Eng-Tips is the largest engineering community on the Internet
Intelligent Work Forums for Engineering Professionals
-
Congratulations waross on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
relative shaft displacement vs bearing housing vibration
- Thread starter kancil
- Start date
electricpete
Electrical
I think that for ball bearing fault detection bearing housing accelerometer measurements are best. Also they are much easier. I haven't ever heard of anyone doing prox probe shaft measurements on ball bearing machine.
Prox probe measurements are generally only performed on fluid-film bearings. In that case shaft vibration is not as easily transferred to the housing and prox probe may give more info about shaft movement within the bearing.
Prox probe measurements are generally only performed on fluid-film bearings. In that case shaft vibration is not as easily transferred to the housing and prox probe may give more info about shaft movement within the bearing.
- Thread starter
- #3
Actually i have conducted experiments to detect ball and roller bearings fault on the raceways and rolling element via measurement of relative shaft displacement. The results are graphically plotted using timebase plot.
The outcome of the experiment showed that positive or negative spikes or both were detected on the plot depend on the fault location relative to probe tip. It seems that this method can be used to evaluate the bearing condition because spikes interval and its direction (positive or negative)indicated the specific bearing fault.
Can any explain for me why this method only used for journal bearing and not used to detect rolling element bearings fault.
The outcome of the experiment showed that positive or negative spikes or both were detected on the plot depend on the fault location relative to probe tip. It seems that this method can be used to evaluate the bearing condition because spikes interval and its direction (positive or negative)indicated the specific bearing fault.
Can any explain for me why this method only used for journal bearing and not used to detect rolling element bearings fault.
dgallagher
Mechanical
The concept of measuring the deflections of a rolling element bearing outer ring during operation was originally demonstrated by Shapiro of the Franklin Institute using strain gages attached to the bearing's outer ring. This concept was further improved upon by G. J. Phillips of the David W. Taylor Naval Research and Development Center using noncontact fiber optic techniques. Bently Nevada Corporation adopted this noncontact concept and, as an alternative to the fiber optic sensor, developed a high-gain, low-noise eddy current proximity transducer to measure the deflections of a rolling element bearing's outer ring.
In practicallity, I would say that it is a very expensive, less reliabile method than using an Accelerometer externally on the bearing cap. Not many end users are willing to drill their bearings to the outer ring due to associated cost and possible fatigue of the bearing. Field experience has shown that traditional vibration analysis and demodulation techniques are preferred on rolling element bearings.
As Pete mentioned above, Prox probes are valid sensors in a journal bearing application, but even then, an Accelerometer may be a sufficient cost effective solution - For a heavy rotor in a light casing like an electric motor, a case mounted accelerometer will provide sufficient information. For a light rotor in a heavy case, a case mounted accelerometer would be deficient.
Cheers,
Dave G.
In practicallity, I would say that it is a very expensive, less reliabile method than using an Accelerometer externally on the bearing cap. Not many end users are willing to drill their bearings to the outer ring due to associated cost and possible fatigue of the bearing. Field experience has shown that traditional vibration analysis and demodulation techniques are preferred on rolling element bearings.
As Pete mentioned above, Prox probes are valid sensors in a journal bearing application, but even then, an Accelerometer may be a sufficient cost effective solution - For a heavy rotor in a light casing like an electric motor, a case mounted accelerometer will provide sufficient information. For a light rotor in a heavy case, a case mounted accelerometer would be deficient.
Cheers,
Dave G.
Prox probes in general start to "roll off " as far as frequency response @ 1 kHz. Accelerometers on the other hand, have a much higher frequency response up to 30kHz for some high frequency accel's. Bearing degradation normally shows up at the higher end of the spectrum 20 to 60 kHz. Ultrasonic detection or stress wave detection circuits ( Peakvue, Spike Energy, and demodulation to name a few) will pick up defects well before traditional vibration measurements. Once bearing faults start to show up in the normal spectrum, it's recommended that periodic vibration readings be trended as you have probably reached stage 2 of bearing failure. As stated in the above post's prox probes are normally utilized to measure shaft displacement not bearings.
Regards,
MICJK
Regards,
MICJK
- Thread starter
- #7
The experiments i conducted are measurement of displacement between the shaft surface and the probe tip by using proximity probes. The readings are not taken through REBAM probes that observed the bearing outer ring deflection.
The outcome of the experiment proved that the measurement of the relative shaft displacement can indicated specific bearing faults at their early stage. My quention is whether this technique has been investigated by other researches and how is their outcome or conclusion?
The outcome of the experiment proved that the measurement of the relative shaft displacement can indicated specific bearing faults at their early stage. My quention is whether this technique has been investigated by other researches and how is their outcome or conclusion?
electricpete
Electrical
The traditional housing accelerometer particularly with demod processing can easily detect bearing defects in the very early stages long before any defect would be visible on a race and long before most people would replace them.
I doubt you can improve on the sensitivity (as Mick mentioned frequency limitations would be wokring against you), but even if you can that really adds no benefit since as I mentioned those defects already are visible long before we call them bad.
What benefit is there assuming you could somehow improve sensitivity unless you can do it at reduced cost or complexity (which is doubtful). What type of application do you have in mind?
I doubt you can improve on the sensitivity (as Mick mentioned frequency limitations would be wokring against you), but even if you can that really adds no benefit since as I mentioned those defects already are visible long before we call them bad.
What benefit is there assuming you could somehow improve sensitivity unless you can do it at reduced cost or complexity (which is doubtful). What type of application do you have in mind?
- Thread starter
- #9
Thanks for your reply.
The measurement of relative shaft displacement is direct and not influence by other vibration sources. As the experimental data analyzed in time domain (timebase plot), only the displacement of the shaft due to bearing faults are identified. If there is other vibration sources, the direction (positive or negative) and interval of the spikes are different.
I also have conducted experiment to detect REB fault throught the bearing housing vibration with accelerometer. The data analyze with envelope analysis, which showed that the fault can be identified easily. But this method also easily influence by other vibration sources which lead to misleading the result.
So the relative shaft displacement method may be can be used with other fault detection methods together to identify the bearing fault.This is because the more information we have will provided us to do a good decision.
The measurement of relative shaft displacement is direct and not influence by other vibration sources. As the experimental data analyzed in time domain (timebase plot), only the displacement of the shaft due to bearing faults are identified. If there is other vibration sources, the direction (positive or negative) and interval of the spikes are different.
I also have conducted experiment to detect REB fault throught the bearing housing vibration with accelerometer. The data analyze with envelope analysis, which showed that the fault can be identified easily. But this method also easily influence by other vibration sources which lead to misleading the result.
So the relative shaft displacement method may be can be used with other fault detection methods together to identify the bearing fault.This is because the more information we have will provided us to do a good decision.
dgallagher
Mechanical
At some point you have to address the real world and get out of the laboratory. An accelerometer costing $100.00 (and requiring no special mounting) will let you know well in advance of a pending bearing problem and a $500.00 proximity probe system (without installation cost) may help verify what I already know from the accelerometer from years of documented experience and success. Successfull implementation of any technology generally means reducing cost and time, not adding to it.
electricpete
Electrical
Your complaint against bearing housing accelerometer measurements as I understand it was
"I also have conducted experiment to detect REB fault throught the bearing housing vibration with accelerometer. The data analyze with envelope analysis, which showed that the fault can be identified easily. But this method also easily influence by other vibration sources which lead to misleading the result."
In a typical industrial machine condition monitoring program with route-based data collection, a high demod magnitude would prompt the analyst to look at the frequency content of the demod spectrum and/or regular spectrum. There will be a distinctive pattern of bearing fault frequency harmonics which certainly cannot be confused with any non-bearing fault. There may be some false alarms on demod magnitude from other non-bearing impacting (for example impacting at rotational speed or looseness). These can likewise be identified by their frequency signature, and they will not contain bearing fault frequency harmonics.
It sounds like interesting research. But as was mentioned, it seems like a cheap, easy, sensitive means for bearing fault detection is already available in housing accelerometer measurements.
"I also have conducted experiment to detect REB fault throught the bearing housing vibration with accelerometer. The data analyze with envelope analysis, which showed that the fault can be identified easily. But this method also easily influence by other vibration sources which lead to misleading the result."
In a typical industrial machine condition monitoring program with route-based data collection, a high demod magnitude would prompt the analyst to look at the frequency content of the demod spectrum and/or regular spectrum. There will be a distinctive pattern of bearing fault frequency harmonics which certainly cannot be confused with any non-bearing fault. There may be some false alarms on demod magnitude from other non-bearing impacting (for example impacting at rotational speed or looseness). These can likewise be identified by their frequency signature, and they will not contain bearing fault frequency harmonics.
It sounds like interesting research. But as was mentioned, it seems like a cheap, easy, sensitive means for bearing fault detection is already available in housing accelerometer measurements.
- Thread starter
- #12
MachineryWatch
Mechanical
I guess I would have to chime in with what would be considered an additional negative for the prox probe/shaft displacement approach. Even when you consider it from strictly the standpoint of permanently installed systems in situations where initial cost of installation is not a factor the prox probe approach for rolling element bearings is not acceptable from a reliability standpoint. Prox probe systems are somewhat unreliable. Invariably, when asked to do testing on a machine that was originally supplied from the manufacturer with proximity probes, I find they haven't been calibrated in years and the levels are significantly incorrect. Many times work is done on the machine and shaft target areas are damaged causing poor/incorrect data. Or the shaft picks up some residual magnetism that creates substantially incorrect data on the prox probes. Even if you have a system for trending the data from the probes at specific bearing related frequencies the trends would be unreliable without constant calibration checks.
An accelerometer based on-line monitoring system does not have these difficulties. It is pretty easy to look at the data from an accelerometer and tell if the accelerometer is good, or visually inspect the externally mounted accelerometer and tell if it is properly mounted, etc. This examination is not so easy with a proximity probe.
For me this reason alone would be enough to steer me away from a shaft displacement based system for rolling element bearings, even if it were an on-line system. Don't mistake what I am saying as meaning that I think prox probes are never helpful. I think there are many good applications for them.
Skip Hartman
An accelerometer based on-line monitoring system does not have these difficulties. It is pretty easy to look at the data from an accelerometer and tell if the accelerometer is good, or visually inspect the externally mounted accelerometer and tell if it is properly mounted, etc. This examination is not so easy with a proximity probe.
For me this reason alone would be enough to steer me away from a shaft displacement based system for rolling element bearings, even if it were an on-line system. Don't mistake what I am saying as meaning that I think prox probes are never helpful. I think there are many good applications for them.
Skip Hartman
- Thread starter
- #14
Can you explain what you mean by "I find they haven't been calibrated in years and the levels are significantly incorrect".
If the shaft is damaged during operation, this will caused an abnormal reading to the prox. probe, which can indicated this damage as a problem in the other machine components but not the bearing.I think this will provide us the information about the machine condition, where a corrective action can be taken. A damage shaft may be caused unbalance to the machine system.
To detect the bearing fault in time domain, only the properties of the bearing and the shaft speed are need to determine the spikes interval and their direction (positive or negative).The prox. probe only measure directly the relative displacement of the shaft, which is caused by the faulty bearing.
If the shaft is damaged during operation, this will caused an abnormal reading to the prox. probe, which can indicated this damage as a problem in the other machine components but not the bearing.I think this will provide us the information about the machine condition, where a corrective action can be taken. A damage shaft may be caused unbalance to the machine system.
To detect the bearing fault in time domain, only the properties of the bearing and the shaft speed are need to determine the spikes interval and their direction (positive or negative).The prox. probe only measure directly the relative displacement of the shaft, which is caused by the faulty bearing.
As Skip stated, prox probes have there usefulness, but I don't feel as though bearing fault detection is one of them. When trying to detect bearing faults, most analysts are looking for signs of "impacting", not relative shaft displacement (motion). As I stated in my first reply, early bearing degradation shows up first at the ultrasonic level which will not be picked up using a Prox probe. This period is where an analyst would like to catch bearing faults. The bearing can then be monitored and trended. If I or any other analyst were to see a rapid upward trend in energy begining in the high frequency range then migrating to lower frequencies (typical of bearing iminent failure pattern) Then a judgement call can be made. Personally, I feel if you see an increase in relative shaft displacement associated with a bearing, then the bearing is on it's "last leg" and should be replaced ASAP. Even with sleeve bearings there are better methods for early detection. Oill analysis for example. When you see babbit material in your oil samples, dependent upon particle size and concentration, severity can be determined, and appropriate action taken, all before relative shaft displacement would have detected the fault.
Regards,
MICJK
Regards,
MICJK
MachineryWatch
Mechanical
Typically the shaft has not been damaged during operation. Rather, the machine has been disassembled for repair and during the repair some minor scratch is created during handling of the shaft by using a dirty sling to lift the shaft or some other reason. I have even seen it on new machines delivered from the factory. Shafts can pick up local residual magnetism from the use of large magnets when attaching devices used for run-out checks and alignment.
These problems are not detected when the machine is restarted because most users do not go completely through a thorough calibration procedure designed to detect and correct all these problems. The reason they do not go through the procedures is because either they do not know about the procedures, or they decide it is not worth the time and expense to do so. Runout (electrical or mechanical) can cause a proximity probe to read either higher or lower than the actual shaft displacement due to phase cancelation or phase addition. In either case it defeats the purpose of the installation by causing the machine protection system to trip too early (nuisance) or too late (catastrophic failure).
The frequency limitations of the proximity probe have already been pointed out and are significant. I have seen high speed centrifugal compressors with speed increasers run to failure because of worn gears. The machine was equipped with a PROPERLY CALIBRATED prox. probe system that never registered an abnormal level because gear mesh frequency was a higher frequency than the highest frequency the probes could detect. We watched the G levels go up for a couple months and told the machine owner that the machine had to be shut down. Two days before failure we read 42 g's acceleration at gear mesh frequency (normal was about 4 g's) yet they said they had to take a chance with it and continue to run it as long as possible. The displacement probes never even shut the machine down before failure!
The point is, why try to force the use of these probes into an area where we already have adequate technology? Accelerometers are very successful at detecting rolling element bearing problems in most applications even without any fancy signal processing. Not only do they do a good job, but they do not have a lot of the inherent problems of proximity probe systems. For the more challenging applications enveloping technology seems to adequately enhance the raw signal to provide early warning of developing problems. As many have said these problems can be detected often before the damage is visible when the bearing is inspected.
Apparently there are some applications of rolling element bearings you are involved with where the use of rugged, lightweight, inexpensive accelerometers cannot provide the proper detection of bearing wear. Perhaps if you shared with us the specific rolling element bearing application you had in mind we would understand better the need to adapt a different technology. For myself (and I am sure many others), the detection of rolling element bearing problems has been one of the least challenging problems in vibration analysis, except for a couple machinery types that present some specific challenges that require a little additional caution when acquiring data and care when analyzing the data.
Skip Hartman
These problems are not detected when the machine is restarted because most users do not go completely through a thorough calibration procedure designed to detect and correct all these problems. The reason they do not go through the procedures is because either they do not know about the procedures, or they decide it is not worth the time and expense to do so. Runout (electrical or mechanical) can cause a proximity probe to read either higher or lower than the actual shaft displacement due to phase cancelation or phase addition. In either case it defeats the purpose of the installation by causing the machine protection system to trip too early (nuisance) or too late (catastrophic failure).
The frequency limitations of the proximity probe have already been pointed out and are significant. I have seen high speed centrifugal compressors with speed increasers run to failure because of worn gears. The machine was equipped with a PROPERLY CALIBRATED prox. probe system that never registered an abnormal level because gear mesh frequency was a higher frequency than the highest frequency the probes could detect. We watched the G levels go up for a couple months and told the machine owner that the machine had to be shut down. Two days before failure we read 42 g's acceleration at gear mesh frequency (normal was about 4 g's) yet they said they had to take a chance with it and continue to run it as long as possible. The displacement probes never even shut the machine down before failure!
The point is, why try to force the use of these probes into an area where we already have adequate technology? Accelerometers are very successful at detecting rolling element bearing problems in most applications even without any fancy signal processing. Not only do they do a good job, but they do not have a lot of the inherent problems of proximity probe systems. For the more challenging applications enveloping technology seems to adequately enhance the raw signal to provide early warning of developing problems. As many have said these problems can be detected often before the damage is visible when the bearing is inspected.
Apparently there are some applications of rolling element bearings you are involved with where the use of rugged, lightweight, inexpensive accelerometers cannot provide the proper detection of bearing wear. Perhaps if you shared with us the specific rolling element bearing application you had in mind we would understand better the need to adapt a different technology. For myself (and I am sure many others), the detection of rolling element bearing problems has been one of the least challenging problems in vibration analysis, except for a couple machinery types that present some specific challenges that require a little additional caution when acquiring data and care when analyzing the data.
Skip Hartman
- Status
Similar threads
- Locked
- Question
- Locked
- Question
- Locked
- Question
