Gupta2009

A control chart guided maintenance policy selection
Suprakash Guptaa
*, Jhareswar Maitib
, Ravi Kumarc
and Uday Kumard
a
Department of Mining Engineering, Institute of Technology, BHU, Varanasi, Uttar Pradesh,
India; b
Department of Industrial Engineering and Management, Indian Institute of Technology,
Kharagpur, West Bengal, India; c
Reliability Engineering Centre, Indian Institute of Technology,
Kharagpur, West Bengal, India; d
Division of Operation and Maintenance, Lulea˚ University of
Technology, Lulea˚, Sweden
(Received 29 July 2008; final version received 7 February 2009)
In addition to adopting new and advanced technology, effective maintenance
management is one of the key parameters for meeting ever increasing
production targets both in terms of quality and quantity. A lot of capital
may be drained off in the absence of sound maintenance policy and there is
ample scope to minimise this loss through proper maintenance decisions. This
study aims to develop a scheme for maintenance policy decisions through a
time-based control chart (t-chart) that monitors the failure process of the
component or system under investigation. On the basis of the nature of the
control chart, eight maintenance policy zones were identified. A case study on
an Armoured Flexible Conveyor (AFC) used in an underground coal mine in
India was undertaken to illustrate the use of the developed scheme for making
maintenance policy decisions.
Keywords: control chart; failure process monitoring; maintenance policy; Weibull
distribution
Introduction
Maintenance is an increasing portion of operating cost and requires to be
thorough planning to ensure that each maintenance dollar is well spent [1].
Therefore, organisations should have a maintenance policy to guide the
maintenance department in planning and scheduling of all maintenance activities.
A maintenance policy is a statement of principle, usually containing a set of rules,
used to guide maintenance management in decision-making for maintenance
planning. There are two aspects of maintenance planning, namely, identification
of correct maintenance actions from an analysis of past and recurrent failures,
and planning and scheduling of all maintenance actions directed to improve the
efficiency of the maintenance function [2]. Maintenance policy normally addresses
maintenance related queries like when to do maintenance, how to do maintenance
or whether the present schedule for maintenance is apt (i.e. cost effective/as per
with the failure process) or needs modification. A suitable failure process
*Corresponding author. Email: suprakash@bhu.ac.in
International Journal of Mining, Reclamation and Environment
Vol. 23, No. 3, September 2009, 216–226
ISSN 1748-0930 print/ISSN 1748-0949 online
Ó 2009 Taylor & Francis
DOI: 10.1080/17480930902916478
http://www.informaworld.com

monitoring scheme may be adopted that will serve as an effective tool to the
management in framing the maintenance policy.
A control chart is an useful tool for process monitoring in any industry. The
basic purpose of developing a control chart is to identify assignable causes of
deviations of process parameters. The principles behind developing a control chart
are (i) all processes behave in a typical manner based on their inherent
characteristics, which are termed as common cause variation and (ii) in the presence
of assignable causes, the process behaviour changes and the symptoms of these
changes are manifested by the process characteristics [3]. Each component has its
own characteristic life and it fails after this because of the common cause of ageing.
As the failure of the components of equipment is a random process and caused by
common causes, control chart principles can be applied for monitoring the failure
process of the components of equipment to detect the presence of assignable causes
and their removal. For example, the assignable causes may be sudden increase of
stress and human error, and common causes are the inherent characteristics of the
machine component.
Xie et al. [4] has adopted the control scheme proposed by Chan et al. [5] for
monitoring the reliability of a component or system. This approach can ease out
the problem of large data requirement in standard control charts like Shewart
control charts and can easily be adapted to failure process monitoring in
maintenance engineering. Cassady et al. [6] have used an "X chart with an age
related preventive maintenance policy for optimisation of the operating cost of
manufacturing equipment. In this article, the control chart principle developed by
Xie et al. [4] is applied to monitor the failure process of the components of an
Armoured Flexible Conveyor (AFC) being used for conveying coal from an
underground Longwall face. A framework for maintenance policy decisions is
developed based on the failure causes of the components of the AFC as depicted
in the control charts.
Development of control chart for monitoring of time between failure
The failures of a component can be described by a continuous statistical distribution
like the Weibull and lognormal distributions. However, the Weibull distribution has
one very important property in that the distribution has no specific characteristic
shape. In fact, depending upon the values of the parameters, it can be shaped to fit a
set of data that cannot be characterised as a particular distribution other than a
Weibull distribution with certain shaping parameters. The construction of control
charts for Weibull distribution based on Xie et al. [4] is given below.
The cumulative distribution function (CDF), F (t) of a two-parameter Weibull
distribution is,
FðtÞ ¼ 1 À exp
t
Z
b
; t 0 ð1Þ
where Z ¼ scale parameter and b ¼ shape parameter.
Xie et al. [4] termed their control chart t-chart to monitor the time between
failures (TBFs) of components. The t-chart is a function of the scale (Z) and
shape (b) parameters of the Weibull Distribution. Like standard control charts,
‘‘the control limits for t-chart are defined in such a manner that the components’
International Journal of Mining, Reclamation and Environment 217

failure process is considered to be out of control if the time to observe
exactly one failure is less than lower control limit (LCL), TL or greater than
upper control limit (UCL), TU’’. As the control limits are set based on the
properties of sampling distribution, there is always a chance of false alarm when
the failure process is characterised by the common causes with a very low
probability value and is termed as Type-I error (a). The acceptable level of Type-I
error (a) depends on the material of the component used, maintenance
effectiveness, human performance and environmental factors. For an acceptable
Type-I error (a), the control limits can be calculated from the following
expressions:
Central Limit; TC ¼ CL ¼ Z À 1 þ 1=bð Þ½ Š ln 2 ð2Þ
Lower control limit, TL ¼ LCL ¼ Z ln 1=1 À a=2ð Þ½ Š1=b
ð3Þ
Upper control limit, TU ¼ UCL ¼ Z ln 2=að Þ½ Š1=b
ð4Þ
Development of a framework for maintenance policy decisions
Monitoring of the component’s failure by t-chart shows its failure behaviour and
helps to take required maintenance actions following the maintenance policy as
given in Table 1. Components’ performance as well as the effectiveness of the
present maintenance schedule is said to be satisfactory when the TBFs points lie
in between the CL and UCL, and average and within control when the TBFs lie
around the CL and above the LCL provided that there are no systematic patterns
on the TBF values. The performance of the present maintenance schedule is
unsatisfactory and the component is deteriorating when the TBFs points are
around the LCL in the t-chart. When the TBFs data show any systematic pattern
it is also a cause of concern, as a pattern reflects the possible presence of
assignable causes. TBFs less than LCL mean that the failure occurrence rate has
increased resulting in a decrease in the failure time. On the other hand, if TBF
exceeds the UCL, this signals a possible process improvement. If this happens,
the management should look for possible causes for this improvement and the
follow-up actions should aim to maintain them. The policy or approach that had
helped this improvement should be maintained or even implemented for other
similar systems or equipment.
Application of control chart based maintenance policy – a case study
The developed control chart based maintenance policy was applied to an AFC
operated in a Longwall mine in India. The schematic for applying the methodology
is shown in Figure 1. The AFC is divided into manageable components and TBF
data for each of the components were collected. The failure data were analysed and
control charts for each of the components were developed. Finally, the components
were classified based on their control chart characteristics and an appropriate
maintenance policy for each class was outlined. The details of the application are
described in the following sections.
218 S. Gupta et al.

Table 1. Control chart characteristics and its meaning in maintenance policy.
Zone no Control chart characteristics Possible causes
Maintenance
policy
0 All points are within control limits
and no systematic pattern exists.
Common causes
of failures
Follow present
maintenance
policy.
I t-chart of TBFs show some points are
outside the upper control
limit (UCL) including points
on the limit line (UCL).
Signals a possible
process
improvement or
adoption of
better
maintenance
scheme or both.
For the ﬁrst two
causes, maintain
the present
maintenance
policy. Follow
the same policy
for alike
components.
Component may
be under-loaded
due to misﬁt
that results in
delegating its
load over others
or over-design.
For the last cause,
redesign system.
II t-chart of TBFs show some
points are outside the lower
control limit (LCL) including
points on the limit line (LCL).
Signals possible
process
deterioration
and component
has reached in
wear out phase
(old) or in burn-
in phase (new).
Depending on the
age of the
component,
replace if it is
old or watch for
some more time
before framing
its maintenance
schedule when it
is new.
Presence of
extraneous
causes such as
roof fall in
underground
coal mines.
Some assignable
causes likely to
be present and
need to be
investigated.
III t-chart of TBFs show all points
within control limits but several
points’ line up consecutively
only between CL and UCL.
Signals possible
permanent
process
improvement,
e.g. adoption of
better
Follow the
changed
maintenance
policy.
(continued)

Table 1. (Continued).
Maintenance
policy
maintenance
policy.
May increase the
periodic
inspection or
replacement
interval.
IV t-chart of TBFs show all points
within control limits but
several points’ line up
consecutively only between
CL and LCL.
Signals possible
permanent
process
deterioration,
e.g. adoption of
inadequate
maintenance
policy or
inferior quality
components/
spares.
Change the
present
maintenance
policy.
May decrease the
periodic
inspection or
replacement
interval.
Institute intensive
quality check.
V t-chart of TBFs show all points within
control limits but there is continuous
rise or fall in a series of points.
Component is
overstressed or
under-stressed,
possibly due to
change in
operating
conditions or
change in
maintenance
crew or spare
parts quality.
Investigate the
changes in
operating
conditions or in
maintenance.
Necessitates
modiﬁcation in
present
maintenance
schedule in
coherence with
the changed
operating
conditions.
VI t-chart of TBFs show all points within
control limits and the points show
the same pattern of change over
equal intervals.
It’s a complex
phenomenon
and is very
diﬃcult to
evaluate the
The only way is to
follow the
changes in TBF
closely, and
make a technical
(continued)
220 S. Gupta et al.

Mine description
A case study was undertaken in two longwall panels (Panel – A Panel – B) of an
underground coal mine in Southern India. The A panel is located between TG
35 LS and MG 40 LS and worked with a face length of 158 m. A total of 434,416
tonnes of coal was extracted in 12 months from this panel. The B panel is located
between TG 41 LS and MG 44 LS and worked with a face length of 136.5 m. A
total of 551,577 tonnes of coal was extracted in 8 months from this panel. Detailed
geology and layout of a coal face can be found in the article of Gupta et al. [7].
Failure data of the AFC used in these two longwall panels were collected through
field visits.
Data collection
Performance monitoring of components calls for sufficient information on the time
to failure characteristics of that unit under the specified conditions of use. Although
data from life tests serve as a good source of this information for Longwall mining
equipment like AFC, this sort of data is not available and normally collected from
the field. For the present study, various failure and maintenance information of the
AFC for a 2 year period was collected from the records kept in the mine, and the
Table 1. (Continued).
Maintenance
policy
causes. One of
the causes of
such behaviour
may be the
periodic
replacement of
another
component
which influences
the failure
characteristic of
the component
of interest.
decision
regarding
maintenance
and/or
operations.
VII t-chart of TBFs show all points within
control limits but the points stick
close to the CL or control limit
lines (UCL or LCL).
Indicating mixing
of failure data
obtained for
different
components or
similar
components
under different
operating
conditions.
Data recording
and
classification
system is
inadequate and
need to be
changed.

concerned persons were consulted for grouping of the data. This period includes the
salvage period between these two panels. Necessary correction for the salvage period
was made to the TBF data for all the components.
Calculation of time between failures
TBFs for each component of AFC were calculated from the collected ﬁeld data in
following steps and tabulated in Table 2.
Figure 1. A ﬂow-chart for control chart based maintenance policy selection.
222 S. Gupta et al.

Table 2. Time between failure data of the components of AFC.
Component no (Ci) Component name Time between failure (h)
1 Adjusting pan 22, 193.75, 1332.75, 850, 433
2 Attachment
bracket
628, 1858.5, 340.5, 657.25, 255.25, 1140.75, 49.5,
594.75
3 Bolts 2282, 1432, 619.25, 1314, 1280
4 Bursting disc 872.25, 1120.25, 699.75, 727.5, 1481.75, 861.25
5 Chain connector 1106.5, 1233.25, 1242.25, 404.5, 1210.25, 1309.75
6 Chain links 16.5, 252.25, 44, 91.25, 340.75, 104, 16.5, 81.75,
59.5, 80.25, 11, 233, 224.75, 170, 126, 113.5,
31.25, 164.75, 112, 71.25, 550, 105, 131.5, 48.25,
70, 16.5, 16.5, 5.5, 154, 27.5, 128.5, 129.75, 27,
53.75, 49.5, 179.25, 110, 117.5, 66, 162.5, 321.5,
178.75, 16.5, 27.5, 16, 27.5, 32.75, 15.75, 81.25,
32.25, 121, 49.5, 98.75, 208.75, 133.75, 82.25, 44,
90.25, 55, 16.5, 30.75, 161.25, 100.75, 381.75, 36,
82.5, 115.5, 99, 227.75
7 Deck plate 406.75, 462.75, 204.5, 413.25, 1069.75, 199.75,
669.75, 609.25
8 Drive shaft 445, 890, 134.75, 714.75
9 Flight bars 236.25, 16, 559.5, 16.5, 120.5, 16.5, 16.5, 32.25, 43,
33, 71.5, 49.5, 16.5, 153.75, 27.5, 16.5, 33,
139.75, 427.5, 16.5, 16, 27.5, 33, 294.75, 15.5,
199, 16.25, 31.75, 337.5, 31, 287.75, 32.75, 22,
81.25, 16.25, 16.5, 70.25, 33, 200.75, 16.25, 60,
55.75, 16.5, 16.5, 2.75, 5.5, 102.75, 11, 49.5, 49,
11, 104, 125.25, 28.75, 16.5, 12.75, 11, 16.5,
131.25, 31.75, 33, 125, 179.75, 81.5, 16.25, 82.25,
63.75, 16.5, 16.5, 217.25, 32.25, 203.5, 16.25,
38.5, 46.5, 16.5, 5.75, 59.5, 32
10 Fluid coupling
(MG)
1074.25, 309.75, 41.25, 700.25, 8.25, 372.5, 609.75,
185.75, 1615.75, 531.25, 27.5, 102.5, 195, 808.75,
38.5, 529.75
11 Fluid coupling
(TG)
406.75, 54.25, 239.75, 11, 466.75, 354.25, 379.25,
562.25, 597.25, 1058
12 Fusible plug 921.25, 3426.75, 241.25, 2321.5, 151.25, 632.25,
115.5
13 Gear box 88.25, 148.5, 39, 100, 235.5, 104, 120
14 Inspection
door
73, 1303.5, 720, 698.25, 33
15 Oil seal (MG) 257.25, 493.5, 394.75, 567.25, 716.25, 548.25, 292,
348.5, 472.5, 358.25, 451.75, 872.75, 567.25, 632
16 Oil seal (TG) 65.25, 273.75, 395.25, 621.75, 678.5, 1035, 413.5,
669.75, 1151.5
17 Pan connector 1123, 3557, 16, 304.75, 147, 173.75, 678, 582.75,
216
18 Ramp pan 177.75, 21.25, 15.75, 30.75, 681.75
19 Spill plate 2688.25, 1124.75, 1139.25, 97.75, 386.25, 124.5,
434.5, 195, 63.25, 148.25, 170.5, 16.5, 340.75,
504
20 Spill pan 2688.25, 2345.25, 527.25, 402.25, 436.25, 281.25
21 Sprocket assembly 628, 1825.5, 619.25, 601.25, 1431.75, 535
22 Sprocket bearings 460.25, 435, 147.5, 238, 745.75, 1041.5
23 Triangular socket 27.5, 16.5, 11, 942.75, 141.5, 1058.5, 637

. Step 1: Note the dates and times of failures from the field record.
. Step 2: Find the number of working days between two successive failures of the
same component (dn). This calculation was done for the total study period, i.e.
from the starting of the first panel to the end of the second panel.
. Step 3: Calculate the equipment working hours (wh) between two successive
failures of the same component by multiplying the number of days (dn) with
16.5, the number of working hours per day. As the mine was operating in four
shifts of 6 h duration and the equipment remains idle in the pre-maintenance
shift. Average equipment operating hours per shift was 5.5 h as half an hour is
lost due to change of shifts.
. Step 4: Calculate the TBFs by subtracting the number of breakdown hours of
the equipment owing to failure of other components within the period of
interest from the equipment working hours (wh). Incorporate proper correction
for the exact time of failure, i.e. in which shift and at what time the failure was
reported on the particular dates.
Analysis of failure data
Once the TBFs for all the components of AFC were calculated, the next step is to fit
the data to a distribution by which they can be best described and subsequently, the
parameters of the fitted distribution are estimation. As the result of trend test suggests
that the data are identically distributed and absence of correlation advocates their
Table 3. List of scale parameter, shape parameter and control limits for the components of
AFC.
Component
no (Ci) Component name
Scale
parameter
(Z)
Shape
parameter
(b) CL UCL LCL
1 Adjusting pan 473.792 0.969862 332.6192 3320 0.5213
2 Attachment bracket 793.152 0.968203 556.8324 5576 0.8625
3 Bolts 1611.63 2.11333 989.1189 3938 70.715
4 Bursting disc 1074 1.40235 669.1386 1871 154.047
5 Chain connector 1311.07 1.04757 890.2909 7951 2.3908
6 Chain links 111.852 1.12578 74.28683 599 0.3161
7 Deck plate 581.54 1.83264 358.2186 1629 15.80
8 Drive shaft 666.497 1.1628 438.1832 3381 2.2707
9 Flight bars 108.879 1.0879 72.82737 618 0.2508
10 Fluid coupling (MG) 582.867 0.791018 460.2906 6343 0.1374
11 Fluid coupling (TG) 479.744 0.79393 378.8492 5175 0.1166
12 Fusible plug 1083.6 0.858649 809.9091 9770 0.4933
13 Gear box 164.128 2.07487 150.7456 408 6.7962
14 Inspection door 615.324 0.810079 477.6849 6330 0.1766
15 Oil seal (MG) 555.376 3.2836 345.4141 987 74.257
16 Oil seal (TG) 707.044 1.2276 457.678 3292 3.2513
17 Pan connector 678.771 0.709884 584.296 9703 0.0616
18 Ramp pan 153.396 0.581485 166.8578 3945 0.0018
19 Spill plate 484.085 0.856476 363.7308 4389 0.2161
20 Spill pan 1244.78 0.962259 877.7954 8857 1.2977
21 Sprocket assembly 1117.23 1.68444 690.9487 3428 22.114
22 Sprocket bearings 591.227 1.45999 370.7966 2155 6.4035
23 Triangular socket 334.65 0.811455 259.7918 3429 0.0974
224 S. Gupta et al.

independency, a classical distribution may therefore fit to the data [8]. Studies by
Ramani et al. [9] and Gupta et al. [10] show that the failure behaviour of longwall
equipment can be best described by a two-parameter Weibull distribution. Mann’s
test was done for goodness-of-fit tests of the TBF data to two-parameter Weibull
distribution. Test results are in favour of accepting the null hypothesis H0:- TBF of
AFC components are Weibull. Parameters of the two-parameter Weibull distribution
for the components of the AFC were estimated using STASTICA (version 5) as given
in Table 3.
Results and discussions
Table 3 presents the control limit values for the components of AFC. The TBF
control charts show the pattern of the failure process of the components. TBF points
in conjunction with the control lines show deterioration or improvement of
components’ failure process. The patterns observed in TBF monitoring are
attributed to maintenance effectiveness or ineffectiveness of the case study
component. t-charts for all the components of AFC were prepared as shown in
Figure 2 for chain links. On the basis of possible control chart characteristics of the
components, they were classified into eight maintenance zones. The t-chart of chain
Table 4. AFC components with their zone of maintenance.
Zone of maintenance Component no (Ci)
Total number
of components Percentage
0 7, 12, 13, 21 4 17.39
I 4 1 4.35
II 6, 8, 9, 10, 14, 17, 18, 19, 23 9 39.13
III 15 1 4.35
IV Nil 0 0.00
V 1, 3, 11, 16, 20, 22 6 26.09
VI Nil 0 0.00
VII 2, 5 2 8.70
Figure 2. T-chart for chain links.

links depicts a number of points below or on the LCL, clearly indicating its berth in
the maintenance zone II. Mostly failures of the chain links were caused because of
ageing as it was old, and sometimes due to the impact of falling rock or coal. The
number of components along with their zones of maintenance is shown is Table 4. Of
the 23 components of the AFC studied, 39.13% components fall under zone II
followed by zones 0 (17.39%) and V (17.39%). Maintenance policy for each of zone
of maintenance is given in Table 1. For example for the components falling under
zone II, the maintenance policies are:
. Depending on the age of the component, replace if it has reached its wear out
phase (old) or watch for some more time before fixing up its maintenance
schedule when it is new.
. Investigate whether there is any assignable cause leading to significant process
deterioration and eliminate the cause.
Conclusions
The general conclusion of this study is that control chart (t-chart) is a very simple
and effective tool for monitoring the failure process of a component. It is helpful to
judge the maintenance effectiveness i.e. whether the present maintenance is in
coherence with the failure process or required any change. It can be used as an
effective tool for maintenance management in making maintenance decisions. This
study shows that the overall status of maintenance in the AFC is not up to the mark.
The management should pay proper attention to this ever neglected but very
important aspect.
Acknowledgements
The authors thank the anonymous reviewers for their constructive comments and suggestions
on an earlier version of the article.
References
[1] D. Nyman and J. Levitt, Maintenance Planning, Scheduling Coordination, Industrial
Press Inc., New York, NY, USA, 2001, p. 228.
[2] B. Bhadury and S.K. Basu, Terotechnogy: Reliability Engineering and Maintenance
Management, 1st ed., Asian Books Private Limited, Delhi, 2003, p. 364.
[3] K. Ishikawa, Guide to Quality Control, Asian Productivity Organization, Tokyo, Japan,
1998, p. 225.
[4] M. Xie, T.N. Goh, and P. Ranjan, Some effective control chart procedures for reliability
monitoring, Reliab. Eng. Syst. Saf. 77 (2002), pp. 143–150.
[5] L.Y. Chan, M. Xie, and T.N. Goh, Cumulative quality controls charts for monitoring
production processes, Int. J. Prod. Res. 38 (2000), pp. 397–408.
[6] C.R. Cassady, R.O. Bowden, L. Liew, and E.A. Pohl, Combining preventive maintenance
and statistical process control: A preliminary investigation, IIE. Trans. 32 (2000), pp. 471–
478.
[7] S. Gupta, N. Ramkrishna, and J. Bhattacharya, Replacement and maintenance analysis of
a Longwall shearer using fault tree technique, Min. Technol. 115 (2006), pp. 49–58.
[8] H.E. Ascher and H. Feingold, Repairable Systems Reliability: Modeling, Inference,
Misconceptions and their Causes, Marcel Dekker Inc., New York, 1984.
[9] R.V. Ramani, A. Bhattacherjee, and R. Pawlikowski, Reliability maintainability and
availability of Longwall mining system, Miner. Resour. Eng. 2 (1989), pp. 3–18.
[10] S. Gupta and J. Bhattacharya, Reliability analysis of a conveyor system using hybrid data,
Qual. Reliab. Eng. Int. 23 (2007), pp. 867–882.
226 S. Gupta et al.

Gupta2009

More Related Content

What's hot (20)

Viewers also liked (6)

Similar to Gupta2009 (20)

Recently uploaded (20)

Gupta2009