Where do ghost components in gearbox vibration spectra arise from and what condition-indicating information do they contain?

Spooky condition?

Is this some kind of homework, or are you looking for the answers of some vibration exam?

If you need help, please elaborate more on the subject, else you will not get much response

Wayne,

"Ghost" frequencies are usually associated with tooth flank surface pattern caused by hobbing (cutting) gear. It could also be caused by ginder if excessive vibration (unbalance/eccentricity) was present during grinding operations. In either case, the presence of Ghost frequencies usually indicates problems with the tooth flank surface quality and manufacturing issues.

Walt

Thanks all. Here are some words I've put together.

A ghost frequency is a gearbox vibration frequency, which does not relate to the geometry of the gearbox. 'Ghost' frequencies are caused by irregularities in gears and usually disappear as the gears wear.

These "ghost frequencies" are caused by irregularities machined into the gears in the manufacturing process. Ghost components are independent of loading, and tend to disappear as the gears wear. These spectral components are not related to the tooth counts of either gear. These components are the result of irregularities in the tooth spacing of one of the gears. The irregularities are the result of machining errors when the gear was made. So new gear sets will sometime exhibit Ghost components but they generally tend to disappear over time since there is no mechanical action that reinforces them
- Reference [] http://www.compsys.com/drknow/aplpapr.nsf/06b6f5a4de2ea...0436010?OpenDocument Defining Analysis Parameter, Alarm Limit & Fault Frequency Sets by Vern Sorenson discusses: - Gear Boxes and the gear mesh frequency or frequencies may require several parameters to identify problems within a multi stage gear box. Gearmesh is the number of teeth on a shaft times the shaft speed. If this is only a single reduction or increaser the gearmesh will be the same for both the pinion and bullgear. With the addition of one or two intermediate gears the number of gear mesh frequencies of concern will multiply by two or three or more depending on the configuration of the gear box
- A malady that is ever present on gearboxes is the ghost frequency. The ghost frequency usually shows up at approximately 50% of gearmesh and is caused from the imperfections of the hobbing tool that cut the gear at the time of manufacturing. I mention this because many analysts have beaten their heads trying to figure out what the unknown frequency is in the spectra that they have no data on and is in alarm level. Every gearbox should have a fault frequency set up for the ghost frequency, not to alert us of problems but to show us what that unknown frequency is

Hobbing is a gear-tooth-generating process consisting of rotating and advancing a fluted steel worm cutter past a revolving blank. In the actual process of cutting, the gear and hob rotate together. The speed ratio of the two depends on the number of teeth to be generated on the gear and on whether the hob is single - or multiple-threaded. The hob cutting speed is controlled by change gears that vary the speed of the hobbing machine’s main drive shaft

- Reference [] Gearbox Spectral Components and Monitoring Methods Timothy S Irwin, P.E.
http://www.vibration.org/Meeting/0903/Gearbox%20Spectra...resentation%20V2.ppt makes reference to ghost frequencies

Oke Wayne, thanks for the links, but from a practical side:

How to put others to do "ghost busting", since in a lot of places monitoring equipment is bought, and someone who looks intelligent is put to pull the chart. Any tips on how to develop a good working monitoring system? IMO sending one person to training and waiting fore a "train the trainer miracle" is not very cost effective, especially if you are far from the "civilized" world.

Wayne,
I do not believe that all gears "wear in" over time. I have worked on gearboxes with high service factor (2.0) design that showed the original tooth flank surface over a very long time. It depends on the gear material and processes to make it and on the machine load conditions.
I have measured gear vibrations under variable load conditions, and sometimes the vibration would change and other times it would not.
I consider gears to be the most complex machine component for vibration measurements and fault diagnosis. I caution about using "rules of thumb" or other genaric facts and assumptions.

Walt

Ghost frequencies (or undulation error) are not very well understood. I don't think there is consensus about "wearing in". Sometimes, apparently the gears do wear in, but sometimes the "peaks" and "valleys" introduced by the hobbing error get steeper/deeper. In our case, we had a substantial presence in the data, but no evidence of abnormal wear. Apparently a few microns across the face of the tooth can be enough to cause the problem.

Regards,
Tim

Having worked in acoustics and vibration for a major gear manufacturer for almost 20 years I can tell you positively that the irregularities that cause a ghost frequency don't always disappear over time. This may be true in relatively soft gearing but in large power transmission gearing that is often case hardened this may not occur. Thus you can't generalize by making the statement it will "usually disappear as the gears wear."

Another inaccurate statement is to state that "The ghost frequency usually shows up at approximately 50% of gearmesh..." This isn't true...it shows up as the product of the number of teeth on the cutting mechanism and the running speed of the rotor. It is quite often well away from the 50% of mesh value.

John

John,

You've told us what it is not likely to be..how about what it IS...What to look for?? In the field we are NOT likely to know anything about the cutting tool used to cut a gear.

If it is NOT gearmesh frequency and NOT an identifiable bearing defect frequency is it likely a "ghost frequency" and how would we know? If it IS a ghost frequency would it be a NON-gearmesh integer harmonic of turning speed, less than or greater than gearmesh?...
...OR???... ---------- ----- -----

As to going away with time...I agree, not always...I have seen numerous drives of various ages with an anomalous peak[s] that apparently has nothing to do with a known tooth count[s], shaft speeds and bearing ID's.

Dana, there are no hard and fast rules as to ghost frequency but rest assured that any reputable gear manufacturer can track back and determine the number of teeth used in a hob or in the index wheel. Admittedly, they are reluctant to provide this information, not because it is proprietary, but because they have to be convinced that you can substantiate that you have a real issue before they get involved with what is a difficult task. Much gearing is outsourced in this country unfortunately compounding the problem. Also, too many people out there suffer from analysis paralysis, wanting an explanation for every single spectral component observed.

Some techniques that can be utilized is to take a time synchronous average of the casing signal that present the "unknown" frequency using a Keyphasor for triggering from both the pinion and the gear. With sufficient averages, the component will average to a small amplitude on the rotor that does not have the default and will sustain itself on the rotor with the fault. The speed of the machine should be constant. You can also hone in to the exact frequency using high resolution zoom and divide by rotor speed looking for an exact integer; again, this should be done on a constant speed machine. This isn't the best way obviously and with some ratios just won't work. It can however be a useful technigue when you can't do a time-synchronous average. Then go back to the gear manufacturer armed with this information. Pose the question did you have approximately "x" number of teeth on the index wheel used in manufacturing this gear.

As to severity, once in a great while (military work) you will find an actual spec addressing limits to be applied to the actual geometry of the gear. Normally you will only have an AGMA or DIN quality spec to work to which doesn't really address the phenomenon. You have to judge by the final signature. In my opinion, I'm less concerned at the source of the component as to the amplitude. Thus, if you determine via some means that 3 g's peak is a suitable level, then regardless of the source it should not exceed 3 g's.

In investigating these phenomenon also keep in mind that specs applied to casing measurements have generally gotten tighter in recent years but this does not always mean a better gear box. In petro chemical gearing for instance, something built to API 613, mesh frequency might commonly be up at 8 to 10 kHz. Years ago allowable levels might have been 10 g's pk, today the allowable levels are much lower. Yet many old gear drives have been giving reliable service for years at 10 g's. Is todays box better, or is the manufacturer paying better attention to the physical mount of the accelerometer?

John

Thanks to all for this discussion.
Regards
Wayne