Fire Safety Journal 63 (2014) 52–68

Contents lists available at ScienceDirect

Fire Safety Journal

journal homepage: www.elsevier.com/locate/firesaf

Reliability of sprinkler system in Australian high rise office buildings

K.A.M. Moinuddin n, I.R. Thomas
Victoria University, P.O. Box 14428, Melbourne MC, Victoria 8001, Australia

art ic l e i nf o a b s t r a c t

Article history: A comprehensive survey of sprinkler systems in high-rise office buildings was carried out to determine the
Received 13 August 2012 reliability of various components of such systems in Australia. Based on the survey data, a fault tree analysis
Received in revised form (FTA) has been used to estimate the overall reliability of these sprinkler systems. Data from 26 buildings were
16 November 2013
collected and are presented in this paper. In addition, data from overseas surveys has also been considered
Accepted 24 November 2013
Available online 15 December 2013
based on their relevance to the office buildings. The analyses are confined only to wet-pipe systems, as these
constitute the vast majority of automatic sprinkler systems in Australia and New Zealand. To develop the fault
Keywords: trees, the designs found in usual practice are considered, rather than the designs just complying the Australian
Sprinkler codes with the minimum requirements. A range of reliability for the sprinkler systems is estimated based on a
Reliability
number of considerations. Sprinkler zone shut off during tenancy changes and out of specification sprinkler
Survey
head appear to be the main factors that may lead to a sprinkler system failure.
Tenancy change
Failure rate & 2013 Elsevier Ltd. All rights reserved.
Probability

1. Introduction significant distress or permanent harmful effects, provided they were

Water is supplied to automatic fire sprinklers (or sprinkler
heads) through a system of piping, and are arranged so that they
are able to automatically distribute sufficient water directly to a
fire to extinguish it or hold it in check until fire fighters arrive [1].
This is achieved by cooling the fire and wetting surrounding
materials in order to make them harder to ignite. As a conse-
quence of this application of water, there have been cases where
the water interferes with the combustion process sufficiently to
reduce the size of the fire and possibly extinguish it.
The effectiveness of a fire-safety system can be considered as
the product of its efficacy and its reliability and this relationship
was first introduced and discussed in [2]. Later the relationship
was formulated in a rigorous manner in [3] and applied it to the
analysis of smoke alarms. A research program [4] was undertaken
to evaluate the efficacy (the degree to which a particular system
achieves an objective, e.g. control or extinguishment of a fire,
given that it operates and delivers the designed amount of water
to the fire) of a sprinkler system in an office building. It was found
that when small office fire occurs, the fire is suppressed almost
instantly. In all open-plan office fire tests, sprinklers had no
trouble in containing the fire, especially above the desk where it
was adjacent to the fire initiation point. Furthermore, fire did not
spread to adjacent workstations or associated combustibles.
During the time of automatic operation of the sprinklers, occupants
of an open plan office area of similar dimensions did not suffer any
n
Corresponding author. Tel.: þ 61 3 99198042 .
E-mail address: khalid.moinuddin@vu.edu.au (K.A.M. Moinuddin).
not involved in the ignition and were reasonably mobile. The structure
of the building was not damaged during the tests and carried the
required loads without any signs of excessive deflection or other
distress. The structural steel members and the composite floor slab
suffered no measurable permanent deflection and would not have
required any form of major repair before reoccupation if fire sprinklers
had been incorporated in a typical office building.
Having established the efficacy of typical sprinkler systems
experimentally, the purpose of the fault tree analysis herein is to
estimate the reliability of a sprinkler system in a high-rise office
building (60 storeys). The reliability of a sprinkler system is defined
as the likelihood that it operates and delivers the designed amount
of water to the fire. The likelihood that sprinkler operates was
reported as nearly 92% by both Kim [5] from US statistics of high-
rise building fires in 1988 and Rohr and Hall [6] from US statistics
of store and office fires in 1999–2002; however it was reported as
96% by Hall [7] from US statistics of store and office fires in 2003–
2007. It can be observed from [6,7] that between 1999–2002 and
2003–2007 period the likelihood that sprinkler operates, when
sprinklers were present in fire area and fire was large enough to
activate sprinkler, has improved from 92% to 96% for stores and
offices. In this paper, the analyses are confined to the wet-pipe
systems as these constitute the vast majority of automatic sprink-
ler systems in Australia and New Zealand. Wet-pipe systems are
installations in which the sprinkler piping network is permanently
charged with water under pressure and are therefore suitable for
use in buildings in which freezing never occurs [8]. The effective-
ness of the wet-pipe sprinkler system in stores and offices has
been reported as 95% by Hall [7], which has been 96% for office
buildings only.

0379-7112/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.firesaf.2013.11.009
 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68
Fig. 1. Legend of symbols.
2. Description of automatic sprinkler system in a typical high
rise building
This section describes the automatic sprinkler system that
is used in a typical Australian high-rise building. A number of
schematics are presented to assist this description. Fig. 1 shows
legends of symbols for various components of a sprinkler system
used in schematics.
2.1. Water supply
According to Building Code of Australia [9] and the sprinkler
standard AS 2118.1-1999 [10], a building over 25 m of height is
required to have water supply from two separate mains. Strictly,
this requirement is difficult to meet and sometimes a concession is
given. For example mains passing along two parallel streets are
considered as separate mains, though strictly they may not be
entirely independent. Alternatively, an automatic pump supply
from private reservoirs (mainly located in the basement) and two
elevated private supplies (mainly located in the rooftop) is also
widely used.
According to AS 2118.1-1999 [10], an office building requires a
light hazard sprinkler system with fast response sprinkler heads at
about 4.6 m maximum spacing. The design criteria for a light
hazard sprinkler system dictates that the water supply must be
capable of flowing 48 l/min per sprinkler head at 100 kPa for the
hydraulically most disadvantaged group of six sprinklers, for a
minimum duration of 30 min. Given the different geographical
locations of cities around Australia and variations in height above
sea level, there are understandable variations in the supply
pressures available in town main water supplies available to fire
sprinkler systems. For the purpose of this study, it is assumed that
a mean average town mains' pressure of 500 kPa is available.
Given that a study case building of 60 storeys is being considered
here and each storey has an average height of 3.7 m (the floor-to-
floor height of office buildings varies from 3.6 m to 4 m, however it
is typically around 3.7 m), the effective building height is approxi-
mately 222 m. When considering that gravitation/elevation head
loss is 9.8 kPa/m and the operating pressure to operate the six
most hydraulically disadvantaged sprinkler is 200 kPa ( þfriction
losses), town mains pressure of 500 kPa can only supply water up
to the eighth floor of the building. Therefore, town mains' pressure
is required to be boosted by a set of pumps, to supply water to the
sprinkler system for storeys above these lowest eight storeys.
Therefore for sprinklers, water is drawn from town mains and
53
either supplied directly to the pump or is stored in basement
reservoirs/tanks and then is pumped to the riser (vertical pipes).
Water flows through non-return valves, pump isolation valves,
pumps, the main sprinkler valve and then alarm valves to the riser.
The pumps are operated from separate pressure switches.
These pumps are either electric or diesel operated. The levels of
charge in the batteries or/and the continuity of the electricity
supply are usually monitored, and battery/power failure is indi-
cated by a local alarm and at the fire indicator panel (FIP).
AS 2118.1-1999 [10] requires that in the case of a multi-storey
building in excess of 75 m (approximately 21 storeys), the sprink-
ler system needs to be divided into stages so that the pressure on
any sprinkler does not exceed 1 MPa (1000 kPa). Considering that
gravitation/elevation head loss is 9.8 kPa/m and the operating
pressure at the highest level (the most hydraulically disadvan-
taged sprinkler) is 200 kPa, then it is likely that the operating
pressure at the lowest level is 200 þ(75  9.8)þ losses E1000 kPa.
It is also important to mention that a typical riser usually
supplies water to a floor area of up to 9000 m2 [10] (usually seven
or eight floors). Thus seven or eight floors comprises a zone.
Thereby, each stage is divided into three zones; in each zone water
is supplied by a separate riser. These risers are controlled by
separate stop valves. At the interface of the stages, usually a plant
room is located where multiple risers emanate from the main riser
(manifold type; see Fig. 2) to supply water to the respective
sprinkler zones within the stage.
For a 50 to 70 storey building, the system can be divided into
the following stages in different ways as shown in Table 1. As
discussed earlier, water can be supplied from the town main
without a pump up to the lowest eight storeys. Therefore, the
lowest stage usually consists of only eight storeys (see designs
given in [12]) for high-rise buildings. For the purpose of this paper,
Option I for a 60 storey building is considered in detail.
The building floors can be supplied with water in two ways:
 Upfeed system – by pump pressure only.
 Downfeed system – by gravity.
These systems are discussed now.
2.2. Water supply for upfeed system
Given the height of this building, there may be a need to have
pump rooms for stage to stage pumping (plant rooms) or high
pressure supply multi-stage pump along with high pressure riser.
AS 2941-2008 [13] allows the use of both single-stage or multi-
stage pumps. In either case, one electrical and one diesel pump
should be used. Single-stage pumps deliver water at a single
pressure only, multi-stage pumps deliver water at different pres-
sures suitable for different stages of a sprinkler system. In surveyed
buildings, the use of single-stage pumps has not been observed for
the upfeed system, rather multi-stage pumps are used. In the latter
system, two multi-stage pumps can supply water to the whole
building. Risers are connected to various stages of the pumps that
supply water to the different building stages with different pres-
sures (see Fig. 3). Extra care needs to be taken when designing this
type of system to ensure that enough elevation head and friction
loss is allowed for the correctly pressure rated pipework valves and
fittings, to ensure that there is no failure. Water is either directly
drawn from the town main or reservoirs/tanks located in the
basement. In these tanks, water is stored from town mains.
2.3. Water supply for downfeed system
A set of very high pressure single-stage pumps is usually used
for a downfeed system covering the entire building. In this case
54 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68
Fig. 2. Multiple risers supplying water (upward) to a high rise building (the photograph is from [11]).
Table 1
Staging options for a 50 to 70 storeyed building.
50 storey 60 Storey
Option I
Lower stage Up to 8 storeys Up to 21 storeys
Mid stage-I 9–29 Storeys 22–42 Storeys
Mid stage -II – –
High stage 30–50 Storeys 42–60 Storeys
Fig. 3. Water supply arrangement for upfeed system with multi-stage pumps.
single-stage pumps supply water directly to two gravity tanks at
the top of the building. The tanks must be placed sufficiently
above the roof to provide the required pressure for the sprinkler
heads located in the top floor. Alternatively, a small low pressure
pump is used to achieve protection for the first few floors at the
top of the building. To ensure that the pressure on any sprinkler
does not exceed 1 MPa, water pressure needs to be reduced to
supply water to a floor 20 storeys below the gravity tank. In a
building with a multiple stage (each stage contains approximately
20 storeys) sprinkler system, a plant room is built above the floor
at the end of each stage. In this floor, a riser from the upper stage
Typical
riser
Valve
manifold
Subsidiary
valve
FBC
70 Storey
Option II
Up to 8 storeys Up to 8 storeys
9–29 Storeys 9–29 Storeys
30–50 Storeys 30–50 Storeys
51–60 Storeys 51–70 Storeys
feeds the riser of the lower stage with the help of one of the
following options:
(a) a pair of pressure reducing valves to reduce water pressure for
the lower stage;
(b) a pair of cell tanks storing required amount of water, from
which water is supplied for the lower stage.
It is a usual practice in the Australian buildings to install cell
tanks or pressure reducing valves in pairs to improve the reliability
of the system. This requirement is not obvious in the Australian
Standard [10].
For both upfeed and downfeed systems in the plant rooms,
water is distributed to the risers to supply water to the respective
sprinkler zones within each stage.
2.4. Overall system required by AS 2118
Two different systems are shown in Figs. 4 and 5. A summary is
given in Table 2.
Like tridents, tappings are taken from each water main and are
used to transfer water for normal supply, sprinklers and fire hydrants.
For sprinklers, water is drawn and either supplied directly to the pump
or stored in basement reservoirs/tanks and then is pumped to the
riser. Water flows through non-return valves, pump isolation valves,
the main sprinkler valve and then alarm valves to the riser. From the
riser through a piping networks water is supplied to the sprinkler
heads and the system is permanently kept under pressure. Once a
sprinkler head is activated, due to the water release through the orifice
pressure drops. As a result the alarm valve opens, pressure switches
activate the pumps and water flows through the alarm valve. While
water flows through the alarm valve, an alarm sounds to the fire
brigade and to the building. A pump bypass system is usually provided
to supply water under mains pressure to the lowest zone in the
building, as a back-up supply. This system supplies water to only a few
storeys (usually a maximum of eight).
 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68 55

Plantroom

2xGravity Tanks

Zone 3

Stage 3
Zone 2

Zone 1 Pressure
Reducing Valve
(PRV)
Plantroom 2

Zone 3

Some system has

Stage 2
2xcell tanks
instead of PRVs Zone 2

Zone 1

Plantroom

Zone 3 Zone
Isolation
Stage Valves
Lower

Zone 2
Riser

Zone 1 Fire Brigade

Connection Level 2

Pressure Pressure
Switch 1 Switch 2
Pump Bypass

Booster
OIV 2 Connection
Outlet Isolation Valve(OIV) 1 NRV 2 Fire Truck
Non-Return Valve(NRV) 1 Pump 2
Pump 1

2 x Inlet Isolation
Valve (IIV)to Pump 1 2 x IIV
Tank Bypass to Pump 2
(some system
don't have
tanks) Sprinkler
2 x Suction tank Water Main
Water Main (may or may not
be independent)
Domestic
Hydrant

Fig. 4. Details of water supply and sprinkler system for a 60 storey building (downfeed system).

In Fig. 4 a set of very high pressure single stage pumps directly 2.4 MPa maximum pressure whereas a two stage pump can
supplies water to the gravity tank placed at the top of the building. reach 41 bar or 4.1 MPa). In each plant room a set of pressure reducing
However if the pressure requirement is high two stage pumps should valves are installed to regulate the water pressure appropriately for the
be used (single stage pumps can approximately reach 24 bar or sprinklers and piping system in the stage below. Alternatively a pair of
56 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68

Zone 3

Stage 3
Zone 2

Zone 1

Plantroom 2

Zone 3

Stage 2
Zone 2

Zone 1

Plantroom

Zone 3

Lower
Stage
Zone 2
Riser

Zone 1
Level 2
Fire Brigade
Connection

Alarm Valves
Drain Valves Drain Connection
(not shown for
Main other two valves)
Sprinkler
Valves
Pressure
Switch
Booster
Pump Bypass

Connection
Outlet Isolation Valves Multi-stage
Fire Truck Non-Return Valves Pump 2
Multi-stage Pump1

Tank by pass 2 x Inlet Isolation

(some system 2 x IIV
Valve (IIV)to Pump 1
to Pump 2
don't have
tanks)
Sprinkler
Water Main
2 x Suction tank (may or may not be
Water Main independent)
Domestic
Hydrant

Fig. 5. Details of water supply and sprinkler system for a 60 storey building (upfeed system).

cell tanks, storing required amount of water, from which water is water at suitable pressures in different heights of the building. At plant
supplied for the lower stage, are installed. rooms, typical risers emanate from the main riser that supply water to
In Fig. 5, a set of multi-stage pumps supplies water to diffe- separate zones of each stage. The number of typical risers per stage is
rent stages over the height of the building. These pumps can provide considered to be three in this study.
 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68
Table 2
Sprinkler water supply systems.
Figure no Type of system
Fig. 4 Downfeed
Fig. 5 Upfeed
3. Fault tree construction
A number of fault trees have been constructed for sprinkler
systems of a 60 storey building shown in Figs. 4 and 5 covering
various stages. The top event is selected as sprinkler does not deliver
water at the most disadvantageous locations of the building when a
fire becomes big enough to activate it. A succinct description of
underlying principles for the construction of fault trees is given in
[14]. They made a number of assumptions in constructing the fault
trees. Most of these are valid for the analyses of this study and
whichever is disputed will be discussed in the relevant section. For
readers' convenience, main assumptions are given below:
 Any sabotage or simultaneous multiple fires break out in
various locations is not considered.
 Failure rate for each component is considered constant.
 Tenancy changes are evenly distributed over all floors of the
building.
 If all components of the sprinkler system work as specified,
designed amount of water will be delivered. Whether the fire
would be contained or extinguished is considered as efficacy
and is excluded from this analysis.
 Despite a sprinkler system may have several alarm valve
stations, only one alarm station is analyzed here and consid-
ered to be representative.
 The pressure switches for each pump are designed in such a
way that if one of the two switches operate, it will activate
the pump.
 During weekly inspections the fuel level in the fuel tank of the
diesel pump is maintained.
 Opening/closing of the gate valves are monitored and an incorrect
position activates an audible alarm at the FIP/control panel.
 If fire occurs, at least one sprinkler head has to open. For
determining combined failure of sprinkler heads three cases will
be considered: (a) failure of a single head, (b) failure of at least two
heads, (c) as suggested in [14] failure of one head OR its surround-
ing all eight heads. More discussion is given in Section 5.3.3.
 Human actions that may result in a sprinkler system being
inadequate for the hazard protected or create shielding have
not been included. However, such actions can be included in an
integrated system modeling. Nevertheless, in this FTA analysis
some human actions related to component reliability have
been considered which are discussed later in this section.
For the upfeed system, zones of stages 2 and 3 are the most
vulnerable. One of the fault trees for a downfeed system where
sprinkler does not deliver water at a location at zone 3 of stage 2 is
presented in Fig. 6. One of the fault trees for an upfeed system
where sprinkler does not deliver water at a location at zone 3 of
stage 2 is presented in Fig. 7.
CREAM [15] is a human reliability analysis tool which identifies
four major cognitive functions related to human function failure –
observation, interpretation, planning and execution. In this study,
some human function failures are considered directly and some
are considered indirectly which are given below:
a. “Fails to observe monitor alarm activation and taking action by
operator” has been directly considered though could not be
57
Type of pump used Remarks
Very high pressure single stage
Just code compliant and usual practice
Multistage
determined from the survey. However from the literature it is
taken as 3.0  10  2 (see Table 3).
b. Occasionally long repair times for components (e.g. diesel
pumps) have been identified and included in the analysis.
These are primarily related to interpretation type (involving
diagnosis error, decision error and delayed interpretation error)
and execution type human errors.
c. Occasionally long replacement times for components have
been identified and included in the analysis. Although avail-
ability of the parts can be the major source of error, some of
these can be related to planning type human error.
In this study the following likely human errors are considered,
but excluded from the analysis:
a. Various valves, diesel pump and diesel generator overhaul may
lead to execution type human error. However in Australia strict
maintenance regime is in place and only licensed technicians
are allowed to carry out these works. The failure probability is
remote (i.e. likely to be in the range of 0.00001–0.0001 i.e. class
5 of [14]) and therefore ignored.
b. A valve may be closed due to either not opened from the last test/
isolation or valve is inadvertently closed during maintenance. This
is quite unlikely given that after the last test/maintenance the
system needs to be pressurized to the appropriate level after such
actions. Furthermore, in Australia, each building needs to report
to insurance company about the commencement and termination
of isolation period and it is a remote possibility that termination
of isolation reporting will take place without proper checking.
Therefore it is not included in the analysis. It is to be noted that
the system is tested weekly (in most buildings), therefore any
error in maintenance can be detected in a week time.
c. Valve tempering has not been considered as the valves are
chained and locked in Australia.
4. Quantitative evaluation/survey
To calculate the top event frequency or probability, each basic
event or undeveloped event needs to be assigned a probability or
frequency. Frequency represents the number of events expected per
unit time period (e.g., fires per year). On the other hand, probabil-
ities are dimensionless and can be used to describe the likelihood of
occurrence of an event ranging from 0 to 1. Zero means the event
will never occur and 1 means the event is certain to occur.
To facilitate the calculation of the top event frequency or
probability, the frequency/probability of the relevant components
failure (failure rate per unit time or per demand) of the system is
required. If any component's failure frequency (λ) and the interval
between two successive maintenance visits (Δt¼t  t0) is known,
the probability of the failure or failure rate per demand at any time
T within the interval, P(to oT ot), can be found by [17]
Pðt 0 o T o tÞ ¼ F T ðt  t 0 Þ ¼ 1 e  λðt  t 0 Þ ð1Þ
Here, exponential distribution is assumed to start at t0 [18], the
time when the first maintenance visit occurs and the component is
found functional. Hence, F T ðt  t 0 Þ is the probability of unavail-
ability [19] or unreliability [20] which represents the probability of
58 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68

failure at any time T in the interval [t0,t]. Eq. (1) can be derived
from equation 7 of [18] the way its equation 6 is obtained from
equation 5 there. Eq. (1) can also be derived from conditional
probability expressed by equation 20 of [18].
If the repair/replacement time is not short, for some compo-
nents quantification of its downtime is necessary. To keep it
simple, F Downtime , the probability of unavailability due to repair/
replacement, is calculated as per the following equation,
Z T Benchmark
1 TR
F Downtime ¼ uðtÞdt ¼ λ ð2Þ
T Bencmark 0 T Benchmark
|fflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflffl}
λT R
where, TR is the time taken by the owner to repair or replace faulty
components, u(t) is the unavailability function and TBenchmark is the
benchmark time taken as one year as λ is expressed per year.
Eq. (2) is developed based on equation 40 and Table 5-3.4 (third
term of the last equation) of [18].
The information on failure rate can be collected in the following
two ways:
a. Evaluating the published surveys and reports.
b. Conducting a survey on a number of high-rise office buildings
in Australia. One of the major focuses should be to find out how
much time is taken by the owner to repair and replace faulty
components.
If failure frequency (λ) is found in the above processes it will
be converted to failure probability (P) by using maintenance frequency
given in AS1851-2005 [21] and Eq. (1). To obtain total probability of
unavailability/unreliability, figures obtained from Eqs. (1) and (2) are
added. In addition two other unavailability figures are added:
a. for various valves, diesel pump, diesel generator and water
tanks due to periodic overhaul and/or cleanup;
b. for town main (water), grid electricity and zone isolation valve
due to burst, grid failure and tenancy changes/sprinkler fit out
(see Section 5.2), respectively.
4.1. Review of the existing data (Australian and overseas)
A collection of failure data from 17 sources is tabulated in
Table 3. Due to the space constraint all references are not listed
in this paper. However complete references can be found in [16].
The majority of the tabulated data came from three sources:
VTT [22], 140 William St [23] and Rasmussen report [24]. Based
on following arguments the literature data can be used for

Fig. 6. Fault tree of a downfeed sprinkler system.

 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68 59

Fig. 6. (continued)

estimating the reliability of sprinkler system in a high-rise office
building:
 VTT data [22] are mainly based on a survey on the sprinkler
systems of 102 sample buildings (sprinklered and not related to
nuclear industries) which represents some 5% of the number of
sprinklered buildings in Finland. Of these, nearly 20% represents
office buildings. However some data is based on the survey from
nuclear power plant (NPP)s, but it is stated in the report that
the sprinkler technology in NPPs is almost identical with the
60 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68
Fig. 6. (continued)
technique used in other commercial and industrial installation
and only maintenance actions and periodic testing are better
controlled in NPP than elsewhere. It is expected that Australia,
Finland and other West European countries use main sprinkler
system components from the same manufacturers (mainly US
based).
 140 William St data [23] are mainly based on interviews with
sprinkler maintenance personnel and inspection of mainte-
nance log book in Melbourne, Australia. Some of the informa-
tion was also supplied by Fire Brigade.
 The data from Rasmussen report [24] are not specific to
sprinkler system. This data were obtained from both nuclear
and non-nuclear sources in US, but were collected for use in
nuclear hazard asssesment, in particular on the critical loss of
coolant accident. This study was a major exercise involving
some 70 man years of work. The failure data provided in [24]
for most of the components are similar to the other available
sprinkler data sources i.e. have the same order of magnitude in
figures.
4.2. Collection of new data
4.2.1. Collection procedure
A survey was conducted on the sprinkler systems in a number
of Australian high-rise office buildings. Considerable effort was
required to obtain the contact details of the building managers/
owners and to obtain their cooperation to carry out this survey.
Once the building managers/owners were contacted, the purpose
of the study was explained and a questionnaire was sent out, in
order for them to fully understand what was needed. Although the
response was less than ideal, it was possible to get information
from 26 different buildings out of nearly 60 buildings contacted.
The identification details of these buildings are confidential and
only the agglomerated data are presented in this paper.
140 William St study [23] found that the probability of
subsidiary isolation valves that remains shut was found to be the
primary reason to cause the sprinkler system failure. Therefore, in
the questionnaire (a copy can be found in [16]) emphasis was
given to following questions:
 How many floors are attached to a subsidiary isolation valve?
 Number of hours a floor is isolated in a year? Is it done during
the day? Or night?
 Number of tenancy change in a year (averaged over the
building life)?
During the visit to the buildings to collect the failure data, the
building manager was interviewed using the questionnaire and
the following documents were consulted (the ones are made
available):
 Asset registry.
 Schematic diagram of the sprinkler system.
 Records of sprinkler component repair/replacement works.
 Records of informing fire brigade and insurer about sprinkler
isolation of certain floors (system impairment) due to tenancy
change etc.
 Records of periodic testing of sprinkler components.
Every building management, at least, made records of periodic
testing of sprinkler components available.
4.2.2. Limitations of the data
Unfortunately, not all of the above documents were kept for the
entire life of many buildings, due to changes in management or
older records had been archived, therefore some buildings only
had records for the last two to five years. In one of the buildings,
access was provided to the archive room and it was found very
difficult to extract the relevant records from the chaotically archived
documents.
Another limitation of this study is that poorly maintained
buildings were generally not interested in taking part in the
survey. It is also worth mentioning that the age of the surveyed
 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68 61

Sprinkler does not deliver

water to zone 3 of stage 3

OR

No water behind Sprinkler head

sprinkler head faulty

OR

Water supply pipe to No water supply to

sprinkler head damaged floor of fire origin

OR

Zone isolation valve closed Riser leak (from valve to No water supply to
for tenancy change floor of fire origin) zone isolation valve

OR

Failure of main No water supply to

control assembly/riser main control assembly

OR OR

No water flow through Main riser leaked/ Inadeqaute water flow from Inadequate pressure
control assembly damaged storage tank/ town mains from pump

2 AND 4

Inadequate flow from Inadequate flow from

tank/ town main 1 tank/ town main 2

3 3

Fig. 7. Fault tree of an upfeed sprinkler system (2, 3 and 4 are the same as Fig. 6).

building varied between 4 and 36 years and until the year of 2007
data was collected.
4.2.3. Determining failure frequency/probability
To calculate the top event frequency or probability, each basic
event or undeveloped events need to be assigned a probability or
frequency. The frequency of the relevant component failure
frequency λ (failure rate) of the system is computed using
total number of component failures
λ¼
total number of that component  total period of data available ðin yearsÞ
ð3Þ
The number of each component was either directly obtained
from the asset registry or counted from the schematic diagram.
During the survey, building facilities/fire service managers were
also asked the time taken to repair or replace faulty components.
This is essential, as it enables quantification of downtimes using
Eq. (2).
5. Result and discussion
5.1. Failure statistics
Failure probability figure for each component has been computed
from 26 buildings. It is important to see whether the number of
surveyed buildings is sufficient to determine the reliability of all
components of the sprinkler system. Therefore the mean value (μ)
and standard deviation (s) data for each component is calculated as
the survey is progressed. Once the values are plotted against the
number of buildings surveyed, if the value does not significantly
change with more buildings surveyed it can be considered that no
more buildings are required to be included in the study.
In Fig. 8, the mean value of the failure probability data for each
component is presented as the survey is progressed. The data is
presented in log-linear scale. For most of the components after 10
buildings, clear trends had emerged. By 20 buildings all profiles,
except a very few, are quite stable.
 62
Table 3
Failure rate per demand of system components from the literature (References 17–35 of [16]) and the current study.

Components Literature data Current study

P0.05 Mean P0.95 n

Alarm valve 0.03  10  4 [17] 10.7  10  4 18.8  10  4 26.8  10  4 24

0.54, 1, 1.67  10  4 [19]a
1  10  4 [23]
Main stop valve 5.48, 54.8, 548  10  3 [19] 1.36  10  3 2.1  10  3 2.88  10  3 23
0.1  10  3 [23]
Zone isolation valve due to tenancy changes 3.6  10  2 [18]b 1.43  10 2
2.3  10 2
3.19  10 2
21
Ordinary Stop valve 0.02  10  3 [17] 0.36  10  3 0.62  10  3 0.86  10  3 25
5.48, 54.8, 548  10  3 [19],
0.1  10  3 [23]
Non-return valve 8.33  10  4 [17] 6.3  10  4 11.9  10  4 17.5  10  4 25
0.3, 1.0, 3.0  10  4 [19]
0.1  10  4 [23]

K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68

Alarm gong 0.13  10  2 [17] 0.82  10  2 2.9  10  2 4.9  10  2 24
2.68, 3.62, 4.81  10  2 [19]
Sprinkler head 420  10  4 [17] Statistically significant data not available
0.01  10  4 [18,19]
782  10  4 [35]
Town main 1.5  10  4 [18] 0.05  10  4 1.58  10  4 3.12  10  4 26
Gravity tank 2.28  10  4 [18] 2.28  10  4 2.28  10  4 2.28  10  4 10
þ3.64  10  4 [28]
Storage tank 0, 0, 7.8  10  4 [19] 1.37  10  3 7.23  10  3 13.1  10  3 11
2.28  10  4 [18]
þ 3.64  10  4 [28]
Water supply line (per m) 2.6  10  6 [29] 3.9  10  6 11.6  10  6 19.3  10  6 26
Back-up batteries for the diesel pump 8.0  10  4 [18] 80.7  10  4 309  10  4 537  10  4 25
2.19  10  4 [23]
Mains power in building 5.0  10  5 [18] 6.0  10  5 36.6  10  5 66.5  10  5 25
Building power generator 1.08  10  3 1.30  10  3 1.53  10  3 14
Pressure switch 10  10  4 [18] 44.0  10  4 78.9  10  4 114  10  4 26
1.0  10  4 [23]
Diesel pump 1.67  10  3 [18] 55.4  10  3 92.0  10  3 129  10  3 26
0.73, 1.25, 1.91  10  3
[19]
3  10  3[23]
0.0, 1.24, 4.82  10  3 [33]
Electric pump 0.3  10  3 [18] 9.0  10  3 17.3  10  3 25.5  10  3 26
0.21, 0.52, 1.08  10  3 [19]
3 3 3
Jacking pump 4.0  10 8.79  10 12.5  10 15
Direct brigade alarm 1  10  4 [18] 54  10  4 117  10  4 181  10  4 21
Pressure reducing valve 0.05  10  3 [23] 0.3  10  3 1.68  10  3 3.06  10  3 8
10  10  3 [25,27]
4.37  10  3 [26]
FIP 6.43  10  2 [21] 7.1  10  2 10.3  10  2 13.6  10  2 24
0.01  10  2 [34]
Back-up batteries/UPS for FIP 0.1  10  2 [18] 0.48  10 2
1.24  10 2
2.01  10 2
24
3.13  10  2 [21]
Monitor alarm/sensor 1.82  10  2 [30] 0.16  10 2
3.59  10 2
7.02  10 2
23
7.76  10  2 [21]
Fail to observe monitor alarm activation 3.0  10  2 [24] Could not be determined
Underground pipe from town mains to tank corroded 1.0  10  4 [32] Could not be determined

a
When three figures are presented they represent P0.05, μ and P0.95.
b
Based on 15 tenancy changes in a 37 storey building in a year. Eight floors are connected by one subsidiary valve. During a tenancy change, this valve is shut for 4.05 days.
 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68 63

1.00E+00
Back-up battery
Back-up battery Zone isolation valve due to
/UPS f or FIP
f or diesel pump tenancy change
FIP Electric pump
Alarm motor

1.00E-01
Monitor alarm Diesel pump

DBA
1.00E-02
Mean values

Pressure
switch
Main stop valve

Storage
Generator tank
PRV Alarm v alv e
1.00E-03
Ordinary isolation valve Non-return
valve

Gravity tank

1.00E-04
Town main
Main power

Water supply line (per m)

1.00E-05
0 5 10 15 20 25 30
Number of building

Fig. 8. Mean failure probability data for each component against the number of building surveyed.

Similarly in Fig. 9, the standard deviation of the failure prob-
ability data for each component is presented as the survey is
progressed in log-linear scale. Again by 20 buildings the standard
deviation profiles are quite stable. Figs. 8 and 9 demonstrate that
the number of surveyed buildings is sufficient to determine the
reliability of all components of the sprinkler system quite
accurately.
To determine the 90% confidence interval (from 5% to 95%
confidence limits) a rigorous analysis is required to determine the
distribution pattern of each component data set. However, a
simplified approach is presented in this paper where it is assumed
that all data sets are likely to follow the normal distribution.
The lower and upper confidence limits (P0.05 and P0.95) can be
calculated as
1:65s 1:65s
P 0:05 ¼ μ  pffiffiffi and P 0:95 ¼ μ þ pffiffiffi
n n
In Eq. (4), n represents the number of buildings from which the
pertinent data is collected.
The results of mean and P0.95 failure probability of each
component are plotted in Fig. 10 alongside tabulating values,
including P0.05, in Table 3. The figures for the same components
from literature are also presented. In Fig. 10, the vertical axis is in
logarithmic scale and the numbers are shown in inverse order i.e.
the longer the column the lower its failure probability. In Fig. 10
besides minimum and maximum values from literature (indicated
as Lit-min and Lit-max, respectively), the most reliable literature
data is also presented as Lit-best. The reasons for which these
are chosen as the most reliable literature data can be found in
[16]. Out of 21 components, 14 components data are found to be
within the same order of magnitude of the data found in the
literature. The remaining seven components' failure probabilities
are found to be higher than in the literature (indicated by ovals in
Fig. 10).
Using sensitivity analysis of the fault tree it is apparent that the
top event is very sensitive to the failure probability of the
following components:
Sprinkler Non-return Pressure Pressure
head valve switch reducing valve
Alarm Zone isolation Gravity
valve valve tanks
However, both survey and literature data reveal that the failure
probability of these components is very small except for the
sprinkler head and zone isolation valve (sprinkler downtime due
to work during tenancy change). These two most important issues
will be now discussed in detail. However, it is assumed in
regulatory and building management circles that if the strict
maintenance regulation and practice are not continued, the failure
probability of other components is likely to rise significantly.
ð4Þ
Unfortunately, “fail to observe monitor alarm activation and take
action by operator”, “wiring burn out” and “underground pipe from
town mains to tank corroded” could not be determined in the survey;
therefore the literature data from Table 3 are used for fault tree
analyses.
5.2. Sprinkler downtime due to work during tenancy change
In most buildings the only means of interrupting the water
flow to a floor during a tenancy change is to turn off the sprinkler
valve on the riser. Fig. 2 shows that eight floors are connected by
one zone isolation valve. Therefore, all connected floors are not
supplied with sprinkler water for a number of days. However, it
has been found from the survey that during the tenancy changes,
sprinkler valves are usually shut during the daytime only. In some
buildings, sprinkler valve isolation was restricted during weekends
only. The buildings have the policies of restriction on sprinkler
isolation time, which varies from 8 to 10 h a day.
The probability that sprinkler valve is turned off, P(V), can be
calculated as

no of changes per yr  no of days per change  no of isolation hrs per day  no of fls connected to a valve
PðVÞ ¼ ð5Þ
total no of fls in the building  no of days per yr  no of hrs per day
64 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68
1.00E+00
1.00E-01
Standard Deviation
1.00E-02
1.00E-03
1.00E-04
1.00E-05
0 5
Fig. 9. Standard deviation of failure probability data for each component against the number of building surveyed.
It has also been found that the building facility/fire service
managers are aware of the dangers during a tenancy change
period. The danger during this period is exacerbated by the fact
that hot-working is also often carried out at the same time in the
unprotected area. In many buildings for such periods, physical
patrolling is arranged. Applying Eq. (5) to surveyed building data,
mean P(V) is found to be 0.023 which is significantly lower than
the value (0.036 for a particular building) found in [23] which may
be attributed to not allowing sprinkler isolation during night time.
140 William St study [23] found that sprinkler valve shut off
during tenancy changes was the main factor that may lead to a
sprinkler system failure. It was also shown by Bugbee [25] from US
fire data between 1925 and 1969 that sprinkler shut off accounts for
35.4% of the sprinkler system failure (the next highest proportion of
the cause was just 13.5%).
5.3. Sprinkler head reliability
5.3.1. Previous study
Nash and Young [26] present some data of sprinkler head
testing from 1960 to 1970 in the UK, where 1967 new as well as
175 old heads were tested. For new heads, a probability of 0.009
was observed for complete blockage, 0.023 for partial blockage,
thus a total failure probability of 0.032. For old heads, the respective
figures were 0.017, 0.034 and 0.051.
Watanbe [27] developed a fault tree of sprinkler system based
on the 100 Japanese fire reports by well-trained investigators. He
found that the failure probability of sprinkler heads is 0.0028.
It should be noted that this is not the failure probability for a single
sprinkler head, rather the combined failure of one or more heads
located in the vicinity of a fire.
Zhuiykov and Dowling [28] presented the test data of 6350
sprinkler heads conducted in Australia in the time interval of
1992–2005. Of those sprinkler heads 500 failed (outside specifica-
tion) resulting in a failure probability of 0.079 which is within
the same order of magnitude as obtained from above [26]. The
Australian Standard AS 1851.3 [29] required testing of representa-
tive samples of sprinklers in a system after 24 years in service, and
10 15 20 25 30
No of building
then at six-year intervals thereafter. It appears that most of these
tested heads are at least 24 years old. Furthermore, these tested
heads are not only from the office buildings rather from all types
of occupancies. Many of these had apparently been installed and
maintained in abnormal environments, such as corrosion, abnor-
mally high temperatures or mechanical abuse compared to office
environments.
5.3.2. Current study
In the current study, sprinkler head testing data was found only
from two buildings. The result is tabulated in Table 4.
A critical failure probability of 0.3 for Building#1 heads is
extremely high compare to the other study. However, the critically
failed sprinkler heads were tested until the temperature of 90 oC
and the tests discontinued if the heads were not activated. There is
a likelihood that those heads could be activated at a temperature
higher than 90 oC and in the event of a fire those late activated
heads (though not ideal) would play some role in controlling the
fire. During one of the sprinklered office fire tests of Proe and
Bennetts [30] due to inadvertent closing of the supply, water was
delivered through a single sprinkler head 2 min after sprinkler
activation. Although the fire got bigger, that single head had no
trouble in controlling the fire.
5.3.3. Discussion
It is clear that a popular perception of high sprinkler reliability
i.e. a failure probability of 1E-06 (which is often used in the risk-
based fire engineering analysis including [23]) is not supported.
The popular perception may be correct at the time of manufactur-
ing, but clearly not after few years of installation and use. Based on
the above discussion the failure probability of a single head is
taken as 0.079 from [28] as the data from the study is Australian
and conservative in a sense that many of the tested heads were
subject to extreme conditions over several years (up to 24 years)
and got degraded compared to the ones exposed to the office
environment.
 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68 65

1.00E-07
Lit-min
Lit-best
Current Survey -mean value
1.00E-06 95% conf idence
Probability of failure Lit-max

1.00E-05

1.00E-04

1.00E-03

1.00E-02

1.00E-01

1.00E+00
Alarm valve

Main stop valve

town main

Gravity tank

Storage tank
Alarm motor
Zone isolation valve

Non-return valve

Water supply line

Pressure reducing
Ordinary isolation
due to tenancy

(per m)
valve
change

valve

1.00E-07
Lit-min
Lit-best
1.00E-06 Current Survey -mean value
95% conf idence
Lit-max
Probability of failure

1.00E-05

1.00E-04

1.00E-03

1.00E-02

1.00E-01

1.00E+00
diesel pump

electric pump

pressure switch

FIP

Direct brigade alarm

Building power

mains power in

back-up batteries for

back-up batteries/UPS

Monitor alarm/ sensor

the diesel pump
generator

building

for FIP panel

Fig. 10. Comparison of results between literature and current data. Ovals indicate failure probabilities in the current data are higher than those found in the literature.

However, failure of a single sprinkler head does not necessarily failure of sprinkler heads, failure of 1 head OR its surrounding all
reflect failure of the sprinkler system. If the head nearest to the fire 8 heads is needed. Based on this concept, failure probability of
fails, the adjacent heads are likely to be activated and can suppress sprinkler heads is calculated as 0.038 (considering failure probability
the fire. On the other hand, adjacent sprinkler heads are likely to of a single head as 0.079). If surrounding heads are denoted as S1, S2,
be from the same batch. Therefore, there is a strong possibility that S3, S4, S5, S6, S7, S8 and the central head as C1 (each of them having
if one head fails the adjacent heads will fail too. failure probability of 0.079) it can be calculated as
For the purpose of this paper, it is considered that to represent the (((((((S1 AND S2) AND S3) AND S4) AND S5) AND S6) AND S7)
combined failure of sprinkler heads located in the vicinity of a fire at AND S8) OR C1 ¼0.038
least two sprinkler heads need to be failed. Therefore the combined Along with these two sets of cases, cases are also considered
failure of two sprinkler heads is considered as 0.079  0.079¼0.0062. where a single head failure may represent the combined failure of
Hauptamanns, Marx and Grunbeck [14] suggested that for combined sprinkler heads located in the vicinity of a fire.
66 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68

5.4. Sprinkler system reliability In all cases mean value of P(V) is taken as 0.023 as found in the
current survey. Similarly, the failure probability of a single head is
It is important to mention that some buildings draw sprinkler taken as 0.079 from [28].
water from only one town main and some of the buildings (usually It can be observed that based on current survey data, the failure
o20 storey) do not have power generator to back up the street probability of the sprinkler system in Australian high-rise building
power supply. This FTA is conducted for a 60 storey office building lies in between 2.1% and 13.4% which is higher than the commonly
which draws sprinkler water from two independent town mains considered values of 5% (non-flashover fire) and 1% (flashover fire)
and has power generator to back up street power supply. [31]. US statistical data [7] of wet pipe sprinkler does not operate
From FTA the failure probabilities of sprinkler system for in office buildings when fire was large enough to activate sprinkler
different systems, for the most disadvantageous zones, are shown (4%) falls within the range found in this study. It is to be noted that
in Table 5. In all cases three situations are considered: whether 'a fire was large enough to activate sprinkler' as well as
'sprinklers did not operate when it should have' are subjective as
 2 head: for combined failure of sprinkler heads, at least 2 heads they depend on the judgement of the fire fighter entering the data.
need to be failed. Ref. [7] also shows that in 2.8% (¼70% of 4%) of office building
 1 head: for combined failure of sprinkler heads, 1 head failure is fires, sprinklers failed due to system shut off and this agrees well
sufficient. with our P(V) data of 2.3%.
 [14] method: for combined failure of sprinkler heads, failure of Records for Class 5 (office) building fires have been collected
1 head OR its surrounding all 8 heads is sufficient. from Metropolitan Fire Brigade (MFB), Melbourne for 2762 fires
that occurred from January 1998 to December 2012. Of these fires
656 were recorded as having sprinklers present and 1717 as having
Table 4 no sprinklers, with the remainder having the presence of sprinklers
Failure rate per demand for sprinkler heads from the current study. not recorded or uncertain. Despite similar subjectivity mentioned
in relation to US data above, taking the data at face value there
Building No of No of head outside No of head Critical failure
no head specification critically faileda probability are three fires out of a total of 67 fires (with sprinklers present where
tested the fire was not classified as too small to require sprinkler opera-
tion) where the sprinklers are classified as not having operated
1 20 12 6 0.3
when they should have representing 4.5% (very similar to US
2 100 2 0 0.0
values). There were two other fires where sprinklers were present
a
Did not go off even at 90 1C temperature while the nominal operating but their performance is not classified in one of three available
temperature was 68 1C. categories (extinguished, prevented spread or failed to prevent spread).

Table 5
Failure probability of sprinkler system.

Type of system Based on mean value Based on P0.95 Based on P0.05

2 Head 1 Head [14] Method 2 Head 1 Head [14] Method 2 Head 1 Head [14] Method

Based on current survey Upfeed (with no basement tank) 4.7% 11.7% 7.7% 6.6% 13.4% 9.5% 2.9% 10.0% 6.0%
Downfeed 3.0% 10.1% 6.1% 4.0% 11.0% 7.1% 2.1% 9.3% 5.2%

Based on literature data Upfeed (with no basement tank) 3.8% 10.9% 6.9% – – – – –
Downfeed 3.2% 10.3% 6.3% – – – – –

0.1

0.09
Failure probability of sprinkler system

0.08

0.07 Failure probability-downfeed

Failure probability-upfeed
0.06

0.05

0.04

0.03

0.02

0.01

0
0 5 10 15 20 25 30
Number of buildings
Fig. 11. Failure probability data for sprinkler systems against the number of building surveyed.
 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68
Table 6
Failure probability of sprinkler system when valves are installed for each floor.
Type of system Based on mean value
2 Head 1 Head [14] Method
Upfeed (with no basement tank) 3.0% 10.1% 6.1%
Downfeed 1.3% 8.5% 4.5%
If it is assumed that in these fires the sprinklers did not perform as
intended (either did not deliver required amount of water or did not
simply achieve the objective of control or extinguishment) then in
five (or 7.5%) of the fires the sprinklers failed to operate effectively.
There is one other fire where the classification was that the
sprinklers failed to prevent spread. If this fire is included in the
sprinklers not effective category then in six fires (or 9% of fires)
the sprinkler system did not operate effectively.
Once the data for five buildings were processed the failure
probability of the sprinkler system was computed. The failure
probability is calculated as more buildings are surveyed. In Fig. 11
the failure probability figure, based on the mean value of each
component and consideration of at least two sprinkler heads need
to be failed, is presented as the survey is progressed. By 20
buildings both profiles are found to be quite stable reflecting that
statistically sufficient number of buildings have been surveyed.
6. Recommendation for improvements
It is found in this study that sprinkler system failure can be
mainly attributed to sprinkler downtime due to work during
tenancy change and the sprinkler head failure. Sprinkler down-
time can be reduced by the following means:
 Installing sprinkler isolation valve for each floor: If all the
surveyed buildings were installed with isolation valve for each
floor, the failure probability would be as per Table 6. It can be
observed that the failure probability of the same sprinkler
system can be reduced between 0.01 and 0.025 which amounts
to an improvement in the range of 10–63% by this measure.
 Use of flexible sprinkler pipes connecting the riser and the
sprinkler head: This will result in relocating sprinkler heads
without isolating the sprinkler system for minor rearrange-
ment of the office. However, the reliability of a flexible pipe
compared to a rigid pipe is not known. Therefore, it is not yet
clear whether such a measure will reduce the overall failure
probability of the system.
The initial testing interval of representative samples of fast-
response sprinkler heads (required for office buildings) is now
reduced to 12 years, then at 12 years and then at six-year intervals
thereafter as per Australian Standard AS 1851-2005 [32]. This will
likely to reduce the sprinkler head failure probability.
7. Conclusions
As the use of a physical survey is an effective means of gathering
detailed information on the reliability of sprinkler system, this
approach was adopted in this study. Although it was challenging to
involve the building owners in the study, it was possible to come up
with significant data by consulting relevant documents. These
documents allowed calculation of the failure probability of various
components of sprinkler system. In the process, time taken to repair
or replace faulty components has been incorporated.
67
Based on P0.95 Based on P0.05
2 Head 1 Head [14] Method 2 Head 1 Head [14] Method
4.2% 11.2% 7.2% 1.9% 9.0% 5.0%
1.5% 8.7% 4.7% 1.1% 8.3% 4.2%
The results show that out of 21 components, 14 components of
the data are found to be within the same order of magnitude of the
data found in the literature. The remaining seven components' failure
probabilities are found to be much higher than the literature data.
The reliability of the sprinkler system is estimated based on
number of considerations. It can be observed that the failure
probability of the sprinkler system in a 60 storey Australian office
building lies in the range 2.1–13.4% which is higher than the
commonly considered values in Australia.
Sprinkler zone shut off during tenancy changes and sprinkler
heads appear to be the main factors that may lead to a sprinkler
system failure. In spite of the improvement in reliability with
respect to the sprinkler zone shut off during tenancy changes, it
remains one of the main causes leading to a sprinkler system
failure. In regard to this, by installing zone isolation valve for each
floor, the failure probability of a sprinkler system can be reduced
between 10% and 63% of its original value.
References
[1] International Fire Service Training Association, Essentials of Fire Fighting, 1998.
[2] I.R. Thomas, Effectiveness of fire safety components and systems, J. Fire
Protect. Eng. 12 (2) (2002) 63–78.
[3] Y. He, D. Nelson, A comparative study of effectiveness of smoke alarms in two
types of buildings, J. Fire Sci. 26 (2008) 415–434.
[4] I.D. Bennetts, K.A.M. Moinuddin, D. Proe, I.R. Thomas, Sprinklered office fire
tests, Fire Mater. 32 (2008) 159–198.
[5] W.K. Kim, Exterior Fire Propagation in a High Rise Building (Masters Thesis),
Worcester Polytechnic Institute, USA, 1990.
[6] J.R. Hall Jr., U.S. Experience with Sprinklers and Other Automatic Fire Extin-
guishing Equipment NFPA Fire Analysis and Research Division, Quincy, MA,
USA, August 2005.
[7] J.R. Hall Jr., U.S. Experience with Sprinklers and Other Automatic Fire Extin-
guishing Equipment NFPA Fire Analysis and Research Division, Quincy, MA,
USA, February 2010.
[8] Standard Australia, Sprinklers Simplified, 0-7337-3037-X, 2000.
[9] Australian Building Codes Board, Building Code of Australia, 2013.
[10] Australian Standard™, Automatic Fire Sprinkler Systems, AS2118.1-1999, 1999.
[11] R. Piccolo, Back to Basics: Sprinkler Design for Multi-Tenant Buildings. PM
Engineer, May 2004, http://www.pmengineer.com/Articles/Feature_Article/
2004/05/05/Back-to-Basics-Sprinkler-Design-for-Multi-Tenant-Buildings,
downloaded on 14 March, 2012.
[12] M. Roy, Zoning of High-Rise Fire Protection Systems. PM Engineer, March
2005.
[13] Australian Standard™, Fixed fire protection installations – Pumpset systems,
AS2941-2008, 2008.
[14] U. Hauptamanns, M. Marx, S. Grunbeck, Availability analysis for a fixed wet
sprinkler system, Fire Safety J. 43 (2008) 468–476.
[15] E. Hollnagel, Cognitive Reliability and Error Analysis Method: CREAM, Elsevier
Science, Inc, Oxford, UK, ISBN 0-08042-848-7, 1998.
[16] K.A.M. Moinuddin, I.D, Bennetts, I.R. Thomas, S. Chea, An Assessment of
Sprinkler System Reliability In High Rise Commercial Buildings, Victoria
University Internal Report, 2007.
[17] V. Beck, Fire code Reform Centre Project 4, Fire Safety Design Solutions Part 1:
Core Model and Residential Buildings Project Specifications, Victoria Univer-
sity Internal Report 1995.
[18] F. Joglar, Reliability, availability, and maintainability, in: P.J. Dinenno (Ed.),
SFPE Handbook of Fire Protection Engineering, 4th ed.,Society of Fire Protec-
tion Engineers, Boston, Massachusetts, USA, 2008.
[19] J. Braband, R. von Hovel, H. Schabe, Probability of failure on demand – the why
and the how, Comput. Safety, Reliab., Secur.: Lecture Notes Comput. Sci. Vol.
5775 (2009) 46–54, http://dx.doi.org/10.1007/978-3-642-04468-7_5.
[20] A. Shetiya, Basic Reliability Mathematics. Lecture Notes of New Mexico
Institute of Mining and Technology. http://infohost.nmt.edu/  olegm/484/
Chap3.pdf, downloaded on 12 September, 2013.
68 K.A.M. Moinuddin, I.R. Thomas / Fire Safety Journal 63 (2014) 52–68
[21] Australian Standard™, Maintenance of fire protection system and equipment,
AS1851-2005, 2005.
[22] V. Ronty, O. Keski-Rahkonen, J. Hassinen Reliability of Sprinkler Systems: [28]
Exploration and Analysis of Data from Nuclear and Non-nuclear installations.
VTT Building Technology, 951-38-6566-5 (http://www.vtt.fi/inf/pdf/), 2004.
[23] I.R. Thomas, I.D. Bennetts, S.L. Poon, J.A. Sims, The Effect of Fire in the Building
at 140 William Street, BHP Research Report BHPR/ENG/R/ 92/044/SG2C, [29]
February 1992.
[24] N. Rasmussen, Reactor Safety Study Report WASH-1400 (NUREG-75/014), [30]
Atomic Energy Commission, Washington DC, 1975.
[25] P. Bugbee, Principles of Fire Protection, 0877653453, 1978. [31]
[26] P. Nash, R.A. Young, Automatic Sprinkler Systems for Fire Protection, 2nd ed.,
Paramount Publishing, Borehamwood, 1991.
[32]
[27] A. Watanabe, Effectiveness of active fire protection systems, in: Proceedings
of the 4th Joint Panel Meeting of the U.S./Japan Government Cooperative
Program on Natural Resources (UJNR). Fire Research and Safety. Tokyo, Japan,
pp. 22–37, February 1979.
S. Zhuiykov, V. Dowling, Maintenance testing of sprinkler heads: qualitative
analysis causes of failures, in: Proceedings of the Eighth International
Symposium on Fire Safety Science, International Association for Fire Safety
Science, pp. 811–822, 2005.
Australian Standard™ Maintenance of Fire Protection Equipment – Automatic
Fire Sprinkler Systems AS 1851.3-1997, 1997.
D.J. Proe, I.D. Bennetts, Six Fire Tests in a Large Sprinkler-Protected Open-Plan
Office, BHP Research Report No: BHPR/SM/R/92/010, May 1997.
Fire Code Reform Centre, Fire Engineering Guidelines, 1st Ed., Sydney,
Australia, March 1996.
Australian Standard™ Maintenance of Fire Protection Systems and Equipment,
AS 1851-2005, 2005.