See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229052516

WEIBULL ANALYSIS OF TIME BETWEEN FAILURES OF PUMPS USED IN AN OIL

REFINERY

Article · January 2003

CITATIONS READS

4 5,333

4 authors, including:

Anwar Khalil Sheikh Muhammad Younas

King Fahd University of Petroleum and Minerals King Fahd University of Petroleum and Minerals
115 PUBLICATIONS   1,571 CITATIONS    26 PUBLICATIONS   255 CITATIONS   

SEE PROFILE SEE PROFILE

Dahham Al-Anazi
Saudi Aramco
5 PUBLICATIONS   40 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Fouling in Heat Ex changers and Maintenance Strategies View project

All content following this page was uploaded by Anwar Khalil Sheikh on 30 May 2014.

The user has requested enhancement of the downloaded file.

 Table Of Contents Search Author Index

The 6th Saudi Engineering Conference, KFUPM, Dhahran, December 2002 Vol. 4. 475

WEIBULL ANALYSIS OF TIME BETWEEN FAILURES

OF PUMPS USED IN AN OIL REFINERY

Anwar Khalil Sheikh1, Muhammad Younas1, Dahham Matar Al-Anazi2

1: King Fahd University of Petroleum and Minerals, Dhahran 31261

2: Saudi Aramco, Dhahran 31311
[email protected], [email protected]

ABSTRACT

This paper is based on the time to failure or time between failure data and associated maintenance
cost for a set of pumps used in an oil refinery collected over a period of five years. Out of a large
number of pumps in operation, a set of pumps are identified which have experienced 3 failures in five
years. The Pareto analysis performed on this set of pumps further narrowed down the group of most
critical (worst performing) pumps. This group of most critical pumps is the primary target for the
further investigation and Weibull reliability analysis. The analysis started by determining the failure
modes of each pump .The highest number of failures are attributed to the failure of pump seals
followed by downtime due to overhaul, malfunction of impeller, failures of bearings, impellers, shaft,
couplings, etc. Seals and bearings are essentially non-repairable items and are currently being
replaced upon failure. It is found that majority of pump failures is because of seals, overhaul,
impellers and bearings.

The Weibull analysis helps to determine the reliability function, R(t) = exp[-(t/η)β] and its parameters
(β,η). The shape parameter β is a function of σ/µ i.e., β = φ(σ/µ), where σ and µ are the mean and
standard deviation of life respectively. The other parameter η  is the characteristic life. By utilizing
Excel spreadsheet, each pump was analyzed and its related reliability parameters β, η MTBF,
standard deviation, mode Tm, and median time T0.5 were determined. The shape parameter β helps to
select appropriate maintenance and part replacement strategies.
The cost analysis is conducted for both repairable parts and non-repairable parts. The accumulative
repairing costs for all modes of failure for the groups of pumps versus the time of operation for the
five years of study are plotted. Also, the average accumulative repairing costs versus the operation
time of each pump is estimated and plotted. This type of cost analysis can help in deciding if we have
to continue to repair the equipment or have to purchase a new one to replace it because repair cost is
becoming much higher than the cost of new pump.

Keywords: Repairable system, reliability, Weibull analysis, reliability model, parameters, Pareto
analysis, non-repairable parts, cost analysis,
Vol. 4. 476 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

‫اﻟﻤﻠﺨﺺ‬

‫ﺘﻌﺘﻤﺩ ﻫﺫﻩ ﺍﻟﻭﺭﻗﺔ ﻋﻠﻰ ﻤﻌﻁﻴﺎﺕ ﺘﺘﻌﻠﻕ ﺒﻭﻗﺕ ﻤﺎﺒﻴﻥ ﺍﻷﻋﻁﺎل ﻭﺍﻟﺘﻜﺎﻟﻴﻑ ﻟﻤﺠﻤﻭﻋﺔ ﻤﻀﺨﺎﺕ ﻤﺴﺘﻌﻤﻠﺔ ﻓﻲ ﻤﺼﻔﺎﺓ ﻨﻔﻁ ﻭﻗﺩ‬
.‫ ﻭﻗﺩ ﺘﻡ ﺍﺨﺘﻴﺎﺭ ﺍﻟﻤﻀﺨﺎﺕ ﺍﻟﺘﻲ ﻟﻬﺎ ﺃﺩﺍﺀ ﻤﺘﺩﻨﻲ ﻟﺘﻜﻭﻥ ﻤﻭﻀﻭﻉ ﻫﺫﻩ ﺍﻟﺩﺭﺍﺴﺔ‬.‫ﺠﻤﻌﺕ ﻫﺫﻩ ﺍﻟﻤﻌﻁﻴﺎﺕ ﺨﻼل ﺨﻤﺴﺔ ﺃﻋﻭﺍﻡ‬

‫ﺘﺒﺩﺃ ﺍﻟﺩﺭﺍﺴﺔ ﺒﺘﺤﺩﻴﺩ ﺃﻨﻭﺍﻉ ﺍﻷﻋﻁﺎل ﻟﻜل ﻤﻀﺨﺔ ﻭﺍﺘﻀﺢ ﺃﻥ ﺃﻜﺜﺭ ﺍﻷﻋﻁﺎل ﺴﺒﺒﻬﺎ ﺤﻠﻘﺔ ﻤﻨﻊ ﺍﻟﺘﺴﺭﺏ ﻴﻠﻴﻬﺎ ﺃﻋﻁﺎل‬
‫ ﺤﻠﻘﺔ ﻤﻨﻊ ﺍﻟﺘﺴﺭﺏ ﻭﺍﻟﻤﺤﻤل ﻫﻲ ﻤﻥ ﺍﻟﻘﻁﻊ ﺍﻟﺘﻲ‬.‫ ﺇﻟﺦ‬... ‫ﺍﻟﺘﻭﻀﻴﺏ ﺜﻡ ﺃﻋﻁﺎل ﺍﻟﻌﻨﻔﺔ ﺃﻋﻁﺎل ﺍﻟﻤﺤﻤل ﻭﻋﻤﻭﺩ ﺍﻹﺩﺍﺭﺓ‬
‫ ﺘﺤﻠﻴل ﻭﺍﻴﺒل ﻴﺴﺎﻋﺩ ﻓﻲ ﺘﺤﺩﻴﺩ ﺩﺍﻟﺔ ﺍﻻﻋﺘﻤﺎﺩﻴﺔ ﻭﻤﻌﺎﻟﻤﻬﺎ ﻭﻤﻌﻠﻡ ﺍﻟﺸﻜل ﺒﺎﻟﺘﺤﺩﻴﺩ‬.‫ﻻﻴﻤﻜﻥ ﺃﺼﻼﺤﻬﺎ ﻭﻴﺘﻡ ﺍﺴﺘﺒﺩﺍﻟﻬﺎ ﻋﻨﺩ ﺍﻟﻌﻁل‬
‫ ﻭﻗﺩ ﺘﻡ ﺘﺤﻠﻴل ﺍﻟﺘﻜﻠﻔﺔ ﻟﻠﻘﻁﻊ ﺍﻟﻘﺎﺒﻠﺔ ﻟﻺﺼﻼﺡ‬.‫ﻴﺴﺎﻋﺩ ﻓﻲ ﺘﺤﺩﻴﺩ ﻨﻭﻉ ﺍﻟﺼﻴﺎﻨﺔ ﺍﻟﻤﻨﺎﺴﺏ ﻭﺍﺴﺘﺭﺍﺘﻴﺠﻴﺔ ﺘﻐﻴﻴﺭ ﻗﻁﻊ ﺍﻟﻐﻴﺎﺭ‬
‫ ﺘﻡ ﺭﺴﻡ ﺍﻟﺘﻜﻠﻔﺔ ﺍﻟﺘﺭﺍﻜﻤﻴﺔ ﻟﻜل ﺃﻨﻭﺍﻉ ﺍﻷﻋﻁﺎل ﻤﻊ ﺍﻟﺯﻤﻥ ﻭﻫﺫﺍ ﺍﻟﻨﻭﻉ ﻤﻥ ﺍﻟﺘﺤﻠﻴل ﻴﺴﺎﻋﺩ ﻓﻲ‬.‫ﻭﺍﻟﻘﻁﻊ ﺍﻟﻐﻴﺭ ﻗﺎﺒﻠﺔ ﻟﻺﺼﻼﺡ‬
.‫ﺍﺘﺨﺎﺫ ﺍﻟﻘﺭﺍﺭﺍﺕ ﺒﺨﺼﻭﺹ ﻤﻭﺍﺼﻠﺔ ﺍﻹﺼﻼﺡ ﺃﻭ ﺍﺴﺘﺒﺩﺍل ﻗﻁﻊ ﺍﻟﻐﻴﺎﺭ ﻋﻨﺩ ﺍﻟﺼﻴﺎﻨﺔ‬

1. INTRODUCTION

Failure history of a set of pumps used at a large Refinery plants obtained from their
Computerized Maintenance Management System [Samaha, 1997] has been utilized to perform
Weibull analysis. These pumps are essentially repairable systems. Every action after
observing a failure is of corrective nature (corrective maintenance situation). Some parts in
these systems need replacement rather than repair. The Weibull reliability analysis [Kolarik,
1995], [Peterlik, 1995], [Lewis, 1987], [Abernathy, 1996], [Kapur, and Lamberson, 1977] has
been used to provide an indication of the equipment failure modes and to assess the
equipment reliability. Such an analysis can help to make the right decision for the pump
replacements and a proper selection of pump maintenance strategy. Analysis focuses on the
most critical bad actors equipment with highest rate of failure and repairing cost.

2. PARETO ANALYSIS

In the Maintenance Management System, the time of failure, type of failure and repairing cost
of each pump is stored. The data period of our investigation is limited to five years, from
January 1995 to January 2000. There are about 300 pumps operating in the plant. Out of
these, 44 are those whose history was reviewed. These forty-four (44) pumps were classified
as the bad actor pumps because each one of them has experienced 3 failures in the five years
period
The Pareto analysis [Samaha, 1997] was used to judge the relative severity of bad actors in
the refinery plant and to identify the most critical bad actors pumps out of these 44 bad actors
pumps .The decision was based on a suitable criteria of certain percentage of repairing cost
(50 %) and certain percentage of number of failures (50%) contributed by this narrowed down
group of pumps. Thus , out of 44 pumps 17 pumps are chosen to be the most critical bad

Top
Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 477

actors and listed in Table 1, they will be referred to as “the most critical bad actors pumps.”
This group of most critical pumps is the primary target for investigation and reliability
analysis. The Weibull reliability analysis will be performed on these pumps by using their
time between failure data.

Table 1 Most Critical Pumps Among The Bad Actor Pumps.

No. of Failures Cost in

Pump # Pump Description Process Type
in 5 years US $

P41 High Lift Pump Seawater pump 7 488,269

P39 High Lift Pump Seawater pump 7 459,617
P40 High Lift Pump Seawater pump 8 319,968
P38 High Lift Pump Seawater pump 7 240,946
P3 Salt Water Pump Seawater pump 9 233,429
P42 Fire Water Pump Seawater pump 10 150,550
P7 Feed Water Pump Seawater pump 6 148,793
P22 Brine Recycle Pump Seawater pump 5 113,789
P29 Distillate Pump Distillate pump 6 111,722
P8 Feed Water Pump Distillate pump 4 146,529
P32 Sump Pump Distillate pump 7 105,965
P15 Super-Heater Pump Distillate pump 8 100,849
P19 Distillate Pump Distillate pump 6 92,735
P33 Sump Pump Distillate water pump 6 48,765
P24 Booster Pump Distillate water pump 7 73,060
P26 Distillate Pump Seawater pump 7 56,327
P18 Booster Pump Distillate water pump 7 84,481

3. DETERMINING FAILURE MODES

The analysis started by determining the failure modes and the time to failure of each pump.
Before we proceed further a brief description of mostly used pumps , their main parts and
associated failure classifications are given below.
The types of pumps that are most commonly used in a Refinery plant are centrifugal pumps.
These pumps use centrifugal action to convert mechanical energy into pressure in a flowing
liquid. The main components of the pump that will be studied in this paper are impellers,
shafts, seals and bearings.

Top
Vol. 4. 478 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

An important aspect of the impeller is the wear rings. If the impeller is too close to the
stationary element, the impeller or the casing will be worn out. The other part is the shaft. It
runs through the center of the pump and is connected to the impeller at the left end. Seal is a
very important part in the pump. Seals are required in the casing area where the liquid under
pressure enters the casing. The last main part of the pump is the bearing. The pump housing
contains two sets of bearings that support the weight of the shaft. The failures causing the
stoppage of the pumps are primarily experienced by these parts and will be termed as failure
modes. There are 12 different failure modes for the most critical bad actors pumps. The
following is the definition adopted to characterize the various modes of failure:

♦ Mechanical Seal The pump failed due to a malfunction of the pump’s

mechanical seal.
♦ Overhaul** The pump failed due to unknown reason. Therefore, it will be
sent for overhaul which include inspection and repairing of
different parts.
♦ Impeller The pump failed because of its impeller failed due to either
corrosion, erosion or cracks.
♦ Bearing The pump failed because of bearing failure.
♦ Shaft The pump failed to operate because of shaft problem, such as
misalignment, vibration, etc.
♦ Suction Valve A failure due to some thing wrong with the pump suction, such
as problems in valve, corroded pipes or slug accumulated in the
suction.
♦ Casing A failure due to defective casing, such as misalignment or
corrosion.
♦ Operation Upset Failure of a pump due to operational mistakes, such as closing
a valve which should not be closed.
♦ Coupling A failure due to coupling distortion or misalignment.
♦ Gaskets A failure due to a gasket rupture or damage caused by leaks.
♦ Control Valve A failure due to malfunction of the control valve due to
pressure or flow in the line of service.
** Overhaul is necessitated due to any unidentified mode of failure, and its impact is the
nonavailability of the pump, therefore as a whole it is being grouped as a failure mode.

Figure 1and Figure 2 show the Pie diagrams of two of these 17 most critical bad actor pumps
for different modes of failures. To develop a comprehensive picture; all modes of failure for
all of these 17 most critical bad actor pumps were determined and sorted by the number of
failures associated with their failure modes; the results are presented in Figure 3 .The highest

Top
Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 479

number of failures are attributed to the failure of pump seals followed by downtime due to
overhaul, malfunction of impeller, and their failures. Seals and bearings are essentially non-
repairable items and will also be discussed separately. The figure also shows the other types
of failures, such as failures of bearings, impellers, shaft, couplings, etc. It is found that more
than 80% of pump failure is because of seals, overhaul, impellers and bearings. The
mechanical seal of pumps have the highest failure mode, which is 36% of the total number of
failures (111). The other highest type of failure is the malfunction (23%) due to unknown
reasons, so it needs overhaul, this means that the pump should be sent to the Machine Shop
for inspection and replacing the defective parts. The other highest of failure modes are due to
the impellers (11.7%) and due to bearings (9%).

4. TWO PARAMETER WEIBULL MODEL AND ITS CHARACTERISTIC

The two parameter Weibull Cumulative Distribution Function (CDF), F(t) defines the fraction
failing or probability of failing before time t (or unreliability at time t ) and has an explicit
equation [Lewis, 1987], [Abernathy, 1996], [Kapur, and Lamberson, 1977]:

β
F (t ) = 1 − e − ( t /η ) (1)

and the reliability function or probability of survival at time t is given by

β
R (t ) = e − ( t /η ) (2)

The shape parameter (β) is a non-dimensional parameter and reflect the type of failure mode,
such as infant mortality (β<1), random (β=1), or wear-out (β>1). The other Weibull
parameter (η) is a scale parameter having the same unit as of t, and is a function of the mean
time to failure (MTTF). For a special case whenβ=1, MTTF = η. The general relationship
between η and MTTF is given by the following equation: [Lewis, 1987]

 1
MTTF = E (T ) = µ = η Γ 1 +  (3)
 β
where Γ(•) is the gamma function.
Another important characteristic of reliability model is its failure rate λ(t), which is defined
as
β −1
 1 dR (t )   β   t 
λ (t ) = −   =    (4)
 R (t ) dt   η   η 

Top
Vol. 4. 480 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

The following are some other statistical characteristics that should be calculated during a
typical Weibull analysis procedure being applied to analyze the machinery time to failure
data:

1. Variance

  2  1 
σ 2 = η 2 Γ 1 +  − Γ2 1 +  (5)
  β  β 

2. Mode
1
 β − 1 β
Tm = η   (6)
 β 
3. Co-efficient of Variation

 2  1
Γ 1 +  − Γ2 1 + 
σ  β   β 
K= = (7)
µ  1
Γ 1 + 
 β

4. Quantile tp is given by
1/ β
 1 
t p = η  ln  (8)
 1− p 
where F(tp)=p.

5. Median T0.5

T0.5 = η (ln 2)1/ β . (9)

The Weibull analysis helps to determine the reliability function, R(t)=exp[-(t/η)β)and its
parameters β , η, MTBF, etc.). From Eq. 7 β = φ(σ/µ), where µ and σ2 are the mean and
variance of life, respectively. The other parameter η , is the characteristic life. By plotting
the time between failure for these pumps, it was observed that there is no underlying trend in
the data so the Weibull reliability analysis can be used. By utilizing Excel spreadsheet,
regression analysis of transformed time to failure data [Lewis, 1987] of each pump was
analyzed and its related reliability parameters were determined. In this analysis each pump is
considered to be a repairable system including all types of failures/ malfunctions. Figure 4

Top
Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 481

and Figure 5 show the linearized Weibull plots for the time between failure data of two of the
most critical bad actor pumps. Table 2 shows the reliability parameters for all of the 17 most
critical bad actor pumps.

Table 2. Reliability Parameters For The Most Critical Bad Actor Pumps.

Standard
Coefficient
Equipment # β η MTBF Deviation Mode,Tm T0.5
of Variation
(months) (months) (months) (months) (months)

P41 0.84 10.17 11.10 13.30 1.19 6.58

P39 1.02 9.03 9.00 8.80 0.98 0.21 6.31

P40 1.71 5.03 4.50 2.70 0.60 3.01 4.06

P38 0.97 7.72 7.80 8.10 1.03 5.29

P3 1.15 7.09 6.70 5.90 0.87 1.21 5.16

P42 1.26 5.96 5.50 4.40 0.80 1.73 4.46

P7 1.48 9.56 8.60 5.90 0.69 4.49 7.47

P22 1.11 6.98 6.70 6.00 0.90 0.89 5.02

P29 0.97 11.43 11.60 11.90 1.03 7.84

P33 0.64 9.28 12.90 20.80 1.62 5.25

P32 1.97 8.47 7.50 4.00 0.53 5.91 7.03

P15 2.09 6.60 5.80 2.90 0.50 4.84 5.54

P19 1.26 8.70 8.10 6.50 0.80 2.49 6.50

P18 0.99 6.57 6.60 6.70 1.01 4.54

P24 1.19 10.28 9.70 8.20 0.84 2.21 7.55

P26 0.93 11.17 11.50 12.30 1.07 7.54

P8 0.86 17.67 19.10 22.30 1.17 11.54

Top
Vol. 4. 482 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

The reliability parameters of the most critical bad actors are very important to study its
reliability and maintenance. These parameters β, MTBF, standard deviation, mode Tm, and
median time T0.5. The shape parameter  is important to determine the type of failure of the
equipment and type of maintenance that should be applied. Ten (10) pumps are found to have
β≥1. The mean time between failures (MTBF) is also calculated for each pump. From the
MTBF, we can predict the failure time of the pump to plan ahead of time the right action to
minimize the plant downtime. The lowest MTBF of the pump is for P40, P42 and P15
(4.5 months, 5.5 months and 5.8 months, respectively), the highest MTBF is for P12, P29,
P26 and P41 (12.9, 11.6, 11.5 and 11.1 months, respectively).

6. WEIBULL RELIABILITY ANALYSIS FOR NON-REPAIRABLE PARTS OF THE

MOST CRITICAL BAD ACTOR PUMPS

The most common non-repairable parts of the most critical bad actor pumps are found to be
seals and bearings. The common practice of maintenance of these parts is replacing them
with new ones when they fail. In this analysis, as shown in Table 2, the most critical bad
actor pumps are segregated based on their process (for example, seawater pumps, distillate
pumps). As shown in Figures 6 and 7, the time to failure data of the seals and bearings of all
the distillate pumps, and seawater pumps is pooled together. The Weibull analysis on seals
and bearings are performed to find the shape parameter and the characteristic life. It is found
that the behavior of the failure of seals for both seawater pumps and distillate pumps are
same. So, the time to failure of the pumps seals are combined together and plotted in one
figure (Figure 6). From this figure, we calculated the characteristic life of pump seals and
found it to be 27.5 months, which is around two-and-a-half year. The other types of non-
repairable parts of the pump are bearings. As shown in Figure 7, the bearings last little bit
longer than seals in service. The characteristic life of the pump bearings are 29.4 months.
The shape parameter for both pump seals and pump bearings are greater than one (=1.76 for
seals and =1.74 for bearings).

7. COST ANALYSIS OF REPAIRABLE SYSTEM (PUMPS)

The cumulative repairing cost for all modes of failure for each pump versus the five years of
operation was plotted. This type of cost analysis can help in deciding if we have to continue
to repair the equipment or we have to purchase new one because of the repairing cost is
becoming much higher than the cost of new pump. This analysis is shown in Figures 8 and 9.
The cost analysis is conducted for both repairable parts (all modes of failures combined
together) and non-repairable parts. The accumulative repairing costs for all modes of failure
for the 17 pumps versus the time of operation for the five years of study are plotted. Also, the
average accumulative repairing costs versus the operation time of a typical pump is estimated
and plotted. However, for non-repairable parts, the repairing cost versus the time between
failures is plotted for both seals and bearings.

Top
Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 483

8. COST ANALYSIS OF NON-REPAIRABLE PARTS

The repairing (replacing) costs for seals of all of the most critical bad actor pumps are plotted
versus the operation time as shown in Figure 10. For the seals of all pumps (non-repairable
parts), it is found that the maintenance cost is increasing significantly after 10 months of
operation. The accumulative cost of the non-repairable parts is including also the cost of
repairing auxiliaries that is related to the pump seal and the seal will not work if this is failed.
So, after the expected life of the seal (27.5 months), the old seals should be replaced with new
ones and the related auxiliaries (tubes, fittings, drains, etc.,) should be inspected and replaced
if it is necessary. Similarly the repairing ( replacing) cost for bearings of all of the most
critical bad actors pumps versus operation time are plotted in Figure 11.

9. DISCUSSION

1. The time to failure or malfunction data of the most critical bad actor pumps were gathered
from computer maintenance system and from on-line monitoring of process variable data
retrieved from distributed control system of the refinery. The period of the investigation
is five years from 1995 to 2000. All of these pumps were analyzed in depth as repairable
systems. Some of their non-repairable parts, such as seals and bearings, were analyzed
separately.
2. Pumps are found to have a large number of bad actors. Our definition of the bad actors
equipment is “the equipment that have more than three failures or malfunctions in five
years.”
3. The Pareto analysis was utilized effectively to identify the most critical bad actors rotating
equipment. The Pareto analysis criteria are depending on two main factors, which are the
number of failures and their associated repairing costs. The criteria used in Pareto
analysis of the pumps is to identify those pumps that contribute to 50% of total number of
failures and /or 50% of the total repairing cost. Accordingly, 17 pumps, were determined
to be the most critical bad actors equipment by utilizing Pareto analysis.
4. The failure modes of pumps are figured out and plotted in Pie charts to visualize the
relative contribution of each failure mode. It is found that the mechanical seals of pumps
have the highest failure mode which is 36% of the total number of failure (111). The
other highest type of failure is malfunctions leading to overhaul (23%). The overhaul
means that the equipment is sent to the Machine Shop with unknown reasons to inspect it
and replace the defective parts. The failure due to impeller (11.7%), bearings (9%) and
shaft (6.3%) are other significant modes of failures.
5. The various reliability parameters and indices of these most critical pumps are β and η.
The Weibull reliability analysis is found to be very beneficial to characterize the
equipment time between failures of the pumps, and to select appropriate maintenance
strategy for a given pump. These parameters are β, η , MTBF, standard deviation, mode

Top
Vol. 4. 484 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

Tm, and median time T0.5. The shape parameter β is important to characterize the failure
rate of the equipment and type of maintenance that should be applied. Ten (10) pumps are
found to have increasing failure rate (β≥1). The mean time between failure (MTBF) is
also calculated for each pump. From the MTBF, we can predict the failure time of the
pump to plan ahead of time the right action to minimize the downtime of the plant. By
using the fitted Weibull model as a predictive tool, the operation management can take the
appropriate decision in advance to avoid any operational upset.
6. In reliability literature there is a wealth of knowledge on preventive maintenance policies
for repairable systems and part replacement strategies [Kolarik, 1995], [Lewis, 1987],
[Sheikh, et al, 2000], [Sheikh, 1990], [Sheikh, 1991], [Blischke, and Murthy, 2000].
A major application of the reliability analysis results is to decide about the appropriate
maintenance strategy. The detail maintenance and availability related analysis of these
pumps is presented in another follow up paper [Sheikh, et al, 2002]. The Weibull shape
parameter β of the most critical bad actors is very important to recommend the appropriate
maintenance strategy. It is generally recommended [Blischke and Murthy, 2000] that; for
β<1 (decreasing failure rate) predictive maintenance, β =1 (constant failure rate)
corrective maintenance, and for β >1 (increasing failure rate) preventive maintenance
could be adopted. Similarly for aging non repairable parts such as seals for which failure
rate is reflected by β >1, an optimal replacement strategy could be either preventive
planned replacements or scheduled replacements at fixed intervals rather than the current
practice of replacing them upon failure. However with advances in condition monitoring
tools and sensors it is often preferable whenever such monitoring is feasible and
economically viable to adopt for the predictive maintenance.
7. The cost analysis is conducted for both repairable parts and non-repairable parts for the
most critical bad actors rotating equipment. A best straight line is fitted to trace the
behavior of spending the maintenance repairing cost against the operation time for
repairable and non-repairable systems. The accumulative repairing costs for all modes of
failure for the 17 most critical bad actors pumps versus the time of operation for five years
of operation are plotted in Figure 9. Also, the average accumulative repairing costs versus
the operation time of a typical bad actors pump is estimated and plotted in Figure
10.However, for non-repairable parts, the repairing costs versus the time between failure
is plotted for both seals Figure 11.The accumulative cost of the non-repairable parts
includes the cost of repairing auxiliaries that are related to the pump seal.

10. CONCLUSIONS AND RECOMMENDATIONS

The Weibull reliability analysis is found to be very beneficial to characterize the equipment
time to failure data and to design appropriate maintenance strategies using Weibull model as a
predictive model [Black, and Geitam, 1983], [Kelly, 1997], [Ben-Daya and Dufuaa, 2000],
[Jeong and El-Sayed, 2000]. Based on this analysis, operation management will be able to
take the right decision in advance to avoid any operational upset and plants downtime.

Top
Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 485

1. The Weibull model can be applied to rotating equipment such as pumps to represent the
three types of failures, which are early failure or wear-in, random failure, and wear-out
failure, by determining the shape parameter β for repairable and non-repairable parts of
each rotating equipment.
2. Pareto analysis is found to be very helpful tool in reliability analysis which is used to
assist the management quickly identify the most critical bad actors rotating equipment,
such analysis is based on a selected criteria depending on ranking the number of failures
and their associated repairing cost for each equipment.
3. Pump failures due to seals are the highest failure modes, which is 36% of the total number
of failures. Accordingly, it is recommended for providing more efforts to investigate the
root cause of seals failure and replacing the existing types of seals to better types. Using
Weibull reliability analysis of non-repairable parts and the methodology explained in
References [Sheikh, et al, 2000], [Sheikh, 1990], [Sheikh, 1991], the future needs of seals
and bearings can be determined for a specified planning horizon.

ACKNOWLEDGEMENT

The authors acknowledge the support provided by Saudi Aramco; Ras Tanura Refinery
management and King Fahd University of Petroleum and Minerals through the research
project FT/2000-06,

REFERENCES

1. Abernathy, R. B., 1996, The New Weibull Handbook, Gulf Publishing Co., pp.2-3.
2. Ben-Daya, M. and S. O. Dufuaa, 2000, “Maintenance modeling area,” in Maintenance
Modeling and Optimization, Kluwer Academic Publishers, pp.3-26.
3. Black, H. P. and F. K. Geitam, 1983, Practical Machinery Management for Process Plants,
Gulf Publishing Company, pp.514-525.
4. Blischke, W. B. and D. N. P. Murthy, 2000, “Reliability : Modeling, Prediction and
Optimization,” John Wiley and Sons,Inc.
5. Jeong, H. S. and E. A. El-Sayed, 2000, “On-line surveillance and monitoring,” in
Maintenance Modeling and Optimization, Kluwer Academic Publishers, pp.309-336.
6. Kapur, K. C. and L. R. Lamberson, 1977, Reliability in Engineering Design, John Wiley,
New York.
7. Kelly, Anthony, 1997, Maintenance Strategy, Butterworth, pp.56-57.
8. Kolarik, W. J., 1995, Creating Quality, McGraw Hill, New York.
9. Lewis, E. E., 1987, Introduction to Reliability Engineering, John Wiley & Sons, pp.96-99.
10. Peterlik, H., 1995, “The validity of Weibull estimators,” Journal of Material Science,
Vol. 22, pp.1972-1976.

Top
Vol. 4. 486 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

11. Samaha, M. E., Oct. 1997, “Effective utilization of equipment failure history through
computerized maintenance management system,” ASME-ASIA ‘97 Congress and Exhibition,
pp. 2-8.
12. Sheikh, A. K., 1990, “A statistical method of calculating the spare parts requirement,”
Symposium on Maintenance, Planning and Operation, Riyadh, Saudi Arabia.
13. Sheikh, A. K., 1991, “Strategies in spare parts management using reliability engineering
approach,” Engineering Cost Production Economics, 1991.
14. Sheikh, A. K., D.M.Alanazi and M. Younas, 2002, “Weibull reliability and maintainability
analysis of pumps used in oil industry,” Submitted to International Journal of Quality and
Reliabilty Management, 2002.
15. Sheikh, Anwar K., Muhammed Younas and Abdul Raouf, 2000, “Reliability based spare parts
forecasting and procurement strategies,” in Maintenance Modeling and Optimization,
Kluwer Academic Publishers, pp.81-107.

P7

17%
33%

33%
17%

Replacing mechanical seal Repairing control valve

Bearing replacement Suction valve

Figure 1. Modes of failure of most critical bad actor pump P7.

P3

22% 11%

23%

22%
22%

Replacing mechanical seal Repairing shaft

Overhaul Suction valve
Operation upset

Figure 2. Modes of failure of most critical bad actor pump P3.

Top
 Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 487

45
36%
40

35

30
Number of Failures

23.4%
25

20

15 11.7%
9%
10 6.3%
4.5%
5 2.7% 1.8% 1.8% 0.9% 0.9% 0.9%
0
Mechanical Seal

Suction Valve
Shaft
Impeller

Discharge Valve
Bearing

Gaskets
Overhaul

Operation Upset

Control Valve
Coupling
Casing
Failure Mode

Figure 3 Failure modes for the most critical bad actors pumps.

Top
Vol. 4. 488 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

1.00

Y = 1.1506 X - 2.2543
0.50 R2 = 0.9539
β = 1.15
0.00 η = 7.09 months
ln ln(1/(1-F(t)))

-0.50

-1.00

-1.50

-2.00

-2.50
0.00 0.50 1.00 1.50 2.00 2.50 3.00
ln(t)
Figure 4. Weibull plot for most critical bad actor pump P3.

1.00

Y = 1.4829 X - 3.348
0.50 R2 = 0.9697
β = 1.48
0.00 η = 9.56 months
ln ln(1/(1-F(t)))

-0.50

-1.00

-1.50

-2.00

-2.50
0.00 0.50 1.00 1.50 2.00 2.50 3.00
ln(t)
Figure 5. Weibull plot for most critical bad actor pump P7.

Top
Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 489

2.00

Y = 1.6643 X - 5.5494
1.00 R2 = 0.9576
β = 1.66
0.00 η = 27.5 months
ln ln(1/(1-F(t)))

-1.00

-2.00

-3.00

-4.00

-5.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
ln(t)
Figure 6. Weibull analysis of seals for all types of most critical bad actors pumps.

1.50

1.00 Y = 1.7384 X - 5.8798

R2 = 0.8778
0.50 β = 1.74
η = 29.4 months
0.00
ln ln(1/(1-F(t)))

-0.50

-1.00

-1.50

-2.00

-2.50

-3.00
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50
ln(t)
Figure 7. Weibull analysis of bearings for all typesof most critical bad actors pumps.

Top
Vol. 4. 490 Anwar Khalil Sheikh, Muhammad Younas, and Dahham Matar Al-Anazi

2000000

1800000 Y = 350273X - 6.99 x 108

R2 = 0.953
Accumulative repairing cost in US$

1600000

1400000

1200000

1000000

800000

600000

400000

200000

0
1994 1995 1996 1997 1998 1999 2000 2001
ln(t)
Figure 8. Accumulative repair costs versus time of operation for the most critical
bad actors pumps (all failure modes).

140000

Y = 21892X - 4.37 x 107

120000 R2 = 0.953
Accumulative repairing cost in US$

100000

80000

60000

40000

20000

0
1994 1995 1996 1997 1998 1999 2000 2001
ln(t)
Figure 9. Average accumulative repair costs versus time of operation for the
most critical bad actors pumps (all failure modes).

Top
 Weibull Analysis of Time Between Failures of Pumps Used in an Oil Refinery Vol. 4. 491

1200000

Y = 22476 X - 162064
1000000 R2 = 0.986
Accumulative repairing cost in US$

800000

600000

400000

200000

0
0 10 20 30 40 50 60
Pumps operation time in months
Figure 10. Accumulative repairing cost for seals of all pumps versus operation time.

140000

Y = 21892X - 4.37 x 107

120000 R2 = 0.953
Accumulative repairing cost in US$

100000

80000

60000

40000

20000

0
1994 1995 1996 1997 1998 1999 2000 2001
ln(t)
Figure 11. Average accumulative repair costs versus time of operation for the
most critical bad actors pumps (all failure modes).

Top

View publication stats