Fire Protection as the Underpinning of Good Process Safety Programs Key Words: fires; protection, fire; building Background James Milke, Ph.D., P.E., Professor and Chair Department of Fire Protection Engineering University of Maryland Numerous accidents involving process systems either include fires as part of the initiating event or occur subsequently. As such, the most effective fire safety strategies are those employed to prevent the fire or mitigate the consequences. While fire protection can also be provided during the response phase, such an approach should only be regarded as a matter of last resort and not be the principal means of providing protection. This paper will concentrate on the involvement of fire protection engineering in the mitigation phase of a process safety program. Fire protection engineers are involved in the assessment of hazards and the selection of fire protection strategies which can reduce the risk to an acceptable level. Fire protection strategies may include the installation of a variety of approaches, such as passive and active fire protection systems, manual intervention and siting. Passive systems include fire rated barriers and protection of openings in those barriers, while active systems include systems such as fire detectors and sprinklers. Manual intervention may include the manual activation of fixed fire protection systems or firefighting activities by facility fire brigades or municipal fire departments. The manual firefighting activities are typically considered to be in the response phase. An analysis of the contribution of a particular fire protection system to the achievement of specified objectives should include an assessment of the effectiveness and reliability of the proposed fire protection systems. These objectives may be implicit, being incorporated into the basis of regulatory requirements in prescriptive codes, or may be explicit where performancebased designs are proposed. Because fire protection systems have many variations, with few standard, one-size-fits-all designs, understanding the performance objectives intended for the system is essential in order to identify the correct type of system, as well as to formulate the best design options for the selected system, as will be outlined in the remainder of this paper.
Passive Protection Methods A fire resistant building assembly has an ability to confine a fire, continues to perform a given structural function, or both [IBC, 2012]. Generally, meeting prescriptive requirements is accomplished by identifying an assembly which has been subjected to a standard test at a laboratory such as UL. An example of a fire resistant assembly with a spray applied fire protection material on a steel column is illustrated in Figure 1 [UL, 2014]. Figure 1. Fire Resistance Rated Steel Column Design (UL 701) The fire resistance rating is derived from results obtained from a standard test [ASTM, 2012]. Performance criteria stipulated in the standard reflect the intended functions of fire resistant assemblies. For example, the barrier function is assessed by criteria limiting the temperature rise on the unexposed side of the assembly to an average of 139 C and 181 C at a single point. These criteria are based on the spontaneous ignition temperature of ordinary combustibles which may be in contact with an unexposed surface of a fire barrier. The structural integrity function is fulfilled if there is no collapse or extraordinary deflection, or may be determined by temperature limits for steel components relating to the decrease in tensile strength with temperature, e.g. for steel columns an average temperature of 538 C and 649 C at a single point and for steel rebars the temperature limit is 593 C. A common misunderstanding is that the fire resistance rating infers the number of hours that an assembly will perform successfully in an actual fire. In ASTM E119, the following important statement is made concerning the relevance of the fire resistance rating acquired [2012]: Sect 1.2: It is the intent that classification shall register comparative performance to specific fire-test conditions during the period of exposure and shall not be construed as having determined suitability for use under other conditions or after fire exposure. It is important to recognize that the temperatures of the structural components and on the unexposed side will depend on the fire exposure. Further, while the test is a large-scale test, no attempt is made to scale the test results to the size of the actual assemblies or replicate the structural end conditions. As such, if the amount of time that a fire resistant assembly will continue to function despite exposure to an actual fire needs to be determined, the fire resistance rating isn t relevant. The time estimate can only be obtained by applying a thermo-mechanical analysis [Milke, 1999]. A
flowchart of this type of analysis is depicted in Figure 2. As indicated in the flowchart, the first step is to estimate conditions posed by proposed design fires. This output of this analysis results in a determination of the time-dependent temperature or heat flux exposure for the assembly. The next step is to conduct a thermal response analysis which involves analyzing the heat transfer of the fire to the assembly, where the output from the fire analysis serves as boundary conditions for the analysis. The thermal response analysis provides a determination of the temperature distribution caused by the fire exposure. These results may be compared to temperature endpoint criteria provided in ASTME E119, or may be used as inputs to the structural analysis in the next step. The structural response analysis consists of an analysis of deflections or stresses developed within the assembly and can be compared to prescribed endpoint limits (perhaps relating to maximum allowable stresses or deflections). Fire Exposure Incident heat, temperature of exposure vs. time Material Properties Thermal Response Temperature of structural elements, unexposed side Structural Response Stresses, deflection, collapse Figure 2. Outline of Thermo-mechanical Analysis for Fire Resistance Assessment As an example application of a performance-based fire resistance analysis, consider the impact of a fire exposing a steel tower supporting a foam turret protecting an off-loading facility of crude oil from a barge docked at a pier. The fire scenario involves a spill of crude oil on the barge. Detection of a fire is accomplished either by security personnel observing the fire or by a drop in pressure in the transfer line for the crude oil. Release of the foam system is done at a watchtower (a different tower than that supported the foam turret). The analysis needs to address whether the foam system can be actuated prior to the steel components of the tower being heated to a failure temperature should a fire be initiated on a barge. While the security guard watchtower is constantly staffed, at times only one guard might be present who is also required to walk the property for a security check. This staffing situation requires that the analysis account for the possibility that the lone guard is at the far end of the property at the time of the fire and thus must estimate the time required for the guard to walk (or run) from the remote location and traverse the watchtower stairs to actuate the foam system.
Radiant Heat Flux (kw/m²) The radiant heat flux from a hydrocarbon pool fire is dependent on the diameter of the pool. In addition, the smoke included in the plume has the ability to decrease the radiation emitted from the source (as compared to a plume that would contain only flames). The radiant heat flux emitted from such a plume is estimated as [Beyler, 2008]: ( ) The radiant heat flux to a target from the flame is estimated using a point source approximation, which is reasonably accurate for engineering purposes [Beyler, 2008]. The estimated radiant heat flux for a fully-engulfed, level barge incident on targets separated by a range of distances from the edge of the barge are presented in Figure 3. 18 16 14 12 10 8 6 4 2 0 RH=100%, 35 C RH=100%, 15 C RH=50% 0 10 20 30 40 50 Distance from Barge (m) Figure 3. Radiant Heat Flux vs, Distance from #6 Fuel Oil Fire However, because the fire would grow to encompass the spill during the early stages, a fire spread analysis is required. Horizontal spread rates are included for several hydrocarbon fuels in Beyler [2008]. The fire is assumed to grow until engulfing the entire barge, though the area of the barge which a spill could occupy is dependent on the heel angle of the barge in the water. An example of the transient radiant heat flux for one heel angle of the barge is presented in Figure 4. This analysis also considers a decrease in the diameter of the pool fire as oil was consumed (assuming no foam discharge). The radiant heat flux on the steel tower is then applied to estimate the temperature rise in the steel supporting elements using an elementary lumped heat capacity analysis to confirm that the tower is able to withstand the thermal insult from a pool fire until the foam system could be discharged [Milke, 2008a]. Active Protection Methods Because all active protection methods require a means of actuation, reviews of active protection methods need to start with the actuation method. Actuators include thermal components of
Radiant Heat Flux (kw/m²) sprinklers, automatic fire detectors and manual approaches. Selection of the best type of mechanism is based on response time, reliability and nuisance alarm considerations. Short response times may require very sensitive detection thresholds, which need to be balanced for an increased propensity for generating nuisance alarms. 10 9 8 7 6 5 4 3 2 1 0 RH=100%, 35 C RH=100%, 15 C RH=50% 0 30 60 90 120 150 180 Time (sec) Figure 4. Radiant Heat Flux with 6 Heel Angle Selection of detectors, i.e. initiating devices, needs to be done to meet identified performance objectives. These performance objectives should relate to having detection of the fire prior to it reaching a specific heat release rate, referred to as a threshold fire size in NFPA 72 [2012]. Selection of an initiating device for any fire scenario should consider the products of combustion to be generated, the concentration or magnitude of such products and the presence of similar products during ambient conditions. A variety of signatures are produced by fires, including thermal energy, aerosols and chemical species. Thermal energy is transported from the vicinity of the fire primarily by convection and radiation. Heat detectors and sprinklers are devices which respond to convected energy and consequently are located on or near the ceiling. Spacing for heat detectors is based on their listed spacing obtained via laboratory tests. Adjustments from the listed spacing are made considering the characteristics of ceiling jets (the initial, thin layer of smoke which travels along the underside of the ceiling) and heat transfer principles. For example, heat detectors placed in spaces taller than 3 m have decreased spacings than the listing because the ceiling jet is cooler in tall spaces than in short spaces. The relationship of the temperature of the ceiling jet with height and radial distance from the plume centerline is depicted in Figure 5. Radiation detectors, such as spark and flame detectors respond to the radiant energy produced by fires in selected wavelengths, typically in the UV or IR ranges. Smoke detectors respond to particulates produced by fires. Spacing for flame detectors is determined for each hazard
protected, considering the radiation spectrum generated for the specific fuel involved and the threshold fire size. The threshold fire size dictates the emissive power of the fire and surface area of the radiator. The analysis method is included in NFPA 72. Figure 5. Ceiling Jet Temperature There are three operating mechanisms for smoke detectors: ionization chamber (ionization smoke detectors), light scattering (photoelectric smoke detectors) and light obscuration (projected beam detectors). These three mechanisms have different sensitivities to the spectrum of particle sizes produced in fires, as summarized in Table 1. Table 1. Smoke Detection Principles Technology Relationship with Particle Characteristics Ionization Chamber MIC n i d 2 Light Scattering s n i di OD Light Obscuration n i di d i : diameter of particles in range i n i: number of particles in range i OD/l: optical density per unit length s: scattering parameter MIC: measuring ionization chamber response Given that flaming fires produce more small particles than large particles, ionization smoke detectors are more responsive to flaming fires, while photoelectric and projected beam detectors are more responsive to smoldering fires or flaming fires with high levels of soot production. i 3
Heat Release Rate (kw) Temperature ( C) Smoke detectors are spaced nominally at 9.1 m for low ceiling heights (3 m or less) and smooth ceilings. As with thermal detectors, adjustments are made for other situations. Where performance-based calculations are proposed in order to estimate the response of ionization or photoelectric smoke detectors, an appreciable amount of uncertainty exists. Much of this uncertainty is due to the calculations being based on an estimate of light obscuration or temperature, even though these metrics do not relate to the mechanisms by which these detectors operate. Some analogies have been included in NFPA 72 and the research literature to indicate the level of correlation of these measures with the operating mechanisms [Fabian and Gandhi, 2007][Milke, et al, 2008]. Selection and spacing of spot detectors in tall spaces may be challenging. For example, consider the case of needing to provide detection in a 10 m tall space where a hazard analysis of exposed electronic equipment demonstrates that detection is needed prior to the fire reaching 500 kw. The design fire involves a growing fire, following a fast-t 2 profile, i.e. one where the heat release rate versus time curve follows a parabola and reaches a value of 1 MW in 150 sec. The smoke layer temperature can be estimated from an elementary zone model analysis [Milke, 2008b]. The heat release rate and resulting temperature of the smoke layer are presented in Figure 6. As indicated in the figure, the temperature is less than 30 C at the time the fire reaches 500 kw. A separate estimate indicates that the level of visible smoke is minimal at that time as well. As a result, detection of an incident with the stated performance objective would need to be accomplished with projected beam detectors at intermediate heights (in order to intercept the smoke plume) or flame detectors. Verification of the adequacy of those alternatives would need to be analyzed, though such is beyond the scope of this paper. 700 600 500 400 300 200 100 HRR Temp 40 35 30 25 20 15 10 5 0 0 0 30 60 90 120 Time (sec) Figure 6. Temperature of Smoke Layer in Example
Suppression systems are selected based on the effectiveness of the agent, collateral damage potential and cost. Water-based systems are typically the default choice given its widespread availability. Automatic sprinklers are known for their relatively high effectiveness [Hall, 2011]. However, as compared to the mid-1900 s when there were 2 sprinkler models, selecting a particular sprinkler model is now more involved given that there are over 1,000 sprinkler models, when considering combinations of orifice size, deflector type, orientation (pendant, upright or sidewall), RTI and operating temperature (and this does not include the variations between manufacturers) 1. Whilethe down-selection is partially based on regulatory requirements noted in NFPA 13, there is still an significant number of options which may be applicable for a particular project. The final selection of the particular sprinkler depends on design objectives relative to response time, water flow requirements, and collateral damage. Annex B of NFPA 13 provides guidance for the selection of sprinklers based on performance objectives. Other aspects influencing the choice of a particular sprinkler include storage height and arrangement and hazard classification. Simulations of suppression by sprinklers are available in CFAST and FDS [Peacock, etal., 2011][McGrattan, et al., 2014]. However, the suppression algorithm in both of these models is based on a limited set of empirical data. Also, as with any simulation, both models include a set of assumptions. In the case of the suppression algorithm in CFAST a decrease in heat release rate and consequent smoke layer can be obtained with even miniscule amounts of water, i.e. sprinkler densities well below realistic sprinkler densities. In FDS, the sprinkler droplets are assumed to form immediately upon exit from the sprinkler deflector. However, the formation of droplets occurs a distance away from the sprinkler and is a function of orifice size and pressure, as summarized in Figure 7 [Ren, et al., 2008][Wu, et al., 2007]. K = 3.2 gal min -1 psi -1/2 1 The number of sprinkler models does not include additional water-based options of foam sprinklers, water spray and water mist nozzles.
Summary Fire protection engineers should be part of an integral team involved in Process Safety Management assessments. The fire protection engineer brings to the team a more thorough understanding of the principles underlying the design bases of fire protection systems. Through the input of fire protection engineers, an improved formulation of a fire safety approach can be developed which will be more effective and thus have a greater likelihood of achieving stated performance objectives. References IBC, 2012. International Building Code, Country Club Hills, IL. ASTM 2012. Fire Tests of Building Construction and Materials, ASTM E-119, American Society for Testing and Materials, West Conshohocken, PA. Beyler, C.L., 2008. Fire Hazard Calculations for Large, Open Hydrocarbon Fires, SFPE Handbook of Fire Protection Engineering, 4 th Edition, P.J. DiNenno (ed.), National Fire Protection Association, Quincy, MA. Fabian, T., and Gandhi, P., 2007. Smoke Characterization Project, Fire Protection Research Foundation. Hall, J.R., Jr., 2011. US Experience with Sprinklers and Other Automatic Fire Extinguishing Equipment, Quincy, MA: NFPA, May. McGrattan, et al., 2014. K. McGrattan, R. McDermott, C. Weinschenk, K. Overholt, S. Hostikka, and J. Floyd, Fire Dynamics Simulator Technical Reference Guide, NIST Special Publication 1018, 6 th Edition, National Institute of Standards and Technology, Gaithersburg, MD. Milke J.A., 2008a. Analytical Methods for Determining Fire Resistance of Steel Members, SFPE Handbook of Fire Protection Engineering, 4 th Edition, P.J. DiNenno (ed.), National Fire Protection Association, Quincy, MA. Milke J.A., 2008b. Smoke Management in Covered Malls and Atria, SFPE Handbook of Fire Protection Engineering, 4 th Edition, P.J. DiNenno (ed.), National Fire Protection Association, Quincy, MA Milke, et al., 2008. Milke, J.A., Mowrer, F.W., and Gandhi, P., Validation of a Smoke Detection Performance Prediction Methodology, Fire Protection Research Foundation, Quincy, MA. Milke, J.A., 1999. An Overview of Analytical Methods to Evaluate the Fire Resistance of Structural Members, J. of Structural Engineering, 125 (10), 1179-1187. NFPA 13, 2013. Standard for the Installation of Sprinkler Systems, National Fire Protection Association, Quincy, MA. NFPA 72, 2013. National Fire Alarm and Signaling Code, National Fire Protection Association, Quincy, MA.
Peacock, et al., 2011. Peacock, R.D.Forney, G.P., Reneke, P.A., Consolidated Model of Fire Growth and Smoke Transport (Version 6) Technical Reference Guide, NIST Special Publication 1026r1, National Institute of Standards and Technology, Gaithersburg, MD. Ren, et al., 2008. N. Ren, A.F. Blum, Y.H. Zheng, C. Do, and A.W. Marshall, Quantifying the Initial Spray from Fire Sprinklers, Fire Safety Science - Proceedings of the Ninth International Association of Fire Safety Science, IAFSS, Karlsruhe, Germany. Wu, et al., 2007. D. Wu, D. Guillemin, D., A.W. Marshall, A Modeling Basis for Predicting the Initial Sprinkler Spray, Fire Safety Journal, Vol. 42, pp. 283-294.