Pump Magazine Publications


“The Essence of Equipment Failure Analysis -

Theory, Approach, and a Case Study”




Sourav Kumar Chatterjee

Manager, Rotating Equipment

Hindustan Petroleum Corporation, Ltd.

Mumbai Refinery, India

July 9, 2002




Failure is incapability of an item to deliver desired level of service as specified by design/expected by user, under specified condition. A thorough analysis of Root Cause of Failure is followed by the detailed Field Case History of a seal failure of a pump at a refinery. Human factor, logistics and team assignment is analyzed, along with tracking technical aspects of a problem. Actual data for a pump operation around the failure period is related to mean time between failures (MTBF) and a follow-up monitoring plan, after problem evaluation and correction, is established. An interesting and informative case for practicing plant engineers, maintenance and operating personnel, to compare notes and learn.




Analysis is a technique where a set of useful information on an event under consideration is compared with a set of design information of same areas and pertaining to same item involved in the event,  - to find out the deviations, followed by a logical conclusion on cause of eventuality, using expert system which possess wide database on similar type of events or using human expertise.


Failure analysis


Failure analysis is an analytical technique used by professionals of all field at various functions to protect against potential problems in process & products.


What is a Potential Failure?


The identifiable & measurable physical condition of an item, which may be equipment / person / system, and which indicate that the functional failure is about to occur or in the process of occurring is known as potential failure.

The term potential implies strong probability of occurrence.




* Temperature of running equipment parts (bearing housing casing lube oil, etc)

* Visible leaks and wear

* Vibration level indicating potential bearing failure

* Wear particles in gearbox oil showing imminent gear failure


What is a Failure Mode?


It is an event, likely the cause for the condition of each failure state. In other words, it is the manner in which an item could potentially fail to meet the functional requirement, or design intent, or both, as defined and/or acceptable to the end user.


Some typical failure modes:

* Bent                  * Incorrect adjustment

* Broken                * Internal leak

* Contaminated          * Jamming


What is a Failure Effect?


Failure Effect indicates the result of failure, and makes us realize the following:

* Evidence that the failure has occurred.

* Safety, environmental & social consequences

* The way in which the production or operation or system is getting affected

* The physical damage caused by the failure

* Action to be taken to repair/revive /cure the system and arrest further deterioration


Some typical failure effects:

* Leakage of pumpage                  * Low pressure

* Low flow                            * No production

* Erratic operation                   * No control

* High vibration                      * Poor performance

* Rough finish                    * Unstable operation

* Operating parameter fluctuation     * Intermittent operation

* Deterioration of product quality



Objectives of Failure Analysis


* To find out Root Cause of failure and remedial actions

* Recognize and evaluate the potential failure modes

* Higher organizational, environmental, social and human security and safety

* Identify actions, which could eliminate or reduce the chance of potential failure from occurring

* Cost control

* Higher productivity

* Documentation of the process for future reference and monitoring


Core View of Failure Event


In an apparent assessment, though failure event leads to losses, hazard, despair, discrimination and all similarly negative notions, - but, ironically, there are few positive features also are inherent in it if followed by Root Cause analysis.




















Pentagon: Modernization




A formalized approach is of utmost necessity to carry out effective and successful failure analysis. Such concept generally comprises of five main activities.


Data collection

Formulation of probable cause areas


Remedial measure

* Documentation and corrective actions


Data Collection


The success of a failure analysis greatly depends on data collection. Out of so much data, the technique of picking up relevant data accurately is a highly skilled job. Many times the analysis sets back as concerned personnel become at a loss to understand what data are required, and how to get it. For an equipment failure following steps may be followed:


1- Identify the equipment & component

2- Find out potential failure mode or failure effect

3- Find out designed parameters (constructional & operational)

4- Note observations on operating parameters (during failure) & constructional parameter on dismantling


Formulation of Probable Causes

Type of equipment and accessories

Constructional features

Service condition

Type of component failed

Nature of failure

Potential failure modes observed before failure

Last maintenance details and MTBF




Probable causes















Remedial Measures

Remedial measures are adopted based on area of root cause and feasibility study for implementation.

* Design problem

* Installation problem

* Assembly problem

* Mal-operation

* Raw material/spare part problem




Documentation and Corrective Actions


Documentation is an arrangement / system to keep useful information in meaningful manner which can be retrieved easily whenever required. It is also information for concerned in relation to pertinent item or event, which is basic requirement for further development, and progress.


Documentation of entire failure analysis event must be done in designated item field and in prescribed format highlighting details of event and total observations, analysis considerations, justification for selecting appropriate measures, implementation details, effect and observations after implementation, update of P&ID / Datasheet / Drawings, indicating cause, date and agency involved.




Case study




Plant, site, or an enterprise modernization means continuous alteration of policies & approaches with the goal of making positive response to the needs, in terms of quality of product, its cost effectiveness, time, availability, and safety.  In course of doing that, a careful study must be carried out to select appropriate measures and to identify its key aspects for successful operation.


Appropriate monitoring of performance of such new systems also has immense importance. Absence of mandatory accessories for operation and monitoring can lead to trouble and hazardous situation. This case study presents a situation where an ECS seal for emission control has failed creating hazard due to improper monitoring system and supporting accessories.


ECS SealFailure Of LPG Pump,Data Collection








1A- Equipment type: centrifugal pump, back pull out design

Tag no.-  14P19

Location  Cr. LPG

Service   C3+C4 (Propane+Butane+propylene)


1B- Mechanical parameters:       

Bearing type  NU310/7310*2

Seal type ECS seal Double tandem

Flushing Plan 02,62 water quenching                                             

Seal box venting to closed flare

Cooling Plan- Plan G

Lubrication type- Oil splash lubrication

Lube oil grade--   Turbinol-68

Suction and discharge nozzle size-6"*300 & 4"*300

MOC of Major parts- SS-410, SS-316, CS


1C- Operation parameters:

     Service fluid: cracked LPG                          

     Temp:     45 C                                               

     Flow: 115 M3/hr                                       

          Sp.gr:   0.49                                                                                

          Diff. head:   75M                                                            

     NPSHr     1.0M                                                       

     Suct.pr: 210 PSI                                             

     Disch. pr: 263.5 PSI                              

     RPM: 1450                                                                                     

     Min. flow:    28 M3/hr

     Vapor pressure at p.t.: 200 PSI      


1D- monitoring facility:

Online primary seal failure detection facility      

Alarm / trip connected seal failure alarm      

Failure detection probability:    Fair



Failure mode and effect: both primary & secondary  (ECS) seal leaked

Time of failure: 1st May 2002 @ 3 AM 

Detected by: Operation personnel 

Immediate action taken: Pump stopped and isolated immediately

Safety hazard: Yes

Environmental Hazard: Continuous leakage of LPG through seal.

Failure Reporting time: 1st May 2002 @ 10 AM   


Input process conditions:   

Suction condition trend:     O.K.

Temperature trend: constant 

Suction flow trend: N/A

Suction source level / pressure: suction drum pressure & level trend constant 


Output process conditions:

Discharge flow trend: though the reflux flow trend found constant,

heavy fluctuations observed in LPG run down flow and back pressure.

Discharge Temperature trend: N/A 

Discharge pressure trend: N/A    


Observations at site:

Cooling/flush Line and jacket condition: cooling water lines found through and clear. Scaling found inside stuffing box jacket.          

L.O Condition: Good No contamination observed.           

Coupling condition: Good and intact            

Foundation condition: OK              

Alignment readings on decouple: within limit             

Suction and discharge piping alignment: no piping stress          

Piping Foundation condition: in order

ECS Seal system: flare vent line found plugged           



Bearing condition: bearings found good and intact, no radial and axial play observed.

Bearing housing condition: OK             

Seal parts condition: heavy pitting on seal ring mating face. Seal ring packing ( "O" ring) totally burnt.  Heat marks on Insert mounting burnt and damaged. Rotary unit springs found broken in pieces.  Dust of carbon found around seal parts. Observations on secondary seal: wave springs broken, bellows found punctured. Rotary face and packing good and intact. Heat marks on shaft at sleeve sitting portion.           

Shaft condition/runout: OK, runout 0.001"           

Impeller / lock nut condition : lock nut intact, impeller found cracked at back shroud sleeve/ bush Clearances: wear marks on sleeve at steam purge bush position

Wearing conditions & clearances: rubbing marks on both wear rings. Clearances found: 0. 045" and 0.050" as compared with designed 0.026" & 0.030" front & back respectively (suction and discharge)          

Condition of other related parts: coupling teethes well. Throat bush clearance also found increased by 0.015             

MTBF: 12 months   

Last PM & observations: 11th April, 2001 BCW lines were clear, coupling condition was good, bearing good, foundation bolts OK, alignment was off realignment was carried out, coupling run out OK.

Last failure details and cause: pump was removed for seal leak on 14/03/2001. Subsequently single seal

was replaced with ECS seal .


Last overhauling details with activities:

Bearings were changed, ECS seal was installed. 

Parts used from OEM/local: OEM   

Vibration trend since last O/H:




Probabale cause Areas 



* Starvation (loss of flow?)

* Bend shaft

* Bearing failure

* Misalignment

* Loose rotor assembly

* Sealing system problem


Observations and conclusions were based on type of equipment and accessories, constructional features, service condition, type of component failed, nature of failure. Potential failure modes observed before failure are depicted in the chart:


Detail Analysis and Discussion


The heat mark on seal parts, sleeve and fatigue failure of wave spring bellow of 2nd seal and spring of primary seal, eventually reveal the parts were exposed to high temperature and high stress causing catastrophic failure. Moreover, the alarm on failure didn't activate which is major flaw in ECS seal system and calls for immediate rectification. It may be noted that this seal was installed during March 2001 and the vapor recovery line has been connected to flare system only on April 2002. During the operation of seal this was kept plugged, as LPG is prohibited item or releasing to atmosphere. It is evident from observations of failed parts that primary seal failed first which could not be noticed, as alarm system didn’t work. The seal kept on running on ECS seal and only on failure it both failures got exposed leading to hazardous situation.        


Failure of Primary Seal


The flushing plan 02.62 (water quench) for this service, always has a tendency towards getting inadequate seal flush. This is because the pump design, which has back wearing and throat bush restriction to stuffing box along with impeller balancing holes. Due to this design, the stuffing box pressure always equals to suction pressure, which is very close to vapor pressure at process temperature. Hence rise in temperature at seal box can create vaporization at seal box and faces leading to loss of seal face lubrication. More dead end vapor recovery system also didn't allow the vapors at primary seal face and got accumulated at ECS seal box, pressurizing ECS seal box and increasing face loading on ECS seal. After some time the heat generated due to seal friction would add more heat to entrapped vapor causing the rise of pressure due to constant volume. This enhanced pressure will act on secondary seal box at O.D and on inner diameter of primary seal  insert squeezing off any possible lubrication film, which was already constrained due to type of flushing plan.


Thus the compression units were subjected to abnormal stress due to increased pressure along with high heat due to lubrication less rubbing of seal faces. In this case the primary seal leak took place due to inductee seal flush (evident from heat mark and carbon dust) followed by reversed pressure, causing damage of o-rings and compression of unit springs. Pitting on the seal face appears to be due to blistering as a result of heat concentration The hairline crack on impeller surface across the radius is also due to corrosion fatigue as it was subjected to cyclic stress due to flow variation within corrosive environment as the H2S, which is present in LPG (15000 PPM).


Failure of ECS seal


This failure was the result of high load on wave spring due to vapor concentration at seal box and rapid wear due to high face loading and lack of lubrication. Actually this seal face has less contact area so that heat generation be less and designed for operating under minimum box pressure. Once first seal is failed, this seal provided service for short period, allowing planned (although urgent) shutdown for seal repair.


The wearing clearances increased due to temporary rotor bow at impeller end while operating under fluctuating load condition away from BEP. Scale formation in stuffing box jacket further caused poor cooling effect and heat dissipation.      


Calculation of heat generation at seal faces:

Pressure-velocity facto (PV)

Heat Generation at seal: Q=C1 x PV x f(Ao), B.T.H/Watt            


b= seal balancing ratio, 0.7     

K=Pr. gradient factor, 0.3 for light liquid         

Psp= spring pressure = 0.45 bar           

Vm=velocity at mean diameter 3.14x65x3000/1000 x 60 = 10.5 m/sec            

f= coefficient of friction, 0.07 for C/TC Combination             

Ao=Seal face area of seal ring = .001 sq. m         

C1= 1 for SI unit           

PV = {(12-0.45) x (0.7-.0.3) + 0.45} x 10.5 = 53.025 bar m/sec    


Q= 1x 53.025 x0.07 x0.001 =0.037 watt/sec=0,037/4.18=0.009 cal/sec *             

Or Q= 0.009 x60 x60=31 cal/hr             

* Cal= watt/J  (J=4.18 Jules/sec)

This undissipated heat will cause rise in temperature of LPG vapor at constant volume and the rise per hour could be calculated by using gas law: P1 x V1 / T1 = P2 x V2 / T2           



Root Cause of Failure


The improper flushing plan and lack of vapor escape feature is the Root Cause of failure of primary seal. The non-function of alarm system and absence of vapor recovery connection are the root causes for ECS seal failure. The lack of cooling due to jacket scale also a cause to accelerate the failure.     


Remedial Measure         




1. Seal flush system modification to API Plan 11 that will maintain higher stuffing box pressure and enough flush.

2. De-scaling of stuffing box jacket and thorough inspection during preventive maintenance to be carried out.

3. The diff. temperature of cooling water to be monitored for effective heat dissipation.

4. The vapor recovery line to be connected properly to flare header.

5. Rectification of alarm annunciation system for seal failure.


Timing Schedule and Team Assignment


Activity no 1,2,4  - by maintenance

Item 3  - by operation / PAD     

Item 1  - in consultation with seal manufacturer during next available opportunity

Item 5 by Instrument section

Remaining items to be implemented immediately




* Six months observation 

* Document and update of records and history log to be maintained after corrective measures are implemented and continued during the following satisfactory operation of period of one year.


Readers Feedback, Questions, Discussion and the Author’ Comments:


“…There are several items that are a bit unclear, and this maybe due to nomenclature usage.  The author assumes the reader knows what a double/tandem seal is…”


“…If double tandem, then why a disaster bushing (plan 62)?  (Here again the nomenclature leads us to some confusion that could have been enhanced by including a simple schematic of the seal…”


“…The cracking of the impeller suction end ring is not covered sufficiently to dismiss total doubts of a failure initiation there.  Did this cause the failure of the throat bushing, was there one? And consequently  - the loss of pressure in the seal cavity?”


“…A great understanding of a seal design is shown in the analysis by Mr. Sourav Kumar Chatterjee.  I am curious of what was the cross-sectional face width of the seal?  In refinery applications for LPG one usually uses 0.125 inch, as a rule of thumb, keeping the balance ratio intact in the seal as manufactured (in this case 70%).  Face cross-section greater than 0.125 in., in extreme cases with stuffing box pressures 150 to 200 lbs, can generate sufficient face-to-face loads to turn the seals on the sleeve/shaft and begin machining themselves loose.  This can take place in a shorter time frame than the dead ended box can affect the seal operations. Would appreciate if Mr. Chatterjee could go over some of these to clear up…”


Author’ Response:


I am thankful for valuable comments. The seal is Emission Containment Seal (ECS) which works on nearly zero (particularly if N2 buffer provision is made) emission. The secondary ECS seal runs dry due to minimized contact area and a special grade material (C/SC) for face combination.


To flare through NRV

To minimize seal box pressure (the vapor from primary seal to be vented to flare) the general arrangement is as follows:








The original seal for this pump was a single seal and having plan 02.62. The 62 is a quenching provision and for LPG service water quenching is given to avoid ice formation in seal area due throttling expansion at close clearance areas when pump remains in stand-by condition. Initially, conversion to ECS seal retained the same plan 02.62 due to following reasons:

*      The same plan had been working satisfactorily for single seal since commissioning in 1994

*       In corporation of plan 11 needs major modification in stuffing box, which is time consuming affair.


Unfortunately, the detail analysis of the effectiveness of 02.62 for ECS primary seal had not been done to pinpoint the differences between the earlier single seal and ECS seal.


The cracks on the impeller are due to corrosion fatigue. Similar cracks were noticed and on micro-structure examination it was determined they had developed because of degradation of bonding due to chemical corrosion and subjected to variable pressure, as there is frequent pressure variation of rundown header.  For other pump the impeller material was changed to CF8M from 410ss and for this pump also same modification is done.



Hairline cracks on  impeller face:




D1   D2   D3

                                                                                                              Seal Box area           Atmospheric side                                                                                                                                                                                                                                                                                                                                                                                                              






Contact Area = ת/4 x D12  - ת/4 x D22           

Exposed area = ת/4 x D12  - ת/4 x D32


Closing force = Exposed area x Liquid pressure at stuffing box

Opening Force= contact area x mean face pressure due to hydrodynamic force


Normally, the closing force is higher than opening force so as to keep face contact stable during adverse operating condition like cavitations, high vibration etc.


Balancing of Seal faces= Opening area/ exposed area. Hence for same seal size an increase in face width will cause increase in opening force-making seal unstable in adverse operating condition. If for same seal size the seal face width is too small also the unbalanced closing force will squeeze the film thus loss of lubrication will take place leading to face damage.


Seal face width is selected based on pressure-velocity (PV) factor, service liquid, and size of the seal.


The stick lip condition, when rotary face tends to transmit torque to stationary face, can take place, if there is too much pre-load spring compression applied, or in case there exists congealing substances (VTB, LSHS etc.) due to improper heating or absence of purging steam) between two faces.


Distortion of carbon face can take place due to heat or impurities causing higher frictional force or higher frictional bonding with mating face.


Increase in throat bush clearance is not very high. Moreover, as the impeller is having balancing holes, the in throat bush area will not affect the seal box pressure remarkably as it is already close to suction pressure. Only the circulation would increase and the liquid film support to the shaft would be absent, leading to lowering of critical speed nominally and the rubbing of wearing surfaces could also take place.


Hope this helps to answer questions mentioned by the readers in the Editorial feedback. I sincerely thank the Pump Magazine publisher and the readers for their comments.


Sourav Kumar Chatterjee

Manager Rotary Equipment

HPCL Mumbai India  


We received a comment on this article form our reader from Spain:


Dear Sir


I  enjoyed reading the two cases Mr. Chatterjee  presented on mechanical seal troubles and solutions for a pump for the  reduced crude at the bottom of a column at a Primary Distiller Unit. I work in a refinery on a problem of continuous failures of the mechanical seal. This mechanical seal began to fail shortly after the installation in 2003. I think there are several factors: one related to the cavitation due to high temperature of reduced crude (624 ºC measured by a temperature transmitter) and low suction pressure (8 psig measured by a pressure transmitter).


The mechanical seals for the petroleum industry are governed by API 682, and the supplier used Plan I with Plan 11 and Plan 53 with a barrier fluid cooled in a convective tank (not forced). We experienced the leakage when using Plan 54 (with a forced barrier fluid through a heat exchanger) after a short time.


The current mechanical seal have bellows in a dual tandem pressurized arrangement with Plan 54 with a synthetic oil (Royal Purple Barrier Fluid GT 910) and with a recently installed steam quench extracted from a saturated steam line and demoisturized through a mechanical separator with a steam trap at the bottom of the separator, although I noticed that the condensate is leaving the mechanical seal directly to the drain (sewage) without a steam trap. Is trap needed to avoid condensate on the mechanical seal chamber? I am concerned if the steam entering mechanical seal chamber is dry, as wet steam would cause problems of coking at seal bellows. How can I make sure steam to the seal chamber is dry?


Considering low NPSH due to vapor pressure of reduced crude at the operating temperature of 624 ºC, operators said that the cavitation of pump occurs mainly when starting the pump.  Do you know what is vapor pressure of a typical reduced crude at the temperature of  624 ºC ? I think it varies according to the crude composition, but an approximate value would help..


I also think that a probable problem could be the diameter of suction pump of 10 inches – and has not been increased when we installed a new pump from the old pump of lower capacity. Thus a new pump has more friction loss leading to less suction pressure and risk of cavitation. The old pump of less capacity in gpm did not have problems with mechanical seal failures. Also, operating at low flow, the valve at the outlet of pump is throttled to less than 50 % of nominal capacity.


Vibration of pump after major repair of mechanical seals and bearings went down from 6.4 to 2.5 mm/sec and acceleration went down from 12.4 to 0.74 G. Are these ok?


There are three pumps in parallel pumping reduce crude from bottom of column. Two of them operate together, and have no seal problems. The third pump operates alone at reduced capacity has mechanical seal failures and control valve pinched. Initially, the pump had the same problem experienced in  Case 1 you presented with the barrier fluid circulating only when the pump was in operation and a failure occurred when there was no barrier fluid at the fluid reservoir and coke was found under the bellows of the secondary seal, which is why the seals faces opened.  


Plan 54 users synthetic oil (Royal Purple Barrier Fluid GT 910) as a barrier fluid, which the supplier claims to be compatible with the reduced crude.


I would appreciate any comments about the probable causes of our continuous mechanical seal failures, and any suggestions you may have to solve the problem.


Best regards


Manuel Luque Casanave

August 14, 2006


Dear Mr. Casanave, - thank you for your comments, and it is good to see that our readers keep in touch with the publications long after they first appear at Pump Magazine. We will let Mr. Chaterjee know your input, as well as are asking our readers to provide any additional comments, thoughts, and ideas. In my view, suction pressure was an issue, as you noted, and maintaining proper level of buffer fluid was an issue. Perhaps an automated level detection in the buffer tank would keep it from running dry. Regarding vapor pressure of oil, it may vary from one site to another, depending greatly on specific composition. I would recommitment you involve your local laboratory to test the oil, and determine its characteristics, including vapor pressure.


Thanks again, for your valuable contribution and an interesting case.


Dr. Lev Nelik, P.E.


Pump Magazine