Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations TugboatEng on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
Ball Bearing Order Tracking
- Thread starter tomhoule
- Start date
There is a short article in the supplement to Energy Tech magazine,February 2005 (issued by ASME Power Division)titled "Detecting an AC Motor Bearing Problem" by Alan Friedman which depicts order tracking diagnostic signatures for an inner race fault. It notes that bearing tones often appear at frequencies between three and twelve times the shaft rate and that they rarely show up as integer multiples of the shaft rate but rather as mixed fractional values of shaft rate(1X) such as 4.6X, 7.2X, 5.8X. When accompanied by exact integer shaft rate harmonics, they may be difficult to separate out. By demodulating the signal, integer shaft rates can be eliminated leaving only the non-integer bearing fault peaks. These bearing peaks may be amplitude modulated by the fault, in which case, rotation-rate-spaced peaks may also show up as sidebands around the bearing defect peaks. Integer multiples of the bearing defect peaks may also show up which are fault harmonics.
MachineryWatch
Mechanical
Alan's article is correct in how it relates the frequencies generated by the bearing defects to the "orders of rotation" of the shaft. Is that the kind of information you are seeking? There is a signal processing method known as "order tracking" which incorporates a reference signal into the processing of the vibration data. This is typically a 1XRPM pulse that is used to create a "phase locked loop". The instrument then generates a sample clock based on this pulse train. The spacing of the pulses vary with the changing speed of the shaft and so does the sample rate used for sampling and digitizing the signal from the vibration sensor. The machine's variation in speed also results in vibration that varies in frequency. A time block of digitized data, that is acquired conventionally, with a fixed sample clock controlling the rate at which the data is sampled, is fine when the machine is operating at a fixed speed, but when the speed varies and this digitized data is transformed to the spectrum, the energy from each spectral peak may be "smeared" over several lines of resolution. This has the effect of lowering the amplitude of the peak beacuse the energy is spread into several FFT "lines" or bins. When the sample rate is varied over time with the shaft speed actually controlling the sample rate the resulting Fourier Transform has the energy from each frequency contained within a single FFT line, so all the amplitude is represented. This is true of the frequencies that are between the orders of rotation, as well, such as bearing defect frequencies. If you can see the true amplitude generated by the defect you have a much better idea of the severity of the defect.
We used to use this technique on Navy shipboard turbine driven pumps and blowers because their governors often did not control speed very precisely. Rolling element bearings did not always create a lot of energy when they have defects and if the speed variations during data acquisition created "smearing" the bearing defects were sometimes lost in the noise. This was in the days when all the data was tape recorded before being "processed" so we needed all the help we could get. Even though our real time analyzers had 72 db dynamic range, the analog tape recorders only had about 40 dB dynamic range, so it was important to preserve any of the amplitude near the noise floor that we had, and "Order Tracking" helped us do that. The data was used for planning an overhaul period that was 6-12 months in the future so detecting a developing bearing fault was important.
When using this technique and setting up for data acquisition you define the FFT range in number of "orders" rather than a frequncy range like 0-1000 Hz or 0-60,000 CPM. Typically, you would define an FFT range as 0 - 20 orders and maybe a second higher frequency range such as 0 - 60 orders, depending on the characteristics of the machine. The problem was that you need to have the tach signal, so setup time to get the data was longer. Sometimes when we could not record a tach signal we could generate one from the vibration signal if the 1X RPM vibration signal was strong enough and we band-pass filtered it so that the 1XRPM signal actually appeared to the instrument to be a pulse train that it could track in the tach input channel.
I hope this helps for your application.
Skip Hartman
We used to use this technique on Navy shipboard turbine driven pumps and blowers because their governors often did not control speed very precisely. Rolling element bearings did not always create a lot of energy when they have defects and if the speed variations during data acquisition created "smearing" the bearing defects were sometimes lost in the noise. This was in the days when all the data was tape recorded before being "processed" so we needed all the help we could get. Even though our real time analyzers had 72 db dynamic range, the analog tape recorders only had about 40 dB dynamic range, so it was important to preserve any of the amplitude near the noise floor that we had, and "Order Tracking" helped us do that. The data was used for planning an overhaul period that was 6-12 months in the future so detecting a developing bearing fault was important.
When using this technique and setting up for data acquisition you define the FFT range in number of "orders" rather than a frequncy range like 0-1000 Hz or 0-60,000 CPM. Typically, you would define an FFT range as 0 - 20 orders and maybe a second higher frequency range such as 0 - 60 orders, depending on the characteristics of the machine. The problem was that you need to have the tach signal, so setup time to get the data was longer. Sometimes when we could not record a tach signal we could generate one from the vibration signal if the 1X RPM vibration signal was strong enough and we band-pass filtered it so that the 1XRPM signal actually appeared to the instrument to be a pulse train that it could track in the tach input channel.
I hope this helps for your application.
Skip Hartman
electricpete
Electrical
What is NVH?
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
MachineryWatch
Mechanical
Noise Vibration and Harshness (NVH).
I think it is mostly a transportation industry term.
Skip Hartman
I think it is mostly a transportation industry term.
Skip Hartman
More specifically, Noise/Vibration/Harshness (NVH) generally refers to passenger automobile noise/vibration/customer annoyance control and is highly oriented toward customer perceived conditions and customer satisfaction issues. Besides employing advanced airborne and structureborne analysis and measurement techniques like Computational Fluid Dynamics(CFD) and modal analysis for passenger compartments, body frames, engines, tires,etc., it has spawned an intricate "jury" process of human qualitative assessments of perceived annoyance measuring scales. Since the automotive industry seems to be funding and/or doing possibly 75-90% of all advanced noise/vibration work worldwide, NVH is now a "big" topic in journals and magazines dealing with noise/vibration technology. For example, the prestigious Sound and Vibration magazine (a freebe for qualified engineers) sometimes devotes whole issues to NVH topics.
".....all the amplitude is represented. This is true of the frequencies that are between the orders of rotation, as well, such as bearing defect frequencies. If you can see the true amplitude generated by the defect you have a much better idea of the severity of the defect. "
I think with extended time synchronous averaging the amplitude of the non-integer harmonics like 3.67X, 11.23X, etc average downward, and the true 1,2,3X etc remain.
I think with extended time synchronous averaging the amplitude of the non-integer harmonics like 3.67X, 11.23X, etc average downward, and the true 1,2,3X etc remain.
MachineryWatch
Mechanical
Tmoose,
The process I described (Order Tracking) is not time synchronous averaging (TSA). You are right that TSA reduces the amplitude of any non-synchronous peaks in the FFT, but TSA is a very different process, than Order Tracking.
TSA essentially times the start of each digitized time block of data acquired so that each block of time data is synchronized with some event. This event is usually the passing of some triggered part of a rotating shaft, but it can be some other event like a trigger based on the AC power supplied to a motor. The multiple blocks of digitized time domain data acquired are then averaged. The block of data that is the result of the averaging is then converted to an FFT that typically only shows harmonics of the trigger event.
The Order Tracking (OT) is quite another process. It does not eliminate the non-synchronous frequencies from the FFT. With OT, the "smearing" associated with a normal FFT captured on machines that are varying in speed results in a much reduced amplitude displayed in the spectrum because the energy at a specific "order of shaft rotation" keeps moving across several lines of FFT resolution. Order tracking varies the sample rate of the analyzer/digitizer at a rate linked to the shaft speed. This results in specific number of samples per shaft rotation rather than a specific number of samples per second. The FFT yields frequency energy that is locked into the same FFT lines instead of smearing across several FFT lines, so it is a much better representation of the true amplitude, even if it is a non-synchronous frequency.
Because the data must be captured with a speed trigger many people make the mistake of thinking OT is the same process as TSA. It most definitely is not the same!
Skip Hartman
The process I described (Order Tracking) is not time synchronous averaging (TSA). You are right that TSA reduces the amplitude of any non-synchronous peaks in the FFT, but TSA is a very different process, than Order Tracking.
TSA essentially times the start of each digitized time block of data acquired so that each block of time data is synchronized with some event. This event is usually the passing of some triggered part of a rotating shaft, but it can be some other event like a trigger based on the AC power supplied to a motor. The multiple blocks of digitized time domain data acquired are then averaged. The block of data that is the result of the averaging is then converted to an FFT that typically only shows harmonics of the trigger event.
The Order Tracking (OT) is quite another process. It does not eliminate the non-synchronous frequencies from the FFT. With OT, the "smearing" associated with a normal FFT captured on machines that are varying in speed results in a much reduced amplitude displayed in the spectrum because the energy at a specific "order of shaft rotation" keeps moving across several lines of FFT resolution. Order tracking varies the sample rate of the analyzer/digitizer at a rate linked to the shaft speed. This results in specific number of samples per shaft rotation rather than a specific number of samples per second. The FFT yields frequency energy that is locked into the same FFT lines instead of smearing across several FFT lines, so it is a much better representation of the true amplitude, even if it is a non-synchronous frequency.
Because the data must be captured with a speed trigger many people make the mistake of thinking OT is the same process as TSA. It most definitely is not the same!
Skip Hartman
MachineryWatch
Mechanical
Speaking of S&V magazine.... there is a good article in the new March 2005 issue. Its title is "Main Principles and Limitations of Current Order Tracking Methods"
Skip Hartman
Skip Hartman
And there's a paper to be delivered in May at the SAE Noise & Vibration Conference entitled "Order Separation Using Multiple Tachometers and the TVDFT Order Tracking Method."
Sounds interesting, but coming from Michigan Tech, it may or may not have any practical application!.
LOL!
Sounds interesting, but coming from Michigan Tech, it may or may not have any practical application!.
LOL!
- Status
Similar threads
- Question
- Locked
- Question
- Locked
- Question
