Repeated shim pack coupling failure due to suspected torsional vibrations

Dear All
We are facing problem of repeated shimpack coupling failure on recycle gas compressor train. 

Equipment History
The equipment was originally commissioned in 2009. In 2011 there was damage to LP compressor rotor (a piece at discharge of 1st stage 1st impeller broke) while in operation and rotor had to be replaced. In 2014 compressor was overhauled again on preventive basis and at that time subject coupling (LP to HP) was found damaged “two of the flexible-disc screws being found sheared off and the disc (shim pack) had a fracture”. In 2014 rotor was replaced with repaired rotor (same that got damaged in operation in 2011) due to fouling on 1st and 2nd stage impeller

May 2017 Failure
In April 2017 compressor train was shut down due to some other problem and On May 4th 2017, compressor was started. Following sequence of events observed during startup
Within 8 to 10 seconds of reaching the full speed on 13,300 rpm, the LP compressor tripped on High Vibration . 
Subsequently, the compressor train was started four times and operated each time for about 40 to 70 minutes, during 4 and 5 May 2017.
On 5 May the it was observed that the System1 was not receiving any vibration data for HP Compressor after the first start up and trip. It was also revealed that HPC was not developing pressure during the last four startups.
Upon investigation, it was observed that the HP Compressor was not rotating during the last four startups since key-phasor was not showing any speed indication and the bearings were not indicating any vibration amplitudes.
Further it was noticed that during the first start up, HPC compressor reached only about 12,600 rpm and then coasted down to zero speed, meanwhile, LPC reached the rated speed of 13,300 rpm and with 8 to 10 seconds tripped on high vibration.
Based on this coupling between LP and HP Compressor was inspected and found badly damaged. Coupling transmission unit was replaced and compressor was started. Alignment readings could not be taken

Oct 2017 Failure

In June 2017, nearly one month after coupling replacement vibration values increased on LP compressor. In October last week compressor train was shut down again due to some other problem and subject coupling was inspected. Coupling was again found damaged. 5 out of 6 bolts on LP compressor side shim pack found broken. Also LP compressor shims were found damaged.As found alignment was checked and found as follows
Vertical
Parallel: HP compressor 0.1mm down
Angular: 0.1mm/100mm close at bottom
Horizontal
Parallel: 0.2 mm
Angular: 0.0mm/100mm

Coupling Manufacturer Feed Back
The couplings damaged in May 2017 and Oct 2017 were sent to coupling manufacturer and as per their feed back torsional vibrations cause loss of pre tension in drive bolt which eventually results in coupling damage. In Oct 2017 damage the HP compressor side drive bolts and shim pack were in tact and as per finding from coupling manufacturer tightening torque on these bolts was found to be 4 NM against design/set value of 10 NM.

Vibration analysis report prepared by our Condition Monitoring section also indicates presence of low low amplitude spikes at gear box at frequency matching with 1st torsional critical

 

what can be possible source/cause of torsional vibrations ?
Please note that the stripper bolts no 4 in coupling drawing which attach transmission unit to hubs are also torqued to 10 NM. Why these bolts are not loosened due to torsional vibrations ?

Regards

FSX

Attachments

Original Post

If you suspect torsional resonance, then measure it! I use a Binsfeld strain gage telemetry system, but there are other products and methods available. I would also test for resonance of the coupling spool piece in the axial direction by impact response method. I found this problem on four boiler feed pumps at a large Co-Gen power plant. I have had excellent success in monitoring shaft coupling (both disk packs) condition using airborne ultrasound measurements. Early coupling damage can be detected while machine is operating, and it more effective than vibration measurements.

Walt

Walt

Thanks for the response. We can proceed for torsional vibration measurement but the confusion in my mind is that is it torsional vibration causing this ?

Out of 3 similar couplings on this train this is the only one failing. The coupling between Gearbox to LP compressor is very similar but obviously has more torque rating. This coupling never had any problem since commissioning. Also Gearbox which is other transmission equipment in the train never had any problem.

Also if the disc pack drive bolts are getting loose in operation due to torsional vibrations causing bolt and shims to fail, I am not understanding why the stripper bolts position no 4 (highlighted green)in attached drawing which have same material and tightening torque as drive bolts (highlighted yellow) are not affected by torsional vibrations ?

As per my understanding stripper bolts are also subjected to same torque and torsional vibrations as disc pack drive bolts. The difference is nature of joint i.e. clamping length , friction condition and misalignment forces acting on drive bolts. It may be the case that 10 NM torque is not sufficient for disc pack drive bolts.

For torsional vibrations even to be fair I don't have any experience. What can be the possible source ? Any symptoms fluctuations in motor current ?

Attachments

FSX posted:

Vibration analysis report prepared by our Condition Monitoring section also indicates presence of low low amplitude spikes at gear box at frequency matching with 1st torsional critical 

 

Can you clarify (with some numbers) what you mean by the above statement?  The first torsional critical is shown to be around 1000 cpm which seems reasonable in my experience but what gear box frequency are you making reference to that matches the 1st torsional?   

FSX posted:

Coupling manufacturer has highlighted that critical speed of HP rotor is close to 3rd Torsional critical. 

Likely just a coincidence.  In my experience I have never seen what is likely a high speed gearbox response to the 3-node torsional critical.  Putting it in simple terms, it just takes a tremendous amount of energy to make a system respond torsionally at 4720 cpm.   Another check, and as the end user, you should also request the torsional study to verify that indeed the mode shape would support high stresses at the coupling.

Your S1 arrangement shows an induction motor.  Is it a VFD?

By the way your S1 trend plots have a comment "spikes due to torsional vibration immediately after startup" but the area of interest is from about 05 May to about 06 July.  I don't think there is sufficient detail in the plots to draw any conclusion about torsionals.  In fact, the phase is virtually constant, also not really supporting torsional.

Back to the bolting, if you found 4 NM and the recommended torque is 10 NM has that been corrected?

 

John Thanks for useful feedback

Motor is not VFD.

Torsional study from the project phase is available and it had been shared with coupling manufacturer and they had  observation that some characteristic values of coupling mentioned in study differ from actual values. I don't have details right now bit will be able to share this by Sunday.

As for torsional oscillations are concerned, we have attributed it to torsion mainly because frequency is matching with 1st torsional critical. Phase may be constant because what is displayed is 1X phase and spikes are synchronous.

anyway as mentioned before I don't have much experience with torsional vibrations so I will appreciate if you help us further in confirming whether it is really torsional

Coming to bolting the drive bolts are factory torqued and painted after torquing. In fact normally coupling manufacturer does not share any information about torquing of these bolts 

The observation of 4 NM torque was on damaged coupling HP compressor side drive bolts when checked by coupling manufacturer. As per them bolts get loosened in operation due to torsional vibration.  Refer to attached snapshot.

As per my understanding stripper bolts highlighted green in attached drawing (these are torqued by us in field) are total 04 on each side and torqued to same value.  These stripper bolts in my understanding are bearing same torque and torsional vibrations (if any) as coupling drive bolts but they have never been damaged.

The drive bolts highlighted yellow in attached drawing having same material and torque as stripper bolts are 03 on each side i.e.03 with guard ring and 03 with spacer. Also these drive bolts have greater clamping length. This makes me think that tightening torque of 10 NM is insufficient for these drive bolts. 

Attachments

Photos (1)
Files (1)

"As per my understanding stripper bolts highlighted green in attached drawing (these are torqued by us in field) are total 04 on each side and torqued to same value.  These stripper bolts in my understanding are bearing same torque and torsional vibrations (if any) as coupling drive bolts but they have never been damaged."

The bolts marked in green are at a greater radius from shaft centerline than those in yellow, so they would be at lower torsional force.

Did you verify distance between shaft ends? Were all parts of the coupling parts (or whole unit) replaced, or just disk packs and bolts? How did you control residual unbalance to precision tolerance? Was shaft alignment measured by dial indicators or laser system? Perhaps need a review of assembly procedure by plant technicians or coupling OEM representative.

Walt

Walt

Distance between shaft ends and coupling prestretch were measured and adjusted  every time during assembly.  As per available record during maintenance before 2014 and May 2017 failures prestretch was adjusted to 1- 1.1 mm as required based on shaft thermal growth data.

However during coupling replacement in May 2017 prestretch was incorrectly set to 0 which may have caused rapid failure after within 5 months.

In 2014 coupling was replaced as complete unit. In May 2017 hubs remained same but transmission unit was changed. Transmission unit includes spacer, both disc packs guard rings and prestretch adjustment shims. Refer to drawing attached previously transmission unit has 458.8 mm long and spans b/w joints highlighted in green. It is marked as item 1 in the drawing and is individually balanced. We never do any work on disc pack.

Alignment was checked using laser.

Coming to the bolted joint again

What I understand from your statement is that for given torque transmitted by the coupling bolts highlighted green being at greater radius will be subjected to less force and same will be the case during torsional vibration/torque reversals i.e. less cyclic forces on bolts in green then in drive bolts

I agree but then why drive bolts highlighted in yellow which are subject to more force during normal torque transmission as well as during torsional vibration are of same size, less in number and tightened to same torque as bolts on outer radius.

I am not sure what is extent of torsional vibrations but my gut feeling is this coupling lacks sufficient margin to account for service conditions. I am afraid there may not enough evidence to suggest torsional vibration is causing this damage.

Expert opinion from you guys will be helpful

FSX,

There may be a correlation between shaft axial position (related to pre-stretch, and thermal growth. The machine ran 2-years without incident, but the replacement LP rotor and overhaul may have affected operational haft axial positions (DBSE), either by change in thrust bearing or change in shaft or casing temperatures. Shaft and casing temperatures can be measured with IR meter during operation. Shaft positions can be measured with optical rule and strobe light during operation. Disk pack axial load (friction) and breakage (cracks) can be measured with airborne ultrasound meter during operation. The pre-stretch may be incorrect for the repaired machine.

I think you are making to much emphasis on torsional vibrations by relying on calculations and without verification by torsional measurements. 

I would not rule out axial resonant motion of the coupling spacer as previously mentioned without impact vibration test. The spacer shaft is the mass and the two disk packs act as axial springs for a spring-mass system. I have seen excitation from 1xSS on 3600 rpm 2000 HP feed water pumps.

I agree that coupling may be marginal for shaft speed and load, so that can be evaluated. The coupling may have been marginally good for the new commissioned machine, but now is marginally bad after various machine rebuilds.

Walt

As per available record during maintenance before 2014 and May 2017 failures prestretch was adjusted to 1- 1.1 mm as required based on shaft thermal growth data.

However during coupling replacement in May 2017 prestretch was incorrectly set to 0 which may have caused rapid failure after within 5 months.

Walt

As for thrust bearing/rotor axial position adjustment is concerned, please note that rotor axial position is first set by adjusting thrust bearing shims and then prestretch is set by measuring distance between hub flanges with both rotors pushed to active side and subtracting from it transmission unit length to get 1.1 mm difference. Any adjustment is made by changing shims at location of bolts highlighted green. 

So thrust bearing/axial position changes are accounted for.

For casing and shaft temperature I will check the suction and discharge temperature trends. There is n visible shaft portion so strobe light can not be used for shaft measurement.

Regarding prestretch I have one observation. As mentioned in drawing coupling axial misalignment capability is 1.1-1.2 mm for continuous service and 1.7 mm for transient. In cold condition required prestretch is 1.1 mm.  This to me is unacceptably close because taking into account field measurement tolerances and rotor axial position changes it is very much possible that coupling is started with axial pull beyond its maximum running capability. Any opinion on this ?

As for torsional vibrations is concerned it is highlighted by us based on gear box spikes and by coupling manufacturer based on fretting around drive bolt holes and washers. Anyway we are still open on this point and that is the purpose of getting expert opinion on this forum

As per disc pack coupling failure mechanism, damage of shims through bolt holes is an indication of imporperly torqued drive bolts. Is it possible that grooving highlighted by manufacturer around drive bolt hole location is due to embedding while tightening a factor to be accounted for in calculating drive bolt stress.

Is it possible that 10 NM torque may not be sufficient for drive bolts. As mentioned before if compared with bolts highlighted green which are subject to less force and are more in number, the drive bolts highlighted yellow should have more tightening torque ?

John

It is a double helical gear box with thrust bearing only in Gear shaft. There is no thrust bearing in pinion. Information on Gear Teeth is attached.

Torsional Critical Analysis report is attached.

As per coupling manufacturer feed back coupling inertia and stiffness values highlighted yellow in study report are different from values give on coupling Dwg. The values are zoomed and attached.

Any advise on possible impact of change in coupling inertia and stiffness on train's torsional behaior will be a favor

Attachments

FSX posted:

John

It is a double helical gear box with thrust bearing only in Gear shaft. There is no thrust bearing in pinion. Information on Gear Teeth is attached.

Torsional Critical Analysis report is attached.

As per coupling manufacturer feed back coupling inertia and stiffness values highlighted yellow in study report are different from values give on coupling Dwg. The values are zoomed and attached.

Any advise on possible impact of change in coupling inertia and stiffness on train's torsional behaior will be a favor

In a double helical, the axial spacing of all the components can be critical.  This is especially true when you have multiple compressors.  The end result of pre-stretch (or even compression) is to have everything expand so that minimal end thrust is applied to the pinion.

On the differing values for coupling stiffness and inertia it is possible that one set of values takes into account the shaft penetration.  Common rules for modeling of couplings split the shaft into two sections, roughly 1/3 and 2/3 of length in the coupling.  Each section contributes differently to the inertia and stiffness.  It is a lengthy topic so I suggest you visit https://dyrobes.com/wp-content...-Analysis_linked.pdf and download the paper.  Shaft penetration in a coupling is well discussed on page 19 (see Figure 4).   

John

Thanks a lot for sharing the useful information regarding modeling of couplings in Torsional Analysis. 

As the coupling being damaged is between LP to HP Compressor I believe your point regarding possible end thrust on pinion is related to vibration spikes on Gear Box. For this I would like to inform you that prestretch of Gear Box to LP Compressor Coupling is adjusted as per thermal growth data provided. If actual thermal growth does not deviate from the one calculated/anticipated there will be no end thrust on pinion. Also the prestretch is adjusted with pinion in middle of float zone in accordance with Gear Box manufacturer instructions. I understand this is intricate activity if such gear boxes are coupled with disc pack couplings.

However please note that these low frequency spikes on gear box are intermittent, if these are due to residual thrust on the pinion in my understanding these will appear at low load as the Gear Mesh Force holding the pinion in its position will be less at low load ? Should we try to find a correlation between motor ampere and gear box spikes ? There is no axial probe on pinion to directly confirm this.

Any other symptoms to diagnose cause of these low frequency spikes  on Gear Box ?

One more point

Both LP & HP compressors have thrust bearings. So I believe any axial force generated from error in prestretch in LP to HP Coupling should not effect pinion. Its only Gear Box to LP coupling that can effect pinion ?

FSX posted:

John

Thanks a lot for sharing the useful information regarding modeling of couplings in Torsional Analysis. 

As the coupling being damaged is between LP to HP Compressor I believe your point regarding possible end thrust on pinion is related to vibration spikes on Gear Box. For this I would like to inform you that prestretch of Gear Box to LP Compressor Coupling is adjusted as per thermal growth data provided.

Carefully check the thermal growth numbers.  The gear manufacturer that I worked for provided data based on a nominal 100 deg F rise.  But if the actual temperature rise was something different, then the end user contractor installing the machine was expected to make appropriate corrections.

If actual thermal growth does not deviate from the one calculated/anticipated there will be no end thrust on pinion. Also the prestretch is adjusted with pinion in middle of float zone in accordance with Gear Box manufacturer instructions. I understand this is intricate activity if such gear boxes are coupled with disc pack couplings.

However please note that these low frequency spikes on gear box are intermittent, if these are due to residual thrust on the pinion in my understanding these will appear at low load as the Gear Mesh Force holding the pinion in its position will be less at low load ? Should we try to find a correlation between motor ampere and gear box spikes ? There is no axial probe on pinion to directly confirm this.

Any other symptoms to diagnose cause of these low frequency spikes  on Gear Box ?

I would look carefully for the presence of the spikes at all four pinion probes on the gearbox.  Perhaps even look at the presence of the spikes on the gear.  In my experience, spiking on probes is typically something from some external cause, and not torsional except in a recip driven machine.  Instrumentation issues are quite common.  When a gearbox is loaded there are huge forces on the rotors generated from the transmitted torque.  I have seen for example, gearboxes where the tangential force at rated load is 10X of the pinion weight.  You can calculate the force on your machine; take the transmitted torque and divide by the pitch radius for tangential force.  The separating force is about 1/3 the tangential with a typical 20 deg pressure angle.  No need for a drawing to get pitch radius; the pinion pitch diameter is twice the center distance divided by the quantity (1 + ratio).  

Do you have casing transducers?  Do you see the presence of spikes on the casing?  You likely should, especially under load.  Oil film thickness is low, typically 2 mils more or less so there is good transmissibility. 

 

John 

Thanks again for useful information and guidance. 

Your point about tangential force to pinion weight ratio is valid in this case. In this case tangential load is 2,175 Kgf whereas static weight load on pinion journals is 33.6 Kgf. Refer to highlighted text in attachments for pinion weight and tangential load info.

As mentioned in the paper shared by you if external axial force is less then mesh centering force pinion will remain centered. in this case mesh centering force will be nearly half of tangential force using 28.3 Deg Helix Angle in formula given in attached paper. This implies nearly 1100 Kgf or 11 KN. The Gear Box to LP compressor coupling can with stand maximum of 1.4 KN with 2.6 mm displacement for transient period. So I believe we can rule out this.

Coming back to the spikes i agree with your point that these may be external. I once had a tough time with spikes on 2 MW integrally geared steam turbine gear casing. I discussed this on this forum also. 

I guess spikes due to large tangential load on pinion in relation to its weight should be high frequency ?But the strange thing is these are matching 1st Torsional Critical. May be coincidence?

We do not have casing mounted transducer but we can check with hand held magnetic transducer with portable data collector. I will get data for all four probes on both pinion and Gear and share but it may take time as I have to obtain this from condition monitoring team. 

Attachments

Photos (3)

John

Attached are trends from pannel showing Pinion and Gear Vibrations, Compressor mass flow and motor current.

Green : Motor Current

Black: Compressor Mass Flow

Blue: LP Compressor Axial Displacement

Yellow & Pink: Pinion Coupling End X&Y probes

Red & White: Gear Coupling End X&Y Probes

It can been that vibration excursions occur without any change in mass flow, motor current and LP Compressor Axial Displacement

 

Attachments

FSX posted:

Please be note that X&Y proximity probes are installed only at coupling end on pinion and gear shafts

There are likely provisions for proximity probes at the "blind" or uncoupled end of the gear and pinion which are used for shop test.  The drilled and tapped hole is then filled with a pipe plug prior to shipment.

This is not going to be an easy task to chase down the problem.  But since you see transient response on all four transducers I would suggest you rule out an instrumentation problem.  There are several things I would suggest.  Look at a trend of the DC gap voltage for all four probes.  If the traces overlay each other put each trend on a single plot so the data is clear.  Reduce the time from a 24 hour trend to something short, but also meaningful as far catching the transient.  What is of interest is the duration of a spike.  For instance if we see that the gear moves, as indicated by a DC gap voltage change, over an interval of milliseconds that means one thing.  If it moves, stays in some position for 5 minutes, and moves back that can mean something else but on a 24-hour plot both types of event may look similar.  You might also look at the dynamic data over a shorter interval and as single plots.  If you have a sharp instrument person have them add the DC supply voltage to the proximitors to the DCS and plot that.  It should be a steady value at about -24 vDC..

In light of activity being seen on all four transducers, another thing I would be checking is cable routing.  Is it possible that all four cables are in a cable tray carrying other cabling, especially something of high voltage.  Check the grounding and power connections at the rack to make sure you don't have any loose connections.  I had mentioned that you might be able to install a "blind" end probe on the pinion or gear; do you have the equipment to do so and have that probe powered and read by an independent instrument?  Or power and read one of the current probes with an independent instrument?  I see back in October 2016 you had an ADRE system.  

I think it critical that you verify any anti-surge mechanisms that are in place are functioning properly.

Lastly, once you rule out possible instrument  problems I'd add that tracing this type problem to root cause often requires specialist people.  Just to give you an extreme case of something similar, I suggest you review the paper at http://www.energy-tech.com/col...6a-7370d299242a.html.  The machine is certainly different than yours but a torsional transient was being caused by transient set up by a electric arc furnace (EAF) that was part of the same plant!

John

Thanks for the detailed repsonse. We will proceed for instrument checks as advised by you.

I have attached shorter interval plots of Dynamic Data (Direct Signal) from pannel capturing an event from 24 hrs plot. Pannel sample rate is 1 sample every 5 seconds so this may not be useful in determining which probe starts responding first. However it can be seen that it takes between 5 to 10 second for values on pinion X probe to increase from 14-15 microns to 21- 22 microns. In some instants value stays for 5 seconds. (Refer to values at top abd bottom of each of 03 cursor lines at last three pages)

Colour Tags are same as previous

Green : Motor Current

Black: Compressor Mass Flow

Blue: LP Compressor Axial Displacement

Yellow & Pink: Pinion Coupling End X&Y probes

Red & White: Gear Coupling End X&Y Probes

I will share System 1 Data plots showing gap voltage trends along with other steadty state data plots (Direct & 1X Trends, Water Fall Plots, Direct & 1X Orbits) once I receive.

Currently  I do not have access to ADRE or raw data. I will try to get short interval data plots with appropriate sampling rate to have some meaning ful information.(Since March 2017 I have been working for another company. Previously I used to have access to raw data and I used to generate data plots using ADRE in my laptop.)

I understand that for measuring Torsional Vibrations we will have to acquire services of specialists.

But as mentioned earlier I fear that  there may not be any severe/unacceptable Torsional excitations. This train had been in operation since 2009 and nothing has been changed. (System 1 was connected in Dec 2016 so data plots prior to that are not available) There had never been any problem to Gear Box or other two couplings in the Train. That is why before going forward with Torsional vibration measurements I want to have some expert opinion.

Gear Box window Inspection was done in April 2017 on account of these suspected "Torsional" spikes. No abnormality was found. Pictures are attached

 

Attachments

Photos (5)
Files (1)

Add Reply

×
×
×
×