Contents
A fire hazard analysis (FHA) is an engineering analysis that determines the expected outcome of a specific set of conditions, called a fire scenario, without analysing how likely that scenario is.[1][2] In a fire risk assessment, by contrast, the results of hazard analyses are weighted by their likelihood and summed.[1] An FHA serves either to determine the hazards present in an existing or planned facility, or to test whether trial designs achieve a set of fire safety goals.[1]
Its steps run from selecting a target outcome and the scenarios of concern, through predicting fire effects and occupant movement, to examining uncertainty and documenting the work.[3] In 2008 the NFPA Fire Protection Handbook described quantitative fire hazard analysis as becoming the fundamental tool of modern fire safety engineering practice and the enabling technology for performance-based codes.[4] In the United States, FHAs were required for Department of Energy facilities and performed for nuclear power plants, and similar techniques were used to regulate fire safety in aircraft and passenger trains.[3][5][6]
Definition and relation to risk
Hurley and Bukowski (2008) divide the methods for estimating the potential impact of fire into risk-based and hazard-based methods. Both estimate the consequences of possible events, but only risk-based methods also analyse how likely the scenarios are.[1] Hall and Watts (2008) define a fire hazard as any condition with the potential to create fire-related harm, and note that the terminology is not standardised.[7] In their usage, hazard analysis describes analyses that do not treat most probabilities explicitly, particularly those of ignition and of occupant characteristics and behaviour. Risk analysis is reserved for analyses that address all relevant probabilities.[7]
Watts and Hall (2002) frame fire risk analysis around three questions: what could happen, how bad would it be, and how likely is it. Fire hazard analysis is a term often used for analyses of the first two questions without the third.[2] Bukowski (2002) describes a fire hazard assessment as a systematic examination of all potential fire hazards that could lead to the consequence to be avoided, including scenarios that have not yet occurred.[8] A fire risk assessment then begins with a fire hazard assessment and weights each consequence by the likelihood of its scenario. A high-consequence, low-likelihood scenario and a frequent, low-consequence scenario can then represent equal risk.[8]
The two approaches also treat assumptions differently. According to Hall (2002b), fire risk assessment predicts what will happen, so it weights scenarios by their probabilities and tends to set assumptions at typical or average conditions. Fire hazard assessment predicts what might happen for which the designer is responsible, which is where the probable worst-case scenario becomes relevant.[9] It addresses only challenging, high-severity scenarios, does not discount them by probability, and declares some of them too challenging, an all-or-nothing approach to scenarios.[9] Conservative assumptions, those that present the greatest fire challenge, are the usual choice in fire protection engineering studies, but usually not the right answer for risk analysis.[10]
Jenaway (2008) uses the terms for fire department planning at community scale. He defines hazard as the level of harm associated with exposure to a fire or its effluent, and risk as the expected loss from a fire of a given severity. Hazard analysis is then an analysis to determine the potential impact of a fire scenario when the hazard is involved in fire.[11] In the chemical process industries, a related evaluation, process hazard analysis, is one programme segment of the process safety management rule of the US Occupational Safety and Health Administration (29 CFR 1910.119), introduced in 1992.[12]
Applications
In its design role, an FHA can be used as part of performance-based design.[3] The performance-based design procedure shown by Watts (2008b), after a 2007 SFPE guide, includes developing design fire scenarios and trial designs and evaluating them against performance criteria.[13] In his chapter on the systems approach to fire-safe design, Watts names the combustion potential of a room as the basis for a fire growth hazard analysis. That potential depends on the fuel load, the interior finish, the air supply and the size, shape and construction of the room.[14]
In the United States, the Department of Energy (DOE) required FHAs for all its facilities. The objectives in DOE 5480.7A, as quoted by Hurley and Bukowski (2008), include minimising the potential for fire, preventing releases of radiological or other hazardous material and providing an acceptable degree of life safety.[3] They also include protecting critical process control and safety systems, mission continuity, and keeping property damage within acceptable levels, for example $150 million per incident.[3] One part of a DOE-required FHA assumes both that automatic systems fail and that the fire department does not respond. This determines a worst-case loss and establishes the real value of these systems.[3]
For US nuclear power plants, NRC Regulatory Guide 1.189 specified the objectives of an FHA in the account by Iqbal and Salley (2004). The analysis considers in situ and transient fire hazards and determines the consequences of fire anywhere in the plant for safe shutdown and for the release of radioactivity.[5] It also specifies fire prevention, detection, suppression and containment measures for each fire area containing equipment important to safety.[5]
Their report, NUREG-1805, provides spreadsheet tools to help NRC inspectors perform FHAs quantitatively.[15] The plant FHA also feeds probabilistic work. The fire PRA method of NUREG/CR-6850 starts partitioning a plant from the fire areas and zones documented in the FHA.[16] It can also use the cable quantities recorded in the FHA to weight ignition sources.[17]
Commercial aviation and passenger rail used fire hazard assessment to achieve safety goals before these techniques became common in buildings. The US Federal Aviation Administration pioneered its use in regulation; Bukowski (2002) cited ASTM E2061 as a procedure for rail vehicles.[6] The International Fire Engineering Guidelines (2005) were produced by bodies from Australia, Canada, New Zealand and the United States.[18] In them, preparing a fire engineering brief includes a systematic review of the potential fire hazards of the building. The factors to consider include the general layout, activities such as maintenance, ignition sources and fuel sources.[19]
Steps and scenario selection
Hurley and Bukowski (2008) list seven steps. The first four are selecting a target outcome, determining the scenarios of concern, selecting methods to predict the growth of fire effects, and calculating the time occupants need to reach safety.[3] The remaining steps are analysing the impact of exposure on occupants or property, examining uncertainty, and documenting the analysis, including the basis for the models and input data.[3] Bukowski (2002) applies the same sequence to transportation vehicles, with the selection of design fires as a separate step.[20]
The target outcome most often specified is the avoidance of occupant fatalities. Another might be to give firefighters protected areas from which to fight fires in high-rise buildings.[3] An insurer might want to limit the maximum probable loss, and a manufacturer might want to avoid failing to meet orders.[3] At a US nuclear plant, the first step is to identify the target, normally a safety-related component, and electrical cables are the primary target in most analyses.[21] NUREG-1805 uses damage thresholds of 370 °C and 10 kW/m² for cables that pass the IEEE-383 fire test, and 218 °C and 5 kW/m² for cables that do not.[21]
Records of past fires and statistics from NFPA or the US National Fire Incident Reporting System on ignition sources, first items ignited and rooms of origin help to identify the conditions to avoid. Anecdotal accounts of single incidents may not represent the major part of the problem.[3] Meacham (2002) describes the survey behind a fire hazard assessment: ignition sources, fuels and their arrangement, compartment configurations and fire safety features. The analyst then assumes ignition or established burning and predicts fire growth, spread and impact.[22] Reviewing loss data helps to avoid focusing on unique hazards while overlooking more common, equally important ones.[22] Tools range from checklists, safety reviews and "what if" questions to hazard and operability studies, failure modes and effects analysis, and fault tree and event tree analysis.[23]
Significant fires seem to involve a series of failures, so the consequences of things not going according to plan must be examined.[3] The 2006 edition of NFPA 101, Life Safety Code, has a performance-based option with eight design fire scenarios, which Hurley and Bukowski (2008) illustrate for a multistorey hotel with meeting rooms.[24] Scenario 8 applies this idea: a common fire starts while the fire alarm system or the sprinkler system, in turn, is rendered ineffective.[24]
| No. | Scenario (summary) | Possible design fire in a hotel |
|---|---|---|
| 1 | Typical fire for the occupancy | Patron smoking in bed; a "sterno-initiated" fire in a meeting room or restaurant |
| 2 | Ultrafast fire in the primary means of egress, interior doors open | Flammable liquid fire in the corridor near an exit door |
| 3 | Fire in a normally unoccupied room that may endanger many occupants | Janitor's closet, oily rags or cleaning fluid |
| 4 | Fire in a concealed wall or ceiling space next to a large occupied room | Drop ceiling above a bathroom, likely an electrical fire |
| 5 | Slowly developing fire, shielded from protection systems, near a high-occupancy area | Maid's cart, or under a display table in a meeting room |
| 6 | Most severe fire from the largest fuel load of normal operation | Furniture stored during remodelling, or stored chairs |
| 7 | Outside exposure fire | Other buildings, or a wildland fire |
| 8 | Ordinary combustibles with each passive or active protection system in turn ineffective (may be omitted if its reliability and the design performance without it satisfy the authority having jurisdiction) | Failure of rated walls or floors, sprinklers or fire alarm, including partial failures such as one sprinkler |
Fire scenarios and design fires
A fire scenario is a set of conditions that defines the development of a fire and the spread of combustion products. It comprises building characteristics, occupant characteristics and fire characteristics, the last describing ignition and growth.[3] Design fire scenarios are the subset of fire scenarios selected to analyse a trial design. They should be based on fires with a reasonable likelihood of developing, covering reasonably expected and worst-case fires.[3] Some risk therefore enters even a hazard-based analysis: a technically plausible but extremely unlikely fire may be left out.[3]
For scenario 8 of NFPA 101, the performance objectives need not be fully met, given the normally high reliability of the systems, but stakeholders should feel that the losses are not catastrophic.[25] A risk assessment would weight such failures by their probability. In a hazard analysis, where the objective is hazard avoidance, the contribution of low-probability events is more subjective.[25] The analysis must also state its bounding conditions, the future changes to the building that it allows for, such as a hotel floor that becomes a meeting room floor.[25] A scenario excluded as too unlikely, such as a tanker truck crashing into the building, implies a risk that stakeholders need to understand and accept.[26]
A design fire curve describes the heat release rate of the fire over time, in four phases: ignition, growth, steady burning and decay.[27] Its role resembles the assumed loading in a structural analysis, and the greatest challenge is not necessarily the largest fire, especially in a sprinklered building.[27] Detector or sprinkler activation may need only the growth phase, smoke control sizing only the maximum heat release rate, and an evacuation analysis the growth and fully developed stages.[27] For a routine FHA at a nuclear plant, NUREG-1805 accepts a broad approximation of burning rates. The heat release rate is not a fundamental property of a fuel and is usually determined from tests.[28]
The choice of growth rate matters most for a realistic prediction of detector and sprinkler activation, the start of evacuation and the first exposure of occupants. In 1972 Heskestad proposed that early fire growth follows a power law in time, an assumption supported by experimental data. For most flaming fires, except flammable liquids and some others, the exponent is 2:[29]
| Symbol | Description | Unit |
|---|---|---|
| heat release rate | kW | |
| fire intensity (growth) coefficient | kW/s² | |
| time from the start of established flaming | s |
| Growth class | Time to 1,055 kW (s) | α (kW/s²) | Typical fuels |
|---|---|---|---|
| Slow | 600 | 0.0029 | Thick solid objects such as a solid wood table or dresser |
| Medium | 300 | 0.012 | Lower-density solids such as upholstered furniture and mattresses |
| Fast | 150 | 0.047 | Thin items such as paper, cardboard boxes and draperies |
| Ultrafast | 75 | 0.19 | Some flammable liquids, some older upholstered furniture and mattresses, highly volatile fuels |
Van Rossum (2027) gives the same four classes, defined by the time to reach 1 MW, and tabulates growth rates for Dutch occupancy types from the national annex NEN-EN 1991-1-2/NB. These values are intended as a safe estimate, because many real fires grow more slowly.[31] The curves assume a reasonably large flaming ignition source; with small sources there is an incubation period before established flaming.[30] They are unsuitable for fires that also spread vertically, as in high racks.[31] Hurley and Bukowski consider the Australian link between growth rate and fuel load unjustified, since 10 kg of wood as a solid cube, a crib or sawdust grows at very different rates.[30]
For fully developed fires, computer models are generally no more accurate than simple hand calculations.[30] Decay is usually omitted because its hazards are small compared with the fully developed phase. Where needed, a temperature decay of 10 °C/min for fires shorter than 60 minutes, and 7 °C/min for fires longer than 60 minutes, can be used.[32] Assuming that an operating sprinkler puts the fire out at once is not conservative.[32] Whether flashover is possible can be checked with correlations such as that of Babrauskas, , the minimum heat release rate in kW for an opening of area (m²) and height (m).[32]
Predicting fire effects: a worked example
Once the design fire is known, the hazards can be predicted. They include radiant heat flux, smoke production, fire plume and ceiling jet temperatures and velocities, toxic species production and the depth of the upper smoke layer.[33] Only those needed for the goal are quantified, so the activation of a thermal detector requires only the plume and ceiling jet conditions.[33] Fire modelling ranges from algebraic equations to computer models, which are particularly useful where many repeated calculations must be performed.[33]
Hurley and Bukowski (2008) give an example of such a calculation. A heat detector is located on a 3 m high ceiling, 2 m from a fire on the floor with a constant heat release rate of 1,000 kW. The room is at 20 °C and large enough that no smoke layer forms quickly.[34] The detector has a temperature rating of 75 °C and a response time index (RTI) of 50 m½s½.[34] Away from the plume axis (r/H above 0.18 for temperature and above 0.15 for velocity), Alpert's ceiling jet correlations give the gas temperature rise and velocity:[33][35]
The algorithm is similar to that of the computer model DETACT-QS. At each time step the ceiling jet temperature and velocity are evaluated, and the detector temperature is advanced according to its response time index.[34] The handbook solves the detector equation step by step with the Euler method:[34]
| Symbol | Description | Unit |
|---|---|---|
| total heat release rate of the fire | kW | |
| height of the ceiling above the fire | m | |
| horizontal distance from the fire axis | m | |
| ceiling jet temperature rise above ambient | °C | |
| ceiling jet velocity at the detector | m/s | |
| , | ceiling jet temperature; ambient temperature | °C |
| detector temperature | °C | |
| RTI | response time index of the detector | m½s½ |
| time | s |
With r/H = 0.67, the correlations give a temperature rise of about 113 °C, so a gas temperature of about 133 °C, and a velocity of about 1.9 m/s (values computed here from the equations above). Hurley and Bukowski report a predicted activation time of 24.000 s with a 1 s time step, 24.200 s with 0.1 s, and 24.230 and 24.229 s with 0.01 and 0.001 s.[34] Solving the same equations reproduces these four values, and the exact solution of the detector equation also gives about 24.2 s.
The ceiling jet equations are valid only for horizontal, unobstructed ceilings without a smoke layer; where a layer forms, higher temperature rises can be expected.[33] Alpert based his correlations on fires of 668 kW to 98 MW under ceilings 4.6 to 15.5 m high, so the 3 m ceiling of the example lies below the tested heights.[35] The method is also quasi-steady: it ignores the transport delay from the fire to the detector.[34]
Toxicity is often assessed with Purser's fractional effective dose, but no single value for design has been widely agreed, even for average populations. Some groups, such as asthmatics, the old and the young, are more susceptible.[36] Egress calculations must include the time for detection and notification, for deciding to act and for movement.[37] In the 2005 international guidelines, such a deterministic analysis is a timeline. For an absolute evaluation, the available safe evacuation time should exceed the required safe evacuation time by a safety margin set in the fire engineering brief.[38]
Models, data and documentation
Choosing a model requires a thorough understanding of its assumptions and limitations. Single-room models suit conditions limited to one enclosed space, while multiple-compartment models are needed where the area of interest extends beyond it.[34] A model should limit burning by the available oxygen and, in multiple compartments, track unburned fuel. Without these features it overpredicts conditions in the room of origin and underpredicts them elsewhere.[34] A constant heat loss fraction ignores that heat loss varies during a fire, and unlimited plume entrainment in tall spaces can make predicted egress times too long or too short.[37]
Even a correct model can give seriously wrong results if the input data do not represent the case. The FHA should therefore list every data value, its source and a discussion of its uncertainty.[37] Only rigorously documented models should be used in applications involving public safety, such as code enforcement or litigation, and ASTM E1472-05 is a guide for documenting fire model software.[37] The documentation of the FHA itself cannot be omitted. It should identify the assumptions, bounding conditions, scenarios considered and limitations to the authorities having jurisdiction, the owner and other parties.[27]
A final review checks whether the results make sense. A counterintuitive result is probably wrong, as when a model predicted temperatures approaching the surface temperature of the sun.[4] Some results only look wrong, such as a falling temperature next to a growing fire, caused by cold outdoor air drawn in through an open door.[4]
Uncertainty and limitations
Uncertainty in an FHA comes from the assumptions in the models, from the representativeness of the input data and, in evacuation calculations, from the variability of real people.[4] The classic treatment in building design is a safety factor large enough that the result stays acceptable even if all uncertainty acted in the same direction. Because the design fires are chosen as the worst likely scenarios, no safety factor is needed for fire development, unless the result is very sensitive to some assumption or data.[4] The FHA report should discuss the representativeness of the data and the sensitivity of the results; where the sensitivity is not obvious, the data should be varied to their limits.[4]
NUREG-1805 recommends bounding calculations, such as door open versus closed or ventilation on versus off, to form a window of credible solutions.[39] Predicted hazards depend on the design fire scenarios analysed. These should be challenging enough to represent a realistic worst case, but not so challenging that their likelihood is too remote.[4] Because a hazard analysis does not weight scenarios by probability, the treatment of rare events remains a matter of judgement.[25] Simple methods may not suffice where geometry is complex, where cost and benefit must be optimised or where predicted values are close to acceptable limits. Even then they remain useful for initial predictions and as a reality check.[4]
In the Netherlands, fire safety engineering is applied mainly to non-standard buildings, where the prescriptive provisions are not effective or do not apply. Examples are tunnels, towers with a floor above 200 m and buildings with large atria.[40] Van Rossum (2027) notes that the quantitative approach there lacks risk-based limit values. The limit values of the prescriptive rules exist, but cannot be translated into unambiguous risk-based ones.[41]
See also
Notes
- ↑ 1.0 1.1 1.2 1.3 Hurley and Bukowski (2008), p. 3-121.
- ↑ 2.0 2.1 Watts and Hall (2002), p. 5-2.
- ↑ 3.00 3.01 3.02 3.03 3.04 3.05 3.06 3.07 3.08 3.09 3.10 3.11 3.12 3.13 3.14 Hurley and Bukowski (2008), p. 3-122.
- ↑ 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 Hurley and Bukowski (2008), p. 3-133.
- ↑ 5.0 5.1 5.2 Iqbal and Salley (2004), p. 1-3.
- ↑ 6.0 6.1 Bukowski (2002), p. 5-227.
- ↑ 7.0 7.1 Hall and Watts (2008), p. 3-135.
- ↑ 8.0 8.1 Bukowski (2002), p. 5-231.
- ↑ 9.0 9.1 Hall (2002b), p. 5-147.
- ↑ Hall (2002b), p. 5-148.
- ↑ Jenaway (2008), p. 12-43.
- ↑ Snyder (2008), p. 12-184.
- ↑ Watts (2008b), p. 1-158.
- ↑ Watts (2008b), p. 1-164.
- ↑ Iqbal and Salley (2004), p. 1-1.
- ↑ Nowlen et al. (2005b), p. 1-4.
- ↑ Nowlen et al. (2005b), p. 6-15.
- ↑ National Research Council of Canada et al. (2005), PDF p. 4 (unnumbered copyright page).
- ↑ National Research Council of Canada et al. (2005), p. 1.2-9.
- ↑ Bukowski (2002), p. 5-229.
- ↑ 21.0 21.1 Iqbal and Salley (2004), p. 2-19.
- ↑ 22.0 22.1 Meacham (2002), p. 5-155.
- ↑ Meacham (2002), pp. 5-155–5-156.
- ↑ 24.0 24.1 24.2 Hurley and Bukowski (2008), pp. 3-122–3-123.
- ↑ 25.0 25.1 25.2 25.3 Hurley and Bukowski (2008), p. 3-123.
- ↑ Hurley and Bukowski (2008), p. 3-124.
- ↑ 27.0 27.1 27.2 27.3 Hurley and Bukowski (2008), p. 3-125.
- ↑ Iqbal and Salley (2004), p. 3-1.
- ↑ Hurley and Bukowski (2008), pp. 3-126–3-127.
- ↑ 30.0 30.1 30.2 30.3 Hurley and Bukowski (2008), p. 3-127.
- ↑ 31.0 31.1 Van Rossum (2027), §8.6.
- ↑ 32.0 32.1 32.2 Hurley and Bukowski (2008), p. 3-128.
- ↑ 33.0 33.1 33.2 33.3 33.4 Hurley and Bukowski (2008), p. 3-129.
- ↑ 34.0 34.1 34.2 34.3 34.4 34.5 34.6 34.7 Hurley and Bukowski (2008), p. 3-131.
- ↑ 35.0 35.1 Alpert (2002), p. 2-19.
- ↑ Hurley and Bukowski (2008), p. 3-130.
- ↑ 37.0 37.1 37.2 37.3 Hurley and Bukowski (2008), p. 3-132.
- ↑ National Research Council of Canada et al. (2005), p. 1.2-22.
- ↑ Iqbal and Salley (2004), p. xl.
- ↑ Van Rossum (2027), §1.5.
- ↑ Van Rossum (2027), §1.4.
References
- Alpert, R. L. (2002). Ceiling jet flows. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 2-18–2-31). National Fire Protection Association; Society of Fire Protection Engineers.
- Bukowski, R. W. (2002). Fire hazard assessment for transportation vehicles. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-227–5-233). National Fire Protection Association; Society of Fire Protection Engineers.
- Hall, J. R., Jr. (2002b). Product fire risk. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-143–5-152). National Fire Protection Association; Society of Fire Protection Engineers.
- Hall, J. R., Jr., & Watts, J. M., Jr. (2008). Fire risk analysis. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 3-135–3-143). National Fire Protection Association.
- Hurley, M. J., & Bukowski, R. W. (2008). Fire hazard analysis techniques. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 3-121–3-134). National Fire Protection Association.
- Iqbal, N., & Salley, M. H. (2004). Fire Dynamics Tools (FDTs): Quantitative fire hazard analysis methods for the U.S. Nuclear Regulatory Commission Fire Protection Inspection Program (NUREG-1805). U.S. Nuclear Regulatory Commission, Office of Nuclear Reactor Regulation.
- Jenaway, W. F. (2008). Needs assessment and hazard analysis. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 12-41–12-50). National Fire Protection Association.
- Meacham, B. J. (2002). Building fire risk analysis. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-153–5-175). National Fire Protection Association; Society of Fire Protection Engineers.
- National Research Council of Canada, International Code Council, Department of Building and Housing, & Australian Building Codes Board. (2005). International fire engineering guidelines (2005 ed.). Australian Building Codes Board.
- Nowlen, S. P., Najafi, B., Wyant, F. J., Forester, J., Kazarians, M., Kolaczkowski, A., Joglar, F., Funk, D., Hannaman, G., & Anoba, R. (2005b). EPRI/NRC-RES fire PRA methodology for nuclear power facilities: Vol. 2. Detailed methodology (NUREG/CR-6850; EPRI 1011989). Electric Power Research Institute; U.S. Nuclear Regulatory Commission, Office of Nuclear Regulatory Research.
- Snyder, M. (2008). Industrial fire loss prevention. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 12-177–12-198). National Fire Protection Association.
- Van Rossum, E. (2027). Introduction to Fire Safety Engineering. Fire Safety Academy.
- Watts, J. M., Jr. (2008b). Systems approach to fire-safe building design. In A. E. Cote, C. C. Grant, J. R. Hall, Jr., & R. E. Solomon (Eds.), Fire protection handbook (20th ed., pp. 1-157–1-170). National Fire Protection Association.
- Watts, J. M., Jr., & Hall, J. R., Jr. (2002). Introduction to fire risk analysis. In P. J. DiNenno, D. Drysdale, C. L. Beyler, W. D. Walton, R. L. P. Custer, J. R. Hall Jr., & J. M. Watts Jr. (Eds.), SFPE handbook of fire protection engineering (3rd ed., pp. 5-1–5-7). National Fire Protection Association; Society of Fire Protection Engineers.