around the bearing and there is contact at the point where the damage exists, two things will happen: there will be a shock wave (also called a stress wave); and the bearing may vibrate (or resonate). The shock wave ripples out from the point of contact very quickly. The vibration that results will be very weak, and thus difficult to detect.
We can calculate, or search bearing databases for the telltale frequencies: the ball-pass inner race frequency, ball-pass outer race frequency, ball (or roller) spin frequency, and cage (or fundamental train) frequency. If you can visualize the shaft turning inside the bearing, and the balls rolling around, there will be a fixed time between each impact. The time will be different depending upon where the bearing is damaged; on the inner race, the outer race, or on the rolling elements themselves.
The good news is that this frequency will always be non-synchronous; it will never be exactly 2.0, 3.0, 4.0 (or any other integer) times the turning speed of the shaft. It will be a non-integer number such as 3.09, 6.71, or 11.43 times the speed of the shaft. That's the good news. It makes it easier to distinguish these sources of vibration from the numerous sources of vibration that occur at exact integer multiples of the running speed-from rotating elements such as pump vanes, fan blades, gear teeth, and so on.
More good news!
Another piece of good news is that when these impacts occur, the vibration that results is not smooth; the vibration will suddenly spike in amplitude before it settles again. That causes harmonics to appear in the spectrum. And even more good news is that under certain conditions the amplitude of those spikes will rise and fall (as a spall on the inner race of the bearing, or the damaged rolling elements, move in and out of the load zone). That causes sidebands to appear in the spectrum.
All of these telltale signs, which we can look for even if we do not know which bearing is installed in the machine, provide an early warning that the bearing is damaged. We can look for these signs well before we would ever hear a change in vibration, even with the best screwdriver.
Some bad news . . .
The bad news is that in the earliest stage of bearing wear, we will not be able to see peaks at these telltale frequencies in the standard velocity spectrum when it is displayed in the linear format that most people use. (In truth, we might see them in a logarithmic spectrum or in an acceleration spectrum.) So we have to look elsewhere. And that's where enveloping (and the other techniques listed previously) can be put to good use.
So, what is the solution?
There are a few ways to tackle this challenge.
1. There are simple meters that focus on higher frequencies can be used to get an indication that a fault exists. However, other fault conditions can be confused with bearing faults.
2. Shock Pulse meters are specifically designed to detect bearing faults. When used properly they provide an affordable way to get started.
3. Ultrasound meters allow you to listen for the presence of high frequencies. It is possible to detect lubrication problems and bearing defects. Again, they offer an affordable way to get started.
4. If you rely on "standard" velocity spectra (in linear format) then you will find it difficult to detect the fault until the fault has become more severe. Switching to log can help, and using units of acceleration and setting a higher Fmax will help.
5. The best solution is to use more sophisticated techniques such as enveloping (also known as demodulation), Shock Pulse (with access to the spectra and time waveforms), Spike Energy, and PeakVue. We'll discuss these techniques in the next article.
In the next article we will explore the techniques described in item 5 above. I hope this article has given you a better understanding of how the vibration changes as a defect grows, and the challenges involved with detecting the fault at the earliest stage.
Jason Tranter is the founder of Mobius Institute and author of iLearnVibration and other training materials and products. Jason has been involved in vibration analysis in the USA and his native Australia since 1984. Before starting Mobius Institute Jason was involved in vibration consulting and the development of vibration monitoring systems. www.mobiusinstitute.com