Analysis of Process Fan Failure & Bearing Housing Case_Study - [PDF Document] (2024)

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

1/16

1 of 16

CASE STUDY

ANALYSIS OF PROCESS FAN FAILURE

AND BEARING HOUSING/SHAFT DESIGN

Ken Singleton

Manager

KSC Consulting LLC, Bristol VA

[emailprotected]

Alan Barbee

Reliability Engineer

Fortron Industries

Wilmington NC

Bob Bracher

ATK Maintenance Engineering Assistant

Radford VA

Bob McGinnis, P.E.

(Consulting to KSC)

McGinnis Engineering LLC, Kingsport TN

Abstract:AnAir Movement and Control Association (AMCA)Arrangement 8 fan handling hot processgas experienced severalbearing failures and ultimately a catastrophic failure requiringshutdown of the

plants production. The investigation found that the 1st criticalof the fan rotor was too close to the

operating speed resulting in excessive bearing loads and shortbearing life. Rotor dynamic modeling was

used to evaluate a new shaft design. A new shaft and one-piecebearing housing were designed, fabricated

and installed, successfully moving the critical speed severalhundred RPM above running speed. The new

design also provided improved maintainability over commerciallyavailable designs, as well as lowering

vibration and significantly improving reliability. This articlediscusses the analysis process, rotor

modeling, bearing housing design process, vibration datacomparison before and after the new shaft and

bearing housing were installed.

Keywords:Bearing housing design; critical speed; failureanalysis; overhung fan; shaft design; rotordynamic analysis.

Background: The AMCA Arrangement 8 fan, shown in Figure

1, was designed for 450F gas temperature. Operating speedwas

1785 RPM, direct driven by a 200 HP motor using a Falk Grid

type coupling with 7 spacer. The bearings were spherical

roller 22218 CCK C3 fit lubricated with Mobilith AW-2grease.

The fan wheel was constructed of Avesta 2204, continuously

welded with a 304L SS hub. The wheel hub fit to the shaftwaswith a tapered interference fit, set screws and an end cap.A

Durametallic Triple Carbon Ring seal provided sealing at the

shaft entry to the fan housing. The fan base was fabricatedof

A36 HRS (hot rolled steel). The fan housing was constructedof

304 SS. Fan performance was 34,200 CFM, SP 22.6, and

operating inlet temperature of 365F to 392F.

Figure 1. AMCA Arrangement 8 Fan

Drawings.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

2/16

2 of 16

Figure 3. Image Taken

Shortly After Fan Failure.

Figure 6. Fan Shaft and Fan

Hub. Shaft Heavy Rub Against

Fan Wheel.

Figure 4. Fan Shaft and Wheel in

Shop After Failure.Figure 5. Mechanical Seal

Housing.

Figure 7. All Fan Wheel

Hubs Failed.

Fan Reliability: Initial involvement on the project for KSC wasproviding assistance improving reliability

of the fans. The plant had nine process fans and improving theirreliability was a priority. One

improvement was replacing the bearing support plates on some ofthe fans with 1-3/4 thick plates

machined flat to 0.002 inch/ft.

Bearing reliability and elevated vibration levels continued tobe the primary concerns for this particular fan.

Analysis of the vibration data identified the need for higherresolution spectra to see the rotor critical.

The fan rotor 1stcritical is indicated at 1965

CPM in the data shown in Figure 2. The

sidebands about 1X and 2X run speed are

the difference frequency between rotor 1X

and the rotor 1stcritical. Amplitude

modulation by the fan rotor 1stcritical

frequency is apparent in the data. The 1st

critical frequency would shift higher and

lower depending on the condition of the

bearings. New bearings, having tighter

clearance and higher direct stiffness, would

shift the critical speed higher. It was found

that by monitoring the frequency of the 1st

critical that as the bearings wore the criticalspeed would movecloser to the run speed

frequency. Vibration levels would begin to

rapidly increase as the critical speed

envelope enclosed the running speed

frequency range. Monitoring the frequency

of the 1stcritical provided another indicator

when to replace the bearings.

Fan Catastrophic Failure: On January 5th, 2006 the fan vibrationlevels were rapidly increasing.

Vibration data was measured several times that day with the lastdata measured about 5:00PM. That

evening the fan failed catastrophically. Fortunately no one wasnear the machine at that time.

One of the first images taken of

the fan shortly after the failure is

shown in Figure 3. Bothbearing housings (cast iron)

fractured. The fan shaft had a

significant permanent

deformation (bend) in the area of

the mechanical seal, see Figure

4. The shaft at the mechanical

seal area, see Figure 5, showed

heavy rub part way around the

circumference of the shaft.

Figure 2. Spectrum at Fan Inboard Brg Horizontal

Shows Fan Rotor 1stCritical at 1965 CPM, 173 CPM

Above Running Speed. Sidebands About 1X and 2X

Run Speed are Modulated by the Fan Rotor 1stCritical

Frequency.

Fan

1X

Fan Rotor

1stCritical.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

3/16

3 of 16

Heaviest rubbing had occurred near the bottom of the sealhousing. The shaft also exhibited heavy rubbing

by the fan wheel, see Figure 6,after all of the wheel hub boltsfailed, see Figure 7. The shaft rub locations

and the bend in the shaft suggested a significant thermal bow inthe shaft may have developed due to

rubbing at the seal.

Two vibration spectra and time waveforms measured shortly beforethe failure are shown in Figure 8 & 9.

The bearing housing vibration does not show a pure sinusoidalresponse but a 1X & 2X response. The

shaft was likely contacting the seal housing when these datawere taken. The bearing housing hold down

bolts may also have loosened.

Near Catastrophic Failure: The fan was returned to service afterreplacing

the bearings, shaft, fan wheel and mechanical seal. Reliabilityof the bearings

was a continuing problem. A near catastrophic failure occurredNovember 6,

2006. The fan was disassembled and physical inspection was madeof the

rotor and bearings. To include all stake holders in theinvestigation and

resolution of the fan reliability problems, the clientsreliability engineer had

assembled a fan reliability improvement team that included thereliability

engineer (team leader), mechanics, operators, maintenanceplanner and the

vibration analyst.

The wheel end bearing was an fa*g 22218 ESK C4 with a steeldouble cage.

The grease was very dark indicating contamination, as shown inFigure 10.

The bearing outer race and housing bore exhibited severefretting corrosion.

The bearing cage at the wheel end exhibited axial wear from therollers

rubbing. The cage bore exhibited rubbing on the inner race. Theinner race

had guide grooves machined for the seals.

The coupling end bearing was also an fa*g 22218

ESK C4 with a steel double cage. This bearing was

the fixed bearing taking the thrust load. The thrust

should have been carried by the higher radially

loaded wheel end bearing. The grease was very

dark indicating contamination. The bearing outerrace and housingbore exhibited severe fretting

corrosion primarily in the top half of the housing.

Static load on the bearing would have been

vertically up due to the overhung load of the

wheel. The bearing cage exhibited axial wear from

roller rubbing. Rubbing had also occurred in the

bore of the cage against the bearing inner race.

The inner race also had guide grooves machined

for the seals.

Figure 10. Fan End

Bearing Nov 6, 2006.

Figure 8. Fan Coupling End Bearing Housing,

Horizontal.

Figure 9. Fan Wheel End Bearing Housing,

Horizontal.

Figure 11. Fan Rotor Removed From Service.

Light Fretting

Corrosion At

Bearin Fit

Greater Wear at Seal

Area nearest Wheel.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

4/16

4 of 16

The fan shaft exhibited lite fretting corrosion at the heavierloaded wheel end bearing fit, as shown in

Figure 11. There was also lite wear at the mechanical seal areaand the seal area exhibited melted material

(product in the air stream). The reported melt temperature was230 Deg C ~ 446 Deg F. There were no

indications of wheel rub or shaft cracks.

The bearings were sent to SKF for analysis and their commentswere as follows:

Classic description of high unbalance load. The cage acts like ahula-hoop orbiting in the bearing about the inner race. The cagebore is very

difficult to lubricate and thus begins to wear.

Consider a bearing with a hardened steel cage as an option222218EK.

The fretting corrosion problem is caused by relative motion ofthe outer race and housing bore. If a

groove is worn only a few thousands deep in the housing, theouter race can hang and not float

axially. If this happens the bearing temperature will rise.

Consider a four bolt housing FSAF 22518.

Pay attention to color of the grease expelled from the seals.Dark color indicates contamination.

Vibration data was measured on the fan inboard bearing housingjust prior to the November 5, 2006 event,

see Figure 12, and Nov 11, 2006 after replacing the bearings,see Figure 13. The data shows the rotor 1st

critical shifting about 120 CPM higher due to new bearings withno wear having higher direct stiffness.

During this time period several changes were made includingreplacing the bearing hold down bolts with

grade 8 bolts and washers, different grease lubrication wastried and bearing clearance was increased

Fan Rotor-Bearing Dynamic Model: The rotor

was initially modeled in November 2006 using

finite element rotor modeling software. The OEM

shaft was 3.187 inch straight shafting. The 1st

critical calculated for the original shaft material of

304SS to just above the fan run speed range as

shown by the undamped critical speed map inFigure 14. Bearingstiffness was estimated to be

about 235,000 lbf/in.

During discussion about machining a replacement

shaft, there were questions about the shaft material

to use. The rotor model was run for shaft material

properties of 304, 316L, 17-4 H 1025 and 4140.

The wheel weight was determined by measurement

and the wheel inertial properties were provided by the fan OEM.The coupling weight, bearing locations

Figure 13. Fan Inboard Bearing Horizontal, High

Resolution Spectrum, Rotor & BearingsChanged After Nov 5,2006 Outage. Rotor 1

st

Critical Shifted to 1980 CPM.

1stCritical 1980 CPM.1stCritical 1860 CPM.

Figure 12. Fan Inboard Bearing Horizontal, High

Resolution Spectrum, Just Before ShutdownNovember 5, 2006.

Figure 14. Undamped Critical Speed Map for The

Original 3-3/16 Fan Rotor.

1st

2nd

3rdCritical

Run Speed

Estimated Brg

Stiffness (K) lbf/in

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

5/16

5 of 16

and shaft dimensions were obtained from an OEM drawing.Calculations were made to determine the

following:

Rotor 1stcritical speed.

Rotor static deflection.

Bearing reaction loads.

Rotor response to unbalance at the fan wheel center ofgravity.Undamped critical speed map of rotor versus bearingstiffness.

Maximum static shaft stress in bending.

Effect on rotor critical speed with 17-4 shaft diameterincreased from 3.187 to 3.25 inch.

Effect on rotor critical speed with 17-4 shaft diameterincreased from 3.187 to 4.00 inch.

Residual static and couple unbalance tolerances for BalanceQuality Grade G2.5 calculated per ISO

1940-1.

As expected, the different modulus of elasticity for each shaftmaterial had little effect on the calculated

shaft 1stcritical speed. Material 4140 had the highest modulusof elasticity and calculated to have a critical

speed about 50 RPM higher than the stainless steelmaterials.

Shaft diameter changes were evaluated for increasing the1stcritical speed. Shifting the wheel end bearing

closer to the fan wheel was also considered. Increasing theshaft diameter from 3.187 to 4.00 inch at alllocations except thecoupling hub and the tapered wheel hub fit and using 17-4 SS waspredicted to raise

the 1stcritical speed 30%. This appeared to be an effectiveoption to shift the critical speed and increase

shaft stiffness but would require lowering the bearing supportplate to accommodate larger bearings and

housings.

New Rotor and Bearing Housing Design Development: Afterevaluating the options to increase the

shaft size to move the 1stcritical speed higher and the bearingand support plate changes, a decision was

made to use a one piece bearing housing. Quotes were requestedfrom a well known bearing supplier for a

tunnel type bearing and shaft assembly. Delivery was severalmonths since these bearing housings are not

a stock item and are manufactured in Europe. Additionally, theshaft options offered by the bearing

supplier did not provide the larger diameter at the fan end thatwas needed to shift the critical speed higher.

In addition, the tunnel bearing housing centerline height wasdifferent than the SAF Pillow Block housings

which would require lowering the bearing support plate.

The clients reliability engineerand the fan reliabilityimprovement team evaluated all these options. After

following a very methodical process, a decision was made for KSCto design the bearing housing and shaft

assembly and to use a local machine shop for fabrication of theshaft and bearing housing assembly. The

client would review and approve the design and drawings.

The team wanted the bearing housing to be

designed for direct swap out of the existing pillow

block bearings, something the off the shelf bearing

housing design would not accommodate. This

meant that the shaft centerline height and bolt

pattern would match the existing SAF 218 bearing

housings and the fans position axially within the

fan housing had to be maintained. The design

would provide a backup plan if the new bearing

housing and shaft design did not work and it was

necessary to reinstall the original shaft and bearing

housings. The team wanted all of the components

to be assembled from one end, so that the fan

wheel did not have to be removed from the shaft Figure 15. Modelof Final Shaft Design.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

6/16

6 of 16

during bearing replacement. Removing the fan wheel from theshaft was a major, time consuming job.

These details were incorporated into the bearing housing andshaft design as well as other features to

optimize lubrication, vibration analysis, and temperaturemonitoring.

Over several weeks the design was developed and refined perinput from the clientsreliability team. The

rotor final design model with the 1stcritical mode shape isshown in Figure 15. The 1stcritical was

predicted at about 2475 RPM providing about 27.7% separationfrom run speed. Rotor weight and bearing

reaction loading calculated as follows:

Shaft wt: 115.35 lbf

Total Rotor wt: 694.35 lbf

Coupling end bearing reaction: -333 lbf

Fan end bearing reaction: 1,027 lbf

The damped unbalanced response for the wheel end bearing isshown in Figure 16and the 3D plot of the

rotor displacement in mils pk-pk at run speed is shown in Figure17.

Rotor Bending Stresses: The team leader asked for the maximumbending stresses in the fan rotor and the

location of those stresses. The stresses were calculated usingDyRoBeS [3,4] rotor modeling software for the

static bending stresses in the original rotor, see Figure 18,and the redesigned rotor Figure 19. However,

additional calculations must be made to evaluate if thesestresses are acceptable.

Figure 16. Undamped Critical Speed Map for

The Original 3-3/16 Fan Rotor.

Figure 17. Shaft Damped Unbalanced Response

to G2.5 Balance Quality Grade at Run Speed.

Figure 18. Static Bending Stresses, OEM Shaft

304SS, Maximum 2417 psi at Wheel End Bearing.

Figure 19. Static Bending Stresses, New Shaft

Design, 316l SS, Maximum 2016 psi Inboard

Wheel End Brg.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

7/16

7 of 16

According to [2], to allow for lack of complete designinformation, design texts, handbooks, and engineering

codes recommend use of safety factors and service factors. Thedesign of machine members is dominated

by the concept of allowable working stress. This stress levelwhich should not be exceeded is a fraction of

either the ultimate, yield or fatigue strength of the material.For static loading, yield or ultimate strength is

needed. Cyclic loading is called material fatigue. Thetransition in strength from static to cyclic loading

strength drops to 90% of the initial static value after about1000 cycles and to about 50% after 1 million

cycles then the curve is fairly flat thereafter. The deratingfactors [2]used to determine the allowable

working stress in bending for the redesigned shaft using 316 SSare as follows:

Se= Endurance limit/fatigue limit of machine component

S'e=Endurance limit/fatigue limit of a rotating beam specimen ofthe same material.

ka= Surface finish derating factor. Polished surfaces have lowerderating factor than rough

machine surface.

kb= Size finish derating factor.

kc= Reliability derating factor.

kd= Temperature derating factor.

ke = Stress concentration derating factor.

kf= Impact derating factor.

The calculated allowable working stress of 7,126 psi was greaterthan the values allowed by MIL STD 167[6]. The factor of safety isthe allowable working stress divided by the stresses in bendingwhich calculated

to 3.5. A factor of 2 it typically recommended.

Bearing Housing Design Process: The bearing housing and shaftassembly, shown in Figure 20, were the

end result of the design process. The bearing housing body wasfabricated from mechanical tubing. The

mounting feet were fabricated of mild steel and welded to themain body. The bottoms of the feet were

machined flat to within 0.001 and coplanar to within 0.0015.Dowelholes were provided in the feet foraccurate location of thebearing housing when mounted to the fan pedestal. The bearing boreswere held to

guidelines published by SKF for size, concentricity, run-out andperpendicularity. Two grease purge holes

were drilled and tapped -14 NPT, centered between bearings andpositioned 45 off axis below center.

Three -28 tapped holes with spot facing were provided (1) topvertically and (2) horizontally each side

for the mounting of accelerometers directly over each bearing.One -18 NPT tapped hole was provided

at each bearing for a temperature probe.

k S' 55,000 35,000k = 0.35S' 0.5 0.5 200,000a e

a

e utS

' 35,000psi FromChart Pg3.22 Ref 2a ek S

35,000' 58,333

0.6e

S

k =0.75 Where d>2.0inb

k =0.702Where Reliability = 99.99%c

620 620k = = 1.1

460 460 100d

T

1 1k = 0.54

1.86e

fk

S = 'e a b c d e f ek k k k k k S S =0.35 0.75 0.702 1.1 0.541.86 35,0000=7,126psie

Where NotchSensitivity

g = 60 From Chart 3-15,Pg60 Ref 2

1 ( 1) 1 0.86(1 1) 1.86

f

f t

k

k g k

7,126Factor of Safety = 3.5

2,016

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

8/16

8 of 16

Figure 22. Detail of Coupling End of Bearing

Housing Showing Seal and Seal Path.

A labyrinth ring was attached to the fan end of the

bearing housing by bolts, as shown in Figure 21.

The labyrinth provided a non-contact seal. The

design maintained 0.010 radial clearance between

the minor diameter of the ring and the outside

diameter of the shaft. The labyrinth ring bore was

profiled to form three recessed grooves. These

recessed grooves provided a non-contacting grease

seal between the outside contaminates and the

bearing.

The seal ring was made from 6061 aluminum to

prevent damage to the shaft if a rub occurred

against the shaft. A V-Ring Dirt Excluder was

added as a supplemental seal to repel dust. The V-

Ring fit snuggly over the shaft and sealed against

the outer face of the static labyrinth ring.

Two grooves were machined in the inner diameter

of the end cap to serve as a type of labyrinth seal, as shown inFigure 22. A

V-Ring Dirt Excluder (14) was also provided as a supplementalseal. Thegrooves in the end cap were sized so that the bore couldbe machined if

required, removing the grooves, to receive a 90 x 110 lip seal.This provided

for use of a lip seal if the labyrinth seal and grease provedinadequate. A seal

path or sleeve was installed with a shrink fit to the O.D. ofthe shaft to

provide a wear surface for the lip seal. A seal path provides ameans of

renewing the shaft seal contacting surface without having toreclaim the shaft

to remove grooving. Adequate hardening along with plungegrinding the

surface of the seal path offers a superior surface, extendingthe life of the

seal.

End Caps: The end caps were tapped for 1/8-27 npt for greaselines, see

Figure 23. Also provided were -28 holes on the face of the endcap for

the mounting of accelerometers to measure axial vibration.

Grease Dams: Grease dams were added inboard of each bearing toretain or

hold grease close to the rolling elements due to the

open space in the housing between each bearing,

Figures 24 & 25.

Sleeve: In order to facilitate removable of theshaft assemblyfrom one end as requested by the

fan reliability team, a spacer sleeve was used

between the bearings, as shown in Figure 26. A

close fit between the I.D. of the sleeve and the

O.D. of the shaft was used to minimize residual

unbalance affects.

The stack-up tolerance between bearings was

maintained by an SKF lock nut, KMFE style with

locking screw.

Figure 21. Seal at Fan

End of Brg Housing.

Figure 20. 3D Model of Bearing Housing and

Shaft.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

9/16

9 of 16

Drawings of the bearing housing and shaft assembly are shown inFigures 27.

Fan Wheel/Hub Mounting: The fan wheel hub

and shaft used a tapered fit. The amount ofadvance controlledthe amount of interference fit

and the compressive forces of the hub to shaft. Calculationswere made for the following:

Hub material 304 SS per the fan OEM drawings.

Two shaft materials were considered, 316 SS and 17-4 PHCondition 1025

The fan hub and shaft used a tapered fit of 0.750 in/ft per thefan OEM drawings, Figure 28. The shaft,

hub and wheel material properties are listed in Table I.

Figure 25.

Grease Dam at

Fan End of Brg

Housing.

Figure 24. Grease

Dam at Fan end of

Brg Housing.

Figure 26. Bearing Spacer Sleeve.

Figure 27. Bearing Housing, Shaft, Fan andBearing Support PlateDrawing.

Figure 28. Shaft Taper Fit Dimensions

Taper .750 in/ft

3.187

Figure 23. End Cap Grease Purge Holes.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

10/16

10 of 16

Modulus @

400 Deg F

2% Yield

Strength psi,

min

Coefficient

Thermal Expansion

Shaft Material 17-4 1025 27.0E6 psi 145,000 6.1E-6 /Deg F

Hub Material 304 SS 27.9E6 psi 30,000 9.0E-6 /Deg F

Shaft Material 316 SS 28.1E6 psi 30,000 9.0E-6 /Deg F

Wheel Material RA2205 26.6E6 psi 65,000 7.81E-6 /Deg F

The team asked about the advantages and disadvantages of astraight fit versus a tapered fit. A comparison

was provided as shown in Table II.

Advantages of a taper fit: The contact area between bore andshaft can be checked

The interference between hub and shaft is controllable bythe

amount of advance of the hub on its shaft

The removal of hubs is easier

Disadvantage of a taper fit:

It is more difficult to accurately machine the tapers of the hubbore

and the shaft.

It is possible to overstress the hub if it is advanced too faron the

shaft taper.

Dirt and surface imperfections restrain hub advance, and cangivethe false impression that the desired interference wasreached.

Advantages of Straight fit: A straight fit is more easilymachined and measured.

Disadvantages of a straight

fit:

Assembly without galling is difficult.

Disassembly without galling is even more difficult.

Maintaining zero clearance between shaft and hub isdifficult.

Locking the hub to the shaft requires a step in the shaft andcreates a

stress riser.

The specific fan hub advance required to provide

adequate interference fit at operating speed and

temperature calculated to 0.0306 inch, Figure 29.

Equations: Equations are from [1]

Table I: Material Properties Used For Calculations

12 d

A it

Where:

A= Hub advance on the taper

t= Taper (inches/foot)

d= Shaft nominal diameter (inch)

i = Interference rate (inches/inch)

Table II: Advantages/Disadvantages Straight & TaperedFits.[1]

Figure 29. Illustration of Split Collar

Installed To Act As Stop When

Installing Heated Hub.[1]

0.0306 inch for 316SS Shaft and 304

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

11/16

11 of 16

Interference rate is the difference between bore and shaftdiameter (diametrical clearance), divided by the

shaft diameter. The interference rate is a ratio of quantitieswith the same units and thus has no unit of

measurement.

The combined hoop and compressive stress in hubs caused byinterference is maximum at the bore:

After completing the fan hub and shaft fit calculations, thefindings were as follows:

1.

Changing the shaft material to 17-4 PH Condition 1025 offeredsignificant advantages in material

properties over 304 SS. But, the differential coefficient ofexpansion between the hub and shaft would

not permit achieving an interference fit at the specifiedrunning speed and operating temperature

without exceeding the yield strength of the hub. The maximuminterference rate at ambient

temperature to avoid exceeding the yield strength calculated to0.0006 inch/inch. But, this would

result in a loose fit of .0016 inch at operating speed andtemperature. In addition to the inability to

maintain low residual unbalance, the loose fit would causefretting corrosion of the fit.

2. For shaft material 316 SS and hub material 304 SS a maximuminterference rate of 0.0006 inch/inch at

ambient temperature calculated to not exceed the specifiedstress limits. At operating speed and

temperature an interference of 0.001 inch was calculated to bemaintained. For the existing wheel and

hub, 316 L SS shaft material appeared to be the optimumchoice.

3.

Even with a larger shaft diameter of 4.000 inch and lessflexure, there was still some concern about

uneven rubbing of the seal material against the shaft causingthermal bowing. Thermal bowing would

cause increased unbalance forces and vibration. Coating the sealcontact area with a ceramic material

would reduce the coefficient of friction significantly andreduce potential thermal bowing and shaft

wear. Ceramic coating the seal area was recommended.

Where:

i= Interference rate (inch/inch)

D= Hub bore (large end)

d= Shaft diameter (large end)

2 2 2 2

2 2 2

( ) ( )

2 ( )

i E d c D d P

d D c

Where:

P= Contact pressure

i= Interference rate (inch/inch)

E= Modulus Elasticity Tension

d= Shaft diameter (large end)D= Hub OD

c= Shaft hole diameter (if any)

D di

d

4

2

3

1 ( )psi

P k

d

D

Where:

= Combined hoop and compressive

stress in hub psi

P= Contact pressure psi

k= d (shaft diameter)/D (hub bore)

d= shaft diameter

D= hub bore

D = Hub bore

d = Shaft diameter (large end)

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

12/16

12 of 16

4. For assembly of the 304 SS hub and 316 SS shaft, use of asplit collar clamped to the shaft was

recommended to provide controlled 0.0306 inch advance of thehub, as shown in Figure 30. For

ambient air temperature of 70 Deg F, the hub required heating toabout 200 Deg F prior to installation.

The hub would be installed on the tapered fit against the splitcollar and the draw bolt hand tightened

to hold the hub in place while it cooled.

5.

Aminimum 70% contact is recommended per

AGMA Standard 9002-A86 for keyed shafts

with tapered fits. If the fit does not meet the

70% minimum, and the fit requires lapping,

the shaft and hub should not be lapped

together. A ring and plug type lap of soft

material such as brass should be used.

A question was raised by the fan reliability team

about welding the fan wheel to the hub instead of

using the bolts originally provided by the fan

OEM. This question was asked since all of the

bolts failed during the Nov 6, 2006 failure. See

Figure 30. A recommendation was made to

increase the hub-to-fan wheel bolts by 50% in

diameter. High strength, corrosion resistant bolts,

should be used and the welding process eliminated. Because ofthe difference in the thermal coefficient of

expansion between 304 SS and RA 2205, welding could cause weldcracking problems over time if the hub

and wheel are welded together.

Bearing Evaluation:Many AGMA Arrangement 8 fans are providedwith spherical roller bearings. Wehave found that the bearings areoften not properly selected for the speed, load and lubrication.This is

especially true for the coupling end bearing which often doesnot have the minimum loading specified by

the bearing manufacture. For this fan the spherical rollerbearing limiting speed and required minimum

loading (using 2% of the dynamic load rating for roller bearingsand 1 % for ball bearings as recommended

by SKF) calculated as shown in Table II.

.Bearing Location Fan OEM Design One Piece BearingHousingDesign

Bearing

Location

Fan OEM

Design

One Piece

Bearing

Housing

Design

Coupling End 22218 EK 6217 Deep Groove

Ball

Fan End 22218 EK 6320 Deep

Groove Ball

Limiting Speed 5300 RPM 5600 RPM Limiting

Speed

5300 RPM 4300 RPM

Static Radial

Load lbf

333 lbf 335.5 lbf Static Radial

Load lbf

1007 lbf 1031 lbf

Bearing Static Load

Rating CO

84,300 lbf 14,400 lbf Bearing

Static Load

Rating CO

84,300 lbf 31,500 lbf

Bearing MinimumRequired Load lbf

1461 lbf 196 lbf BearingMinimum

RequiredLoad lbf

1461 lbf 391 lbf

Did not meet

Minimum load

for Grease

Meets Minimum

Load

Did not meet

Minimum load

for Grease

Meets

Minimum

Load

Problems Encountered During The Project: All projects havehiccups as did this one. Obtaining

accurate information from the fan OEM, the rotor assembly, androtor balancing presented some

difficulties. During initial rotor balancing, there was damageto the shaft bearing fits. The images of the

Figure 30. Nov 6, 2006 Failure, All Fan Hub BoltsFailed.

Table II: Comparison of OEM Specified Spherical Roller Bearingand Ball Bearings Used inCustom Design.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

13/16

13 of 16

Figure 31. Journal Showing Balance Machine

Roller Damage to Bearing Fit (Coupling End).

Metal Shaving Embedded in Journal.

Figure 32. Journal Showing Balance Machine

Roller Damage to Bearing Fit, Wheel End. This

Roller Track Appeared Wider or the Shaft

Shifted Axially During Balancing Operation

Generating Two Tracks.

shaft bearing fits, Figures 31, 32show scoring damage to thejournals by the balancing machine rollers.

Scoring of the shaft journals is always a concern when usingrollers in a balance machine to support the

shaft. To minimize scoring some actions that can be taken are asfollows:

The balance machine pedestals and rollers should be alignedperpendicular to the shaft centerline.

Misalignment of the rollers to the shaft can reduce theroller/shaft contact area and increases the

stress level. Misalignment can also cause the rollers togenerate an axial thrust increasing scoring

damage.

Lubricating oil should be applied to the shaft and rollers atregular intervals to maintain an oil film.

Felt wipers in good condition should be installed. The feltshould contact the rollers to clean the

rollers and minimize metal particles being carried between therollers and the shaft.

The shop performing the rotor assembly and balancing

did not follow ISO 1940-1 process to determine the correctbalance tolerance. The balance machine

printout showed that the residual unbalance tolerance used wasfor a center hung rotor which was then

divided by 2 to obtain the tolerance at each bearing journal.The residual unbalance tolerance should have

been calculated per ISO 1940-1 to G2.5 for an overhungconfiguration, as shown in Figure 33. The

calculations are as follows:

b = 9.625

c = 30.682

d = 14.187

Note that the static residual unbalance for the overhung wheelis about 1/10 of

the tolerance for the wheel if is center hung. Comparing thecalculatedresidual unbalance tolerances to the shops balance reportshowed the

following:

Balanced to 25.893 oz-in Right Side

Balanced to 20.027 oz-in Left Side.

These values were far in excess of the calculated maximumresidual

unbalance values:

Static unbalance = 0.65 oz-in

Maximum couple residual unbalance = 2.1 oz-in.

Two balance facilities were used but the quality of work did notmeet expectations at either facility.

2 2

per

static

U dU

C

Permissibleresidualstaticunbalance

Permissibleresidual unbalance

Permissibleresidualcoupleunbalance

static

per

couple

U

U

U

3

2 4

per

couple

U dU

d

6716.015 2.5 5.61

1800perU oz in

5.61 9.6250.65

2 2 30.682staticU oz in

5.61 3 9.6252.1

2 4 14.187coupleU oz in

b

d

c

Figure 33. Schematic of

Fan Rotor For Balance

Calculations.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

14/16

14 of 16

Installation of New Bearing Housing and Shaft: The speciallydesigned bearing housing and rotor were

installed in Sept 2007. Images of the beefed up bearing supportplate, access panel in the fan housing

(added to permit in-place balancing and inspection of thewheel), and the bearing housing and shaft are

shown in Figures 34-37.

The rotor 1stcritical speed,

measured in the vibration data.

was very close to the frequency

predicted by the rotor model.

The spectra plots with log

magnitude scaling, see Figure

38, were measured on the fan

outboard bearing housing

horizontal direction and showsthe rotor 1stcritical speed of

2351 CPM shortly after startup.

Figure 34. Photos of Newly Installed One

Piece Bearing Housing.

Figure 35. Photo of Newly Installed One

Piece Bearing Housing With

Accelerometers and Temperature

Sensors Installed.

Figure 36. Close Up Photo of Newly InstalledOne Piece BearingHousing With

Accelerometers and Temperature Sensors

Installed.

Figure 37. Photo of Modified FanHousing to Install Access Panelfor In-

Place Balancing & Inspection.

Figure 38. Frequency Spectra at Fan OB

Brg Housing, Hor, Original Shaft &

Bearings Top Plot, New Design Brg

Housing and Rotor Lower Plot.

1X

1X

Rotor 1stCritical (as low as1860 CPM) With OriginalShaft &SAF 518 Housing &

22218 Spherical Roller

Rotor 1stCritical 2351 CPM

with Custom Design OnePiece Bearing Housing, Shaft& BallBearin s.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

15/16

15 of 16

Before & After Vibration: The overall and 1X vibrationin/sec pk are shown in Figures 39 & 40.

Conclusions:

1.

The specially designed one piece bearing housing, shaft andbearing assembly successfully raised

the fan rotor 1stcritical from 1860 RPM to about 2350 RPMproviding a separation margin of

around 24% from run speed. The design provided severaladvantages over an off-the-shelf one

piece bearing housing and shaft assembly. The major advantageswere (1) direct bolt in-place

with hold down bolt holes matching the base plate existingholes; (2) lower bending stresses

generated by the overhung load, (3) improved maintainability(designed to permit bearing

replacement without pulling the complete rotor or fan wheelremoval), and (4) ball bearings met

the minimum load guidelines specified by the bearing manufacturefor grease lubrication.

2. Bearing housing vibration levels were reduced afterinstallation of the one piece bearing housing

and shaft and continued to remain low until the plant ceasedoperation in 2008.

3.

Problems were encountered during the project. Some of theseproblems had to do with the quality

of work performed at the shops which included excessive runoutof the tapered shaft fit on the 1 st

shaft, unacceptable balance quality at two balancing facilities,and inaccuracy of some field

measurements.

4. Important steps that should be considered for similarprojects:

a. Detailed analysis of the fan should include but not limitedto vibration, bearing calculations,

bearing lubrication requirements, bearing housing mounting,bearing support flatness and

rigidity, foundation, accurate rotor/bearing model, fanperformance curves, review of

maintenance history and OEM critical speed calculations.

b. Develop an accurate layout drawing of the fan base, bearingsupport plate, hold down bolt

locations and bolt size, the fan housing, wheel dimensions,wheel axial position, the inlet cone

position and overlap.c. During fabrication and machining,witness dimensional inspections of the bearing housing

and shaft at the machine shop. Shaft runout measurements shouldbe done with the rotor

supported in V-blocks at the journals per API 687. Dont trust acoordinate machine

measurement machine to accurately measure shaft runout andfits.

d. If using a contract balance shop, conduct detailed inspectionof the balancing facility to

determine training level of personnel, condition of thebalancing machine, most recent

calibration date and capability to perform the requestedbalancing work.

Figure 39. Chart Fan Bearing Overall Vibration

Original Design and Modified Design.Figure 40. Chart Fan 1XVibration in/sec pk

Modified Bearing Housings & Shaft.

7/25/2019 Analysis of Process Fan Failure & Bearing HousingCase_Study

16/16

16 of 16

References:

1. Michael M. Calistrat,Flexible Couplings Their DesignSelection And Use, ISBN 0-9643099-3.

2. Hindhede, Uffe, Zimmerman, John R., Hopkins, R. Bruce,Erisman, Ralph J., Hull, Wendell C.,

Lang, John e., Machine Design Fundamentals, A PracticalApproach, ISBN 0-471-04136-X.

3. Chen, W.J., Ph.D., P.E., Gunter, C.J.,Ph.D.,www.DyRoBeS.com

4. Chen, W.J., Ph.D., P.E., Gunter, C.J., Ph.D., Introduction toDynamics of Rotor-Bearing

Systems, ISBN: 978-1-4120-5190-3.5.

Gunter, Edgar J., Understating Amplitude And Phase in RotatingMachinery, RODYN Vibration

Analysis, Inc., Charlottesville, Virginia, Vibration Institute33 Annual Meeting June 24-26, 2009.

6.

Szenasi, Fred R., P.E., Torsional Analysis of Variable FrequencyDrives, Engineering Dynamics,

Inc. San Antonio, Texas