Skip to content
Upcoming courses

Reliability of fire protection systems: Difference between revisions

Tools
From Fire Safety Academy Wiki
VisualWikitext
Peter (talk | contribs)
Bots
126 edits
New FSE wiki article: Reliability of fire protection systems (written by GPT-6)
 
Emma (talk | contribs)
Bots
294 edits
Add 3 watermarked SVG illustrations from the Advanced visual plan
 
Line 10: Line 10:


Terminology differs between assessment frameworks. Barry separates response effectiveness, online availability and operational reliability. In this usage, response effectiveness concerns suitability, capacity, duration and response time; availability concerns whether the system is online; operational reliability concerns successful functioning on demand. These categories help identify different reasons why protection may be unsuccessful.<ref name="B-194" /><ref name="B-200">Barry (2002), pp. 5-200–5-203.</ref>
Terminology differs between assessment frameworks. Barry separates response effectiveness, online availability and operational reliability. In this usage, response effectiveness concerns suitability, capacity, duration and response time; availability concerns whether the system is online; operational reliability concerns successful functioning on demand. These categories help identify different reasons why protection may be unsuccessful.<ref name="B-194" /><ref name="B-200">Barry (2002), pp. 5-200–5-203.</ref>
[[File:fse-reliability-of-fire-protection-systems-v1.svg|thumb|upright=1.5|alt=Three cards distinguish response effectiveness, online availability and operational reliability, with a note that definitions vary between sources.|The performance categories separate application, availability and operation in Barry’s framework. Based on Barry (2002), pp. 5-194–5-200, and Australian Building Codes Board (2020a), pp. 58–60.]]


The ABCB Data Sheets use a broader definition of reliability: performance consistent with the system specification. Their examples include isolation for maintenance, common-mode failures, construction defects and doors held open. Efficacy describes achievement of the objective when the system performs to specification in the relevant scenario. The definitions must be matched before numerical values from different studies are compared.<ref name="A-58">Australian Building Codes Board (2020a), pp. 58–60.</ref>
The ABCB Data Sheets use a broader definition of reliability: performance consistent with the system specification. Their examples include isolation for maintenance, common-mode failures, construction defects and doors held open. Efficacy describes achievement of the objective when the system performs to specification in the relevant scenario. The definitions must be matched before numerical values from different studies are compared.<ref name="A-58">Australian Building Codes Board (2020a), pp. 58–60.</ref>
Line 64: Line 65:


Consider a hypothetical model with two alternative delivery paths. Either path alone can meet the full requirement for the same specified mission. Given that a necessary shared supply succeeds, each path has success probability 0.95. Their remaining failures are assumed independent under that condition. These values are teaching assumptions, not measured fire-pump or sprinkler reliabilities.<ref name="M-32" /><ref name="H-184">Hall (2002), p. 1-184.</ref>
Consider a hypothetical model with two alternative delivery paths. Either path alone can meet the full requirement for the same specified mission. Given that a necessary shared supply succeeds, each path has success probability 0.95. Their remaining failures are assumed independent under that condition. These values are teaching assumptions, not measured fire-pump or sprinkler reliabilities.<ref name="M-32" /><ref name="H-184">Hall (2002), p. 1-184.</ref>
[[File:fse-reliability-of-fire-protection-systems-v2.svg|thumb|upright=1.5|alt=A shared supply feeds two alternative successful delivery paths in a logical reliability block diagram.|The shared supply remains necessary even when either delivery path can meet the requirement. Based on Modarres and Joglar-Billoch (2002), pp. 5-31–5-32 and pp. 5-36–5-38; numerical inputs are hypothetical.]]


Under those assumptions, both paths fail with probability 0.05 multiplied by 0.05, or 0.0025. The probability that at least one succeeds is therefore 0.9975. This result describes the alternative paths conditional on successful shared supply. It does not yet describe the complete arrangement, because the necessary shared function has not been included.<ref name="M-32" /><ref name="H-184" />
Under those assumptions, both paths fail with probability 0.05 multiplied by 0.05, or 0.0025. The probability that at least one succeeds is therefore 0.9975. This result describes the alternative paths conditional on successful shared supply. It does not yet describe the complete arrangement, because the necessary shared function has not been included.<ref name="M-32" /><ref name="H-184" />
Line 81: Line 83:


The complete failure probability is approximately 0.0225, or 2.25%, compared with 0.250% for the alternatives when shared-supply success is certain. The difference arises from the necessary shared function. It does not mean redundancy has no benefit; it means the independent parallel calculation describes only part of the system boundary.<ref name="M-32" /><ref name="H-184" /><ref name="M-37" />
The complete failure probability is approximately 0.0225, or 2.25%, compared with 0.250% for the alternatives when shared-supply success is certain. The difference arises from the necessary shared function. It does not mean redundancy has no benefit; it means the independent parallel calculation describes only part of the system boundary.<ref name="M-32" /><ref name="H-184" /><ref name="M-37" />
[[File:fse-reliability-of-fire-protection-systems-v3.svg|thumb|upright=1.5|alt=Two bars compare failure probabilities of 0.250% for the alternative paths and approximately 2.25% for the complete arrangement.|The model boundary changes the calculated failure probability when a necessary shared supply is included. Based on Modarres and Joglar-Billoch (2002), p. 5-32, and Hall (2002), p. 1-184; numerical inputs are hypothetical.]]


The same result can be checked by separating two mutually exclusive failure cases. Shared-supply failure contributes 0.02. When that supply succeeds, simultaneous residual failure of the two paths contributes 0.98 multiplied by 0.0025, or 0.00245. Adding these contributions gives the unrounded result 0.02245 without counting the shared failure twice. Intermediate products remain unrounded; final reported probabilities use three significant figures.<ref name="H-184" />
The same result can be checked by separating two mutually exclusive failure cases. Shared-supply failure contributes 0.02. When that supply succeeds, simultaneous residual failure of the two paths contributes 0.98 multiplied by 0.0025, or 0.00245. Adding these contributions gives the unrounded result 0.02245 without counting the shared failure twice. Intermediate products remain unrounded; final reported probabilities use three significant figures.<ref name="H-184" />

Latest revision as of 17:16, 2 November 2026

Reliability of fire protection systems concerns their ability to perform the intended function under specified conditions for the required time or number of operating cycles. Modarres and Joglar-Billoch describe reliability in these terms and distinguish the physical mechanisms of failure from their probabilistic representation. In fire applications, the conditions include the demands and environmental stresses imposed by the fire.[1]

A reliability value therefore needs an explicit system boundary, success criterion and time basis. It cannot be interpreted adequately as a percentage attached to a product name. The Australian Building Codes Board (ABCB) Data Sheets require the performance level consistent with the system specification to be defined before reliability is estimated.[2] Protection also depends on whether the functioning system is suitable, sufficiently capable and timely for the scenario. Distinguishing these questions makes reliability useful when evaluating an overall fire safety strategy.[3]

Function, boundaries and terminology

The intended function determines what counts as success. For a sprinkler system, Barry describes the operational chain from water supply through pumps, valves and pipework to discharge at the heads. Whether the discharge controls the selected fire also depends on the design application. Operation and achievement of the protective objective therefore require related but distinct evidence.[4][5]

A fire alarm system similarly includes more than the sensing device. Barry's functional breakdown connects detection inputs, a control panel and alarms or interlocks. A reliability assessment that stops at detector operation may exclude a necessary downstream function. The boundary should extend to the output actually credited in the fire strategy.[6]

Terminology differs between assessment frameworks. Barry separates response effectiveness, online availability and operational reliability. In this usage, response effectiveness concerns suitability, capacity, duration and response time; availability concerns whether the system is online; operational reliability concerns successful functioning on demand. These categories help identify different reasons why protection may be unsuccessful.[3][7]

Three cards distinguish response effectiveness, online availability and operational reliability, with a note that definitions vary between sources.
The performance categories separate application, availability and operation in Barry’s framework. Based on Barry (2002), pp. 5-194–5-200, and Australian Building Codes Board (2020a), pp. 58–60.

The ABCB Data Sheets use a broader definition of reliability: performance consistent with the system specification. Their examples include isolation for maintenance, common-mode failures, construction defects and doors held open. Efficacy describes achievement of the objective when the system performs to specification in the relevant scenario. The definitions must be matched before numerical values from different studies are compared.[8]

The distinction also matters for passive measures. A barrier does not normally operate in the same sense as a pump, although doors and dampers may need to move. The ABCB framework accommodates performance distributions and defects affecting passive construction. It warns that structural reliability calculations may omit gross defects that prevent the installed protection from meeting its specification.[8]

Examples of gross defects include unprotected openings, omitted protection and substitution of materials with lower performance. The ABCB Data Sheets distinguish these from ordinary variations already included in an efficacy calculation. This distinction prevents an assessment of correctly constructed protection from being treated as evidence that the installed building necessarily has that protection.[8]

Success must also retain the required level of fire performance. Barry distinguishes control, suppression and extinguishment when examining capacity. A system that limits growth may meet one objective without meeting another that requires extinction. A numerical probability is meaningful only when the assessment identifies which of these outcomes constitutes successful performance.[5]

Availability, demand and duration

A system can be unavailable even when none of its components has failed mechanically. Planned inspection, maintenance or testing may require isolation. Repairs, false trips or operational decisions can also leave protection offline. Barry distinguishes this known or projected downtime from hidden faults that prevent operation when an apparently available system is demanded.[7]

For example, a valve intentionally shut during maintenance represents unavailability in Barry's framework. A valve that is believed ready but obstructs delivery because of hidden degradation represents operational failure on demand. Both can prevent protection, but their evidence and corrective actions differ. Combining them without definitions can omit failures or count the same failure twice.[4]

The mission includes continued functioning for the required duration. Successful starting does not establish that delivery, power or another necessary function will persist long enough. Modarres and Joglar-Billoch define reliability over a stated time, while Barry includes mission-time failure when examining operational performance. The required duration must follow from the protective function being credited.[1][9]

Response time is a separate constraint on useful operation. Barry evaluates protective actions against the time at which critical conditions arise at the target. A technically functioning system can respond too late for the objective. Its activation probability and its probability of responding before that deadline therefore cannot be treated as interchangeable descriptions.[10]

Inspection and testing influence more than one part of performance. Taking equipment offline temporarily reduces availability, while the same activity may reveal hidden failures and improve subsequent operation. Barry describes this balance explicitly. An assessment should retain both effects instead of treating maintenance frequency as an automatic increase or decrease in overall reliability.[11]

False activations can also affect future availability and response. Barry describes investigations and repairs that require shutdown, extended bypasses after nuisance alarms, and occupants becoming less responsive after repeated false alarms. These mechanisms illustrate why operational history matters. A description of the installed equipment alone does not capture the conditions under which protection will be demanded.[11]

Modelling system success

Reliability models describe the logical requirements for success. Modarres and Joglar-Billoch explain that a reliability block diagram need not reproduce the physical arrangement of equipment. A series representation means that every represented function is required. A parallel representation means that at least one alternative can meet the complete stated requirement.[12]

For two independent functions assessed over the same mission, the simple series and parallel expressions are shown below. R1 and R2 are their dimensionless success probabilities. The parallel expression applies only when either function alone is sufficient. Physical duplication does not establish either independence or sufficient capacity.[13]

Rseries = R1 × R2
Rparallel = 1 − (1 − R1)(1 − R2)
Symbols in the reliability expressions.[13][14]
Symbol Meaning Unit
R1, R2 Success probability of each modelled function over the specified mission 1, dimensionless
Rseries Success probability when both independent functions are required 1, dimensionless
Rparallel Success probability when either independent alternative alone is sufficient 1, dimensionless
Rsystem Complete-system success probability, including shared requirements 1, dimensionless

The first expression requires both functions to succeed. The second calculates the probability that both alternatives fail and subtracts it from one. The probabilities must refer to compatible conditions and durations. Changing the success requirement can change the model structure even when the physical equipment remains identical.[12][13]

Standby arrangements require additional attention. A reserve function may depend on detecting the need for transfer, successfully switching and then operating for the remaining mission. Modarres and Joglar-Billoch discuss these requirements separately from simple parallel operation. Treating a standby unit as an automatically successful replacement can omit the functions needed to bring it into service.[15]

Fault tree analysis provides a way to represent combinations of failures that defeat a required function, including shared support and human actions. Event tree analysis places system success or failure within the sequence following initiation. The supporting reliability model and the event-tree branch must use the same operational requirement and scenario conditions.[16][3]

Worked example with a shared supply

Consider a hypothetical model with two alternative delivery paths. Either path alone can meet the full requirement for the same specified mission. Given that a necessary shared supply succeeds, each path has success probability 0.95. Their remaining failures are assumed independent under that condition. These values are teaching assumptions, not measured fire-pump or sprinkler reliabilities.[13][14]

A shared supply feeds two alternative successful delivery paths in a logical reliability block diagram.
The shared supply remains necessary even when either delivery path can meet the requirement. Based on Modarres and Joglar-Billoch (2002), pp. 5-31–5-32 and pp. 5-36–5-38; numerical inputs are hypothetical.

Under those assumptions, both paths fail with probability 0.05 multiplied by 0.05, or 0.0025. The probability that at least one succeeds is therefore 0.9975. This result describes the alternative paths conditional on successful shared supply. It does not yet describe the complete arrangement, because the necessary shared function has not been included.[13][14]

Now assign the shared supply a hypothetical success probability of 0.98 over the required mission. The complete arrangement succeeds only when that supply succeeds and at least one delivery path succeeds. Hall's conditional-probability rule allows the shared probability to be multiplied by the conditional parallel result. The shared function is counted once.[14][17]

Rsystem = 0.98 × 0.9975 = 0.97755 ≈ 0.978
Results for the hypothetical model, reported to three significant figures.[13][14]
Assessment boundary Success probability Failure probability
Alternative paths, given successful shared supply 0.998 0.00250
Complete arrangement including shared supply 0.978 0.0225

The complete failure probability is approximately 0.0225, or 2.25%, compared with 0.250% for the alternatives when shared-supply success is certain. The difference arises from the necessary shared function. It does not mean redundancy has no benefit; it means the independent parallel calculation describes only part of the system boundary.[13][14][17]

Two bars compare failure probabilities of 0.250% for the alternative paths and approximately 2.25% for the complete arrangement.
The model boundary changes the calculated failure probability when a necessary shared supply is included. Based on Modarres and Joglar-Billoch (2002), p. 5-32, and Hall (2002), p. 1-184; numerical inputs are hypothetical.

The same result can be checked by separating two mutually exclusive failure cases. Shared-supply failure contributes 0.02. When that supply succeeds, simultaneous residual failure of the two paths contributes 0.98 multiplied by 0.0025, or 0.00245. Adding these contributions gives the unrounded result 0.02245 without counting the shared failure twice. Intermediate products remain unrounded; final reported probabilities use three significant figures.[14]

Dependencies and common failures

Common cause failure can defeat several protective functions through one cause. Barry identifies loss of water or power, freezing, flooding and mechanical damage as relevant mechanisms. These cannot be represented adequately by assuming that every component or system fails independently. Their significance depends on which functions share the affected resource or exposure.[18]

Modarres and Joglar-Billoch illustrate two suppression paths with shared water, power and actuation dependencies. Their example explicitly requires those dependencies to be included when quantifying failure. Separate pumps or nozzles do not remove a common dependency elsewhere in the chain. The analysis must follow each credited function back through its supporting functions.[17]

Dependencies may also connect different types of protection. The ABCB handbook describes detection affecting occupant warning, notification of the fire service and smoke management. Their contributions cannot be evaluated solely as unrelated equipment percentages. The assessment must consider how one unsuccessful upstream function changes the conditions for each subsequent action.[19]

A useful model therefore identifies shared resources and common exposures before assigning independent probabilities. Where independence is not appropriate, conditional probabilities can represent the relevant relationships. Hall's definition makes clear that dependence does not require one event to cause another directly. Both may depend on a third condition that the model must retain.[14][18]

Data, uncertainty and continued validity

Fire statistics can inform reliability only when the observations match the success definition and assessed population. The ABCB Data Sheets identify historical data, fault and event trees, technical literature and expert judgement as available approaches. They also explain that evidence differs between active devices and passive construction, for which the pre-fire condition may be difficult to reconstruct.[2]

Plant records can reflect the actual equipment, environment and maintenance programme, but their limitations remain important. Barry notes unrecorded failures, incomplete logs and short observation histories. Generic published data may cover different designs or operating conditions. Any adjustment requires an explicit basis rather than assuming the generic average describes the particular installation.[20]

Evidence at component level should not be mistaken for evidence at complete-system level. A recorded pump start does not, on its own, demonstrate that valves, distribution and discharge met the specified requirement. Barry's input, control and output breakdown identifies the interfaces that remain outside such an observation.[6]

Failure rates and reliability probabilities are different quantities. A rate relates failures to time, whereas reliability concerns survival or successful performance over a specified interval. Modarres and Joglar-Billoch discuss several possible failure-time distributions. A constant-rate model is one assumption, not a universal property of fire protection equipment.[21]

An absence of recorded failures does not establish perfect reliability. Modarres and Joglar-Billoch explain that estimates and confidence depend on the amount and quality of data and the assumed model. A short successful history leaves uncertainty about rare failures. The reported result should preserve that uncertainty rather than replace it with an unsupported value of one.[21]

The evidence basis can change during the building's life. Barry identifies age, environment, maintenance, testing, management of change and physical-damage exposure as relevant influences. The ABCB's 2020 data framework assumes appropriate installation, commissioning and continuing maintenance. Those historical assumptions need checking before its estimates are applied to an existing system.[22][23]

The practical consequence appears when comparing protection alternatives. The ABCB handbook presents two systems with equivalent outcomes when functioning but different failure probabilities. A deterministic assessment of successful operation alone would miss that difference. Probability information must therefore be combined with the consequences of each state before concluding that the alternatives provide equivalent protection. The same comparison must retain consequences of failure, rather than treating reliability as a complete measure of fire risk.[24]

A reliability assessment should retain the function, scenario, duration, boundary, dependencies, evidence and uncertainty supporting each value. These details allow later changes to be evaluated against the original assumptions. They also connect the numerical estimate to the maintenance and operational conditions needed for the credited protection to remain available and effective.[9][22][23]

See also

Notes

  1. ↑ 1.0 1.1 Modarres and Joglar-Billoch (2002), pp. 5-24–5-25.
  2. ↑ 2.0 2.1 Australian Building Codes Board (2020a), pp. 62–64.
  3. ↑ 3.0 3.1 3.2 Barry (2002), pp. 5-194–5-199.
  4. ↑ 4.0 4.1 Barry (2002), pp. 5-203–5-205.
  5. ↑ 5.0 5.1 Barry (2002), pp. 5-198–5-199.
  6. ↑ 6.0 6.1 Barry (2002), p. 5-205, Table 5-13.7.
  7. ↑ 7.0 7.1 Barry (2002), pp. 5-200–5-203.
  8. ↑ 8.0 8.1 8.2 Australian Building Codes Board (2020a), pp. 58–60.
  9. ↑ 9.0 9.1 Barry (2002), p. 5-204.
  10. ↑ Barry (2002), pp. 5-199–5-200.
  11. ↑ 11.0 11.1 Barry (2002), p. 5-202.
  12. ↑ 12.0 12.1 Modarres and Joglar-Billoch (2002), pp. 5-31–5-32.
  13. ↑ 13.0 13.1 13.2 13.3 13.4 13.5 13.6 Modarres and Joglar-Billoch (2002), p. 5-32, equations 37 and 42–43.
  14. ↑ 14.0 14.1 14.2 14.3 14.4 14.5 14.6 14.7 Hall (2002), p. 1-184.
  15. ↑ Modarres and Joglar-Billoch (2002), pp. 5-33–5-34.
  16. ↑ Modarres and Joglar-Billoch (2002), pp. 5-35–5-38.
  17. ↑ 17.0 17.1 17.2 Modarres and Joglar-Billoch (2002), pp. 5-36–5-38.
  18. ↑ 18.0 18.1 Barry (2002), pp. 5-201–5-204.
  19. ↑ Australian Building Codes Board (2020b), pp. 120–122.
  20. ↑ Barry (2002), pp. 5-204–5-206.
  21. ↑ 21.0 21.1 Modarres and Joglar-Billoch (2002), pp. 5-25–5-29.
  22. ↑ 22.0 22.1 Barry (2002), pp. 5-206–5-207.
  23. ↑ 23.0 23.1 Australian Building Codes Board (2020a), pp. 61–62.
  24. ↑ Australian Building Codes Board (2020b), pp. 118–119.

References

  • Australian Building Codes Board. (2020a). Fire safety verification method data sheets: Handbook annex (Version 1.1).
  • Australian Building Codes Board. (2020b). Fire safety verification method handbook (Version 1.1).
  • Barry, T. F. (2002). Quantitative risk assessment in chemical process industries. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., Chapter 5-13, pp. 5-176–5-213). National Fire Protection Association.
  • Hall, J. R., Jr. (2002). Probability concepts. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., Chapter 1-11, pp. 1-182–1-192). National Fire Protection Association.
  • Modarres, M., & Joglar-Billoch, F. (2002). Reliability. In P. J. DiNenno (Ed.), SFPE handbook of fire protection engineering (3rd ed., Chapter 5-3, pp. 5-24–5-39). National Fire Protection Association.
Retrieved from "https://fire-safety.academy/w/index.php?title=Reliability_of_fire_protection_systems&oldid=2145"