Downtime Priorities, Jack-Knife Diagrams, Business Cycle

Downtime Priorities,
Jack-knife Diagrams,
and the
Business Cycle
Peter F. Knights
Associate Professor, Mining Centre, Faculty of Engineering, Catholic University of Chile,
e-mail: knights@ing.puc.cl

Abstract 1. Introduction
Pareto histograms are commonly used to determine maintenance In the late 19 th Century, the Italian Engineer Vilfredo Pareto (1842-
priorities by ranking equipment failure codes according to their 1923) constructed histograms of the distribution of wealth in Italy and
relative downtime contribution. However, these histograms do not concluded that 80 percent of the country’s wealth was owned by 20
readily identify the dominant variables influencing equipment percent of the nation’s population. This trend was later found to be
downtime, namely the failure frequency and mean time to restore representative of the distribution of other data populations. The 80:20
service (MTTR) associated with each failure code. This paper rule, and variations such as ABC analysis that uses an 80:15:5
advances an alternative method for analysing equipment downtime classification rule, or “top ten” list generation, are now routinely used
using logarithmic (log) scatterplots. Log scatterplots preserve the in many fields of study. As applied to the field of maintenance
engineering, Pareto analysis is commonly used for identifying those
ranking scheme of Pareto histograms, but provide additional
failure codes responsible for the majority of equipment maintenance
information content regarding failure frequencies and MTTR. By
cost or downtime. Based on the failure codes identified, action plans
applying limit values, log scatterplots can be divided into four
can be elaborated to lower maintenance costs or improve equipment
quadrants enabling failures to be classified according to acute or availability and reliability.
chronic characteristics and facilitating root cause analysis. In
However, Pareto Analysis suffers from a number of deficiencies:
addition, the graphs readily permit the identification of problems
Firstly, maintenance costs and downtime are the product of two
affecting system reliability, availability, and maintainability. Depending
factors; the number of failures that occurred in a particular time frame
on the relative importance of maintenance and repair costs to the
and the average associated repair cost, or mean time to restore service
economic consequences of failures (including lost production costs), (MTTR). A Pareto histogram based on downtime (or cost) alone cannot
a V-shaped limit can be established for determining downtime determine which factor, or factors, are dominant in contributing to the
priorities. These graphs have been christened “jack-knife” diagrams downtime or cost associated with particular failure codes.
after the shape of the V-shaped limit. As prices vary over the business Secondly, Pareto histograms prepared for failure codes ranked
cycle, the jack-knife limit can change orientation, re-focusing according to repair cost, equipment downtime, failure frequency and
downtime priorities to reflect changes in business priorities. By MTTR will generate four quite distinct lists of failure code priorities.
graphing the trend of failure data over successive time periods, log These four lists must be combined in some way to generate a reduced,
scatterplots provide a useful visual means of evaluating the manageable set of downtime priorities. Which failures should be
performance of maintenance improvement initiatives. Since their selected in order to best improve business performance? A graphical
introduction in 1999, log scatterplots and jack-knife diagrams have means for simultaneously visualizing equipment availability, reliability
been profitably employed by a number of mining companies and and maintainability problems as a function of equipment failure codes
mining equipment suppliers world-wide. would be extremely useful to this process.

15
 Downtime Priorities, Jack-knife Diagrams, Business Cycle

Thirdly, Pareto analysis of equipment downtime may miss or functional failure type. More detailed Pareto graphs are prepared
identifying: (i) individual events having high associated downtime or for the downtime associated with those components or functional
(ii) frequently occurring failures that consume relatively little downtime failures judged to be the most significant contributors of downtime.
yet cause frequent operational disturbances. An example of the former There are two potential problems with the use of stratified Pareto
is the failure of the transmission in a mechanical mining truck. An analyses: (i) because hierarchical Pareto graphs are only prepared
example of the latter are repairs to the truck’s driving lights. Failures for the significant contributors of system downtime, failures
that frequently re-occur often have significant hidden costs. For associated with less significant components or functional failures will
example, if the truck has to return to the truck shop to have a light not be explored. It is possible that we may miss identifying a
replaced, the time lost travelling to and from the shop may dramatically component, or failure mode that offers significant potential for
increase the opportunity costs associated with lost production. reliability improvement. (ii) The same failure mode may appear in
Fourthly, when there is a lot of data to analyse we commonly use several of the lower level Pareto histograms. We may fail to identify
data stratification or hierarchical decomposition techniques. A Pareto or underestimate the relative importance of these common cause
histogram is prepared for downtime data grouped by major equipment failure modes.

14.0% 100.0%
90.0%
12.0%
80.0%
10.0% 70.0%
60.0%
8.0%
50.0%
6.0%
40.0%

4.0% 30.0%
20.0%
2.0%
10.0%
0.0% 0.0%
1 2 11 3 10 7 12 8 5 15 6 9 4 17 14 16 13
Failure Code
Figure 1: Pareto hisogram of unplanned shovel electrical downtime

Code Description Quantity Duration % Time % Cum.

(min)
1 Electrical inspections 30 1015 13.0 13.0
2 Damaged feeder cable 15 785 10.1 23.1
11 Motor over temperature 36 745 9.6 32.6
3 Change of substation or shovel move 27 690 8.8 41.5
10 Overload relay 23 685 8.8 50.3
7 Auxiliary motors 13 600 7.7 58.0
12 Earth faults 7 575 7.4 65.3
8 Main motors 12 555 7.1 72.5
5 Power cuts to substations 21 395 5.1 77.5
15 Air compressor 8 355 4.6 82.1
6 Rope limit protection 10 277 3.6 85.6
9 Lighting system 26 240 3.1 88.7
4 Coupling repairs or checks 15 225 2.9 91.6
17 Over current faults 6 220 2.8 94.4
14 Control system 7 165 2.1 96.5
16 Operator controls 5 155 2.0 98.5
13 Miscellaneous 9 115 1.5 100
TOTAL 270 7797 100

Table 1: Unplanned shovel electrical downtime.

16
 Downtime Priorities, Jack-Knife Diagrams, Business Cycle

Finally, Pareto histograms are not generally useful for trending D = ∑ Downtimei (3)
i
comparisons. It can be difficult to directly compare ranked histograms
of costs or downtime for two different time periods since the relative
The total number of failures is:
position of failure codes can change from one period to the other.
N = ∑ ni (4)
This paper outlines a simple, but powerful way of analysing data i
in order to address these shortcomings. The paper builds on and
updates earlier work published by Knights (2001, and 1999). Letting Q be the number of distinct failure codes used to
categorise the downtime data, the threshold limit for acute failures
2. Logarithmic Scatterplots can be defined as:
The most convenient way of introducing the new methodology is or LimitMTTR = D (5)
N
via an example. Table 1 lists the unplanned downtime recorded for
electrical failures in a fleet of 13 cable shovels at an open pit copper and the threshold limit for chronic failures can be determined as:
mine located in northern Chile. The data was collected over a one-
Limitn = N (6)
month period. Figure 1 shows the Pareto histogram for the unplanned Q
electrical failures, with failure codes ranked in descending order
according to their downtime contribution. Applying the 80:20 rule, it is In the case of the unplanned electrical failures for the fleet of
evident that priority should be given to failure codes 1, 2, 11, 3, 10, 7, shovels, D=7797 minutes, N=270 and Q=17. Therefore, the limit value
12, 8 and 5 (failure code 1, electrical inspections, is associated with for acute failures is 7977/270 =28.9 minutes and the limit value for
unscheduled call-outs). Of these, maintenance can do little to reduce chronic failures is 270/17=15.9 repairs.
the downtime associated with failures codes 3 (substation changes
or shovel moves) and 5 (substation power cuts). 4. Identifying Reliability, Availability and
Maintenance downtime can be represented by the equation: Maintainability Problems
Downtimei = ni X MTTRi (1) Having identified the threshold limits, logarithmic scatterplots
provide an easy means for identifying reliability, availability and
where Downtime i is the downtime associated with the ith failure maintainability problems. Strictly speaking, reliability is a probability
code and ni and MTTRi represent the number of failures, and the of survival and a function of time. A common industry practice is to
mean-time-to-restore service respectively. use the mean time between failures (MTBF) as a measure of reliability.
Figure 2 shows an alternative means of presenting the failure data The MTBF is defined as;
listed in Table 1. An x-y scatterplot is used to plot mean downtime MTBF = Operating time (7)
against the number of unplanned failures for each failure code. Curves N
of constant downtime are represented by a family of hyperbolae as
shown. It can be seen that the failures that consume most downtime Shovel availability, reliability and maintainability are interrelated
are those associated with failure codes 1, 2 and 11. Thus the order of by the approximation;
priority observed in the Pareto analysis is preserved, however a
clearer picture is available as to which factor - failure frequency or Availability ≈ MTBF (8)
MTTR - is dominant. MTBF + MTTR
A disadvantage of Figure 2 is that the curves of constant downtime The relationship shows that shovel availability can be enhanced
are hyperbolae and can be difficult to plot. A solution to this is to take by increasing MTBF, decreasing MTTR, or a combination of both.
the logarithm of equations (1) and (2). Thus: Returning to the logarithmic scatterplot, the chronic failures are
log (Downtime i) = log(n i) + log(MTTR i) (2) those that contribute most to the number of observed failures, N.
These are the electrical failures that most affect the reliability from
where log refers to log10 . If an x-y graph is prepared of log(ni)
the shovel fleet under examination (see Figure 5). Finding solutions to
against log(MTTRi), the curves of constant downtime now appear as
the root causes of these failures will most increase shovel MTBF.
straight lines with uniform negative gradient (see Figure 3). These Similarly, Figure 6 shows those failures that most affect shovel fleet
logarithmic (or “log”) scatterplots can be very simply prepared using
availability. In this case, the position of the diagonal downtime limit
commercial spreadsheet software such as Excel™ by positioning the
was determined from the Pareto analysis as the line defining 80% of
mouse over the axes of the x-y graph such as that shown in Figure 2
fleet downtime. Finding solutions to the root causes of these failures
and then using the right-hand mouse button to select logarithmic
will most increase shovel fleet electrical availability. The failure code
scales for the axes. Log scatterplots simplify the identification of those
9 (lighting system) is an interesting case. It was identified in Figure 5
failures which contribute most to total equipment downtime, whilst
as a reliability problem; solving this problem would increase MTBF
continuing to permit the visualisation of the influence of failure
and could therefore be expected to also increase fleet electrical
frequency and MTTR.
availability. In fact, on average repairs to the lighting system require
Repairs that require lengthy downtime can be considered acute very little time. If it were possible to eliminate these failures, the MTTR
problems. Those failures that frequently reoccur (i.e. high n) can be for the remaining electrical failures would actually increase. Referring
considered chronic problems. By determining threshold limits, the log to equation (8), the effect of increased MTBF is largely negated by the
scatterplot can be divided into four quadrants, as shown in Figure 4. increase in MTTR and comparatively little availability gain can be
The upper quadrants denote acute failures, whilst the right-hand expected by eliminating shovel lighting failures.
quadrants denote chronic failures. The upper right-hand quadrant is
Figure 7 shows the acute failures that most affect shovel electrical
a region of acute and chronic failures
maintainability, and Figure 8 shows the combined effect of the
3. Limit Determination reliability, availability and maintainability limits. From these it can be
seen that, if the root causes of failure codes 15, 16 and 17 are
Thresholds can either be absolute values determined by company addressed, although MTTR will be somewhat reduced, fleet availability
policy, or relative values that depend on the relative magnitudes and will not be significantly affected. The reason for this is that failure
quantity of data. One approach for determining relative values is to codes 15, 16 and 17 occur infrequently and their elimination will not
use average values as follows. significantly affect fleet MTBF.
The total downtime, D, consumed by unplanned failures is given by: There is another motive besides fleet availability for addressing

17
 Downtime Priorities, Jack-Knife Diagrams, Business Cycle

Figure 2: X-Y dispersion plot of mean repair times versus number Figure 5. Shovel Fleet Reliability Problems
of failures

Figure 3: Log Scatterplot of mean repair times versus number Figure 6. Shovel Fleet Availability Problems
of failures

Figure 4: Log Scatterplot showing limit values Figure 7. Shovel Fleet Maintainability Problems

19
 Downtime Priorities, Jack-knife Diagrams, Business Cycle

acute failures; repair costs for these failures are frequently more equipment availability and reliability, over repair costs.
expensive. Not only are direct costs higher (repairs may require more However, when copper prices fall, the cost of production becomes
labour hours, specialist services or the costly expediting of spares), more significant. Thus in commodity price troughs, controlling and
but the opportunity cost of lost production per failure is higher. With reducing maintenance and repair costs will be as important as
this in mind, we are now ready to answer the question as to which maintaining equipment availability and reliability.
failures should be prioritizing in order to best improve business Figures 9 and 10 illustrate these two scenarios. Figure 9
performance. corresponds to a scenario where the opportunity cost of lost
production far exceeds shovel repair and maintenance costs. Figure
5. Downtime Priorities and the Business
10 corresponds to the scenario of a commodity price squeeze causing
Cycle production cost concern.
The discussion so far has assumed that the electrical failures For ease of construction, the availability limit has been defined as
causing unplanned downtime of the shovel fleet are (i) evident to the the line of constant downtime equal to the product of the two
operators and (ii) do not cause significant safety or environmental threshold limits calculated in equations (5) and (6). Note that the
hazard. Hidden and potentially hazardous failure modes should be availability limit effectively separates the acute and chronic quadrants
addressed and can be identified by applying Reliability-Centred into two areas: acute A and B and chronic A and B respectively. The
Maintenance (RCM) decision logic such as that developed by expression for the availability limit is:
Moubray (1997).
Assuming that the hidden and hazardous failure modes have
already been filtered from the shovel electrical downtime data,
maintenance priorities will be determined solely on the basis of the
{ ni X MTTRi = D
ni = Limitn
Q
where 0 < n i < Limitn
where n i > Limitn } (9)

economic consequences of failures and their associated repair costs. in the case of Figure 9 and:

{ }
The economic consequence of a failure includes the opportunity cost
MTTRi = LimitMTTR where 0 < n i < Limitn (10)
of lost production, the extension of fixed costs such as operator
salaries which must be paid irrespective of equipment downtime, the ni X MTTRi = D where ni > Limitn
Q
cost of maintaining an increased number of spares as a result of the
failure and the cost of maintaining redundant equipment capacity in in the case of Figure 10. The resulting graphs have been christened
order to mitigate the effects of lost production. “jack-knife” diagrams because of the inverted V shape of the limits.
In some industries, the economic consequences of plant or Because the analysis of unplanned shovel downtime due to
equipment downtime for critical equipment far outweigh repair and electrical failures was undertaken during a time of very low copper
maintenance costs. For example, the mining industry is subject to prices, the jack-knife limits shown in Figure 10 were used to establish
highly cyclical commodity prices. With copper prices currently at five- downtime priorities (see Table 2).
year highs, the opportunity cost of lost production will significantly It is valid to question the decision not to include all of the events
exceed the direct cost of repair and maintenance. In such affecting shovel reliability in the prioritized list. In effect, this means
circumstances, it is desirable to prioritize production, hence adding those events classified as “Chronic type B” to priority list.

Code Description Quantity Duration % Time Av. Time

ACUTE & CHRONIC FAILURES
1 Electrical inspections 30 1015 13.0 33.8
10 Overload relay 23 685 8.8 29.8
Sub total 21.8 63.3
ACUTE FAILURES
2 Damaged feeder cable 15 785 10.1 52.3
7 Auxiliary motors 13 600 7.7 46.2
12 Earth Faults 7 575 7.4 82.1
8 Main motors 12 555 7.1 46.3
15 Air compressor 8 355 4.6 44.4
17 Over current faults 6 220 2.8 36.7
16 Operator controls 5 155 2.0 31.0
Sub Total 41.7 339
CHRONIC FAILURES - TYPE A
11 Motor over temperature 36 745 9.6 20.7
3 Change of substation or shovel move 27 690 8.8 25.6
Sub total 18.4 46.3
CHRONIC FAILURES - TYPE B
5 Power cuts to substations 21 395 5.1 18.8
9 Shovel lights 26 240 3.1 9.2
Sub total 8.2 28

Table 2: Electrical maintenance problems prioritised according to Jack-Knife Principles.

20
 Downtime Priorities, Jack-Knife Diagrams, Business Cycle

6. Root Cause Analysis and Remedial

Action
Once a prioritised list of failure codes has been identified,
hypotheses can be made about the possible cause (or causes) of each
problem. Experienced maintenance and operating personnel are
indispensable to this process, since familiarity with the machine, the
operating environment and with maintenance and operating practices
is required.
The root causes of equipment downtime are associated with the
following broad categories; equipment design, purchase, storage,
installation or start-up, operation or maintenance (Moore, 2002). Each
of these categories can be further subdivided; for example, root
causes associated with maintenance include: planning and
scheduling problems; resource availability problems (labour, tools,
documentation, shop space); inadequate inspection or PM practices;
and poor work quality.
Figure 8. Combined Reliability, Availability and Chronic repairs are often associated with design problems
Maintainability Limits (material quality defects or design inadequacies), inappropriate
operator practices or poor quality control in upstream processes. Two
good examples of chronic repairs are provided by the data:

Figure 11: Trends in unplanned failures for

BE 495-B Cable Shovel
Figure 9: Jack-knife diagram for commodity price cycle peaks Motor over-temperature alarms (failure code 11) can result from poor
blast fragmentation, shovel abuse or high ambient temperatures. In the
former two cases, corrective action should be directed at mine
operations.
Outages to the shovel lighting system (failure code 9) typically
result from wiring damage due to structural vibration or poor filament
reliability. Redesign of the wiring harness may be one way of tackling
this problem.
Following the identification of the factors contributing to the root
causes to each failure code, a set of corrective actions should be
formulated to eliminate or mitigate the factors causing unplanned
downtime. Some maintenance actions may necessitate investment on
the part of the mine. An estimation of the expected reduction in
downtime allows the Maintenance Department to undertake a
cost/benefit evaluation of implementing the maintenance action plan.
If the cost savings are projected over say, a 3-year period, an NPV
can be calculated for the maintenance project. The advantage of this
approach is that it permits executive management to evaluate
maintenance projects alongside competing project alternatives.
Maintenance need no longer be perceived as a costly overhead, but
Figure 10: Jack-knife diagram for commodity price cycle troughs
as a strategic tool to maximise asset utilisation.
7. Trend Plots
A further benefit of logarithmic scatterplots is that they provide a
useful means of visualising trends in maintenance performance. For
example, Figure 11 shows the evolution of four failure codes from a
Experience with analysing larger sets of downtime data has shown
BE 495-B cable shovel working at an open pit copper mine in Chile.
that the priority list simply grows too large. To this extent, it is
Unplanned failures were analysed for a period of three years, 1997 to
desirable to focus on only those chronic failures that have highest
1999 inclusive. The threshold limits used in the graph were calculated
availability impact.
relative to the total unplanned failure data set for the three year period.

21
 Downtime Priorities, Jack-knife Diagrams, Business Cycle

It can be seen that significant improvement has been made with availability, and maintainability. Jack-knife limits can be established
respect to two of the failure codes over the period of the study. for determining failure priorities. The orientation of the jack-knife limit
Unplanned failures to the shovel lubrication system were chronic in depends on the relative importance of repair costs to the economic
1997 and 1998, and not classified in 1999. Similarly, the total downtime consequences of downtime, which in turn is affected by changes in
due to failures of controls in the operator cabin has decreased. the business cycle.
However, unplanned failures to the swing system (comprising the two Log scatterplots are not necessarily intended to replace traditional
swing motors, spur gears and main ring gear) are obviously an area Pareto analysis techniques. The two techniques can be beneficially
of concern, increasing from acute in 1997 to chronic and acute in 1999. used in parallel. In addition, elements of the methodology outline in
Likewise, unplanned stoppages due to motor over-temperature alarms the paper, such as the classification of acute and chronic failures, can
(alarms) also increased in both frequency and duration (data was not be usefully adapted for use in conventional Pareto analysis.
available for the 1999 period to confirm this tendency). Since their inception in 1999, log scatterplots and jack-knife
Jack-knife trend plots provide a very useful visual means of diagrams have been profitably employed by a number of mining
analysing and controlling maintenance performance. It is easy to see companies and mining equipment suppliers. In Chile, a large
the “wins and losses” of the maintenance department. Another maintenance service provider now regularly used jack-knife diagrams
potential application of jack-knife trend diagrams is to the preparation to monitor equipment downtime for controlling maintenance and repair
of maintenance budgets. A log scatterplot of the repair costs incurred (MARC) contracts.
during the most recent time period could assist a maintenance
manager to fix performance targets for forthcoming periods. It is 9. Acknowledgements
postulated that windows-based software could be developed to help The author would like to thank Komatsu Chile and Modular Mining
automate this procedure. Using a mouse, the points representing Systems Chile for supporting the development of the log scatterplot
failure codes in the log scatterplot could be selected and dragged to methodology. The paper has also benefited from the thesis work of
desired target positions. The software could then automatically
Cristian Aranguiz and Carlos Turina, final year students of the Mining
calculate the resulting cost and downtime reductions, as well as
Centre of the Catholic University of Chile.
display the corresponding operating budget for the maintenance
department. 10. References
8. Conclusions Knights, P.F. (1999) “Analysing Breakdowns” Mining Magazine, Vol.
181 No. 3, pp.165-171, September.
The use of logarithmic (log) scatterplots overcomes many of the
shortcomings of Pareto analysis for generating lists of equipment Knights, P.F. (2001) “Rethinking Pareto Analysis: Maintenance
downtime priorities. Log scatterplots preserve the ranking scheme of Applications of Logarithmic Scatterplots”, Journal of Quality in
Pareto histograms, but provide additional information content Maintenance Engineering, Vol 7, No.4, pp252-263.
regarding failure frequencies and mean time to restore service. A Moore, R. (2002) “Making Common Sense Common Practice:
methodology is provided for calculating thresholds for classifying Models for Manufacturing Excellence”, Butterworth Heinemann.
failures according to acute or chronic characteristics. The graphs Moubray, J. (1997) “Reliability-Centered Maintenance”, 2nd Ed.,
readily permit the identification of problems affecting system reliability, Industrial Press Inc., New York.

22