Extended travel distance in residential apartment buildings - A comparative risk model Emanuel Grunnesjö Department of Fire Safety Engineering and System Safety Lund University, Sweden Brandteknik och Riskhantering Lunds Tekniska Högskola Lunds Universitet Report 5439 Lund 2014 Thesis has been funded by Olsson Fire & Risk ii Extended travel distance in residential apartment buildings - A comparative risk model Emanuel Grunnesjö Lund 2014 iii Title Extended travel distance in residential apartment building - A comparative risk model Author Emanuel Grunnesjö Report 5439 ISSN: 1402-3504 ISRN: LUTVDG/TVBB-5439—SE Number of pages 158 (including appendices) Illustrations where not specified are made by: Emanuel Grunnesjö Keywords Extended travel-distance, risk model, residential building, performance-based design, alternative solution, apartments, fire safety engineering, fire risk analysis. Abstract Prescriptive building codes have been used for many years and created a common sense regarding what an adequate fire safety design and an acceptable level of risk is. Introducing performance based building codes brought architectural flexibility, more bespoke design, and cost-benefit analysis into fire safety design previously not possible in a number of different building types. Alternative Solutions are now common place in almost every residential apartment building in Australia and often includes a number of typical non-compliances such as extended travel distances between apartments and exits. However, it is not that easy to accurately evaluate the level of safety provided by a fire safety design in a residential building. The reasons for that are many such as the complexity of the system (both active and passive systems), the different aspects affected during a fire (all sub-systems in the IFEG) but also the harsh commercial conditions during a residential building project. The objective of this thesis was therefore to investigate if (or to what degree) the inclusions of passive and active systems compensate for the increased risk due to the extended travel distance for any given (non-compliant) building solution. This was done through the development of a comparative risk model for residential buildings with a focus on the impact of extended travel distance on occupant life safety. The developed risk model indicates that the main contribution to the risk to occupants when travel distances are increased in residential buildings can be derived from the increased number of apartments. By adding more apartments connected to a corridor, the probability of a fire to occur on that level, as well as the consequence for all scenarios when smoke leaks into the corridor are increased. Therefore the relative risk nearly increases as the square of the number of apartments increase. The difference in travel time due to the extended travel distance was shown to not have any measurable impact on the risk level for corridor lengths between 7 m and 25 m in length. © Copyright: Brandteknik och Riskhantering, Lunds tekniska högskola, Lunds universitet, Lund 2014. Brandteknik och Riskhantering Department of Fire Safety Engineering Lunds tekniska högskola and Systems Safety Lunds universitet Lund University Box 118 P.O. Box 118 221 00 Lund SE-221 00 Lund [email protected] Sweden http://www.brand.lth.se www.brand.lth.se Telefon: 046 - 222 73 60 http:// www.brand.lth.se/english Telefax: 046 - 222 46 12 Telephone: +46 46 222 73 60 Fax: +46 46 222 46 12 iv Extended travel distance in residential apartment buildings – A comparative risk model Acknowledgements Between the apartment in Manly and the office in Neutral Bay, Sydney there are a few big hills to climb when you are riding on your bicycle to get to the office. Some are so steep that you have to cycle so slow that you feel like you are almost standing still and not progressing. When writing this thesis and developing the risk model there have been challenges of the same order when I had to slow down to a pace where I did not even know if I was progressing. The thesis is now done and I want to thank everyone who has been involved for making it possible in the end, and would especially want to acknowledge: Per Olsson – for coming up with the initial idea and inspiring me through brainstorming sessions and good encouragement. Carl Voss – for convincing me to stay in Australia and making it possible to co- operate with Olsson Fire & Risk when undertaking my dissertation thesis. Mark McDaid – for supervising me throughout the process and taking time for discussing ideas and giving me both good information and inspiration. Jonathan Barnett – for helping me to realise what a reasonable scope is and giving me valuable comments throughout the work with my thesis. Håkan Frantzich – for being my remote supervisor and despite the geographical distance giving me good food for thought and ensuring a great quality of the thesis. My hope is that this work can bring some light and guidance on how to evaluate fire safety in apartment buildings from a more holistic risk perspective and that you all enjoy reading it. Emanuel Grunnesjö – Sydney, 2014-04-10. v 0 Summary Summary The objective of this thesis is to investigate if (or to what degree) the inclusion of passive and active systems compensates for the increased risk due to the extended travel distance for any given (non-compliant) building solution. This was done through the development of a comparative risk model for residential buildings with a focus on the impact of extended travel distance on occupant life safety. The model was developed using an event tree approach together with an ASET/RSET analysis (comparing Available Safe Egress Time with Required Safe Egress Time) and included 240 scenarios. Probabilities were determined from data, fault tree analyses and engineering judgement. Consequences were determined through modified ASET/RSET analyses using fire modelling in Fire Dynamics Simulator (FDS), previous studies and engineering judgement. The model was configured in an Excel spread sheet by using the add-in software PrecisionTree and @Risk from Palisade Tools. For each evaluated building solution the model generates an average risk value estimating the expected number of fatalities per year at one typical level of the apartment building and a curve in an F/N diagram. The values generated by the risk model should not be interpreted as accurate estimations of the absolute risk, but as comparative values of the risk to occupants based on detailed analyses for several scenarios including both fires in the corridor and the apartments with safety systems both working and failing. The result generated by the model was shown to be sensitive to a number of variables, and these fixed values variables were therefore replaced by probability distributions. By iterating the result of the model using a Monte Carlo simulation with values from the applied distributions it was shown that the generated risk can be characterised as a triangular probability distribution for less complex buildings, and as natural-like distributions for more complex buildings. The model can be used for comparing building solutions and for quantifying the change in risk level when different fire safety measures are provided or removed. This is done by dividing one generated risk value with the other and calculating a risk ratio. The model cannot verify whether the risk level in the evaluated building solution is acceptable or not. The model can only tell how large the risk level is compared to the risk level for another building solution. If the model is used to compare an alternative building solution with a prescriptive compliant building solution it appears that a safety factor should be applied. Alternatively the probability distributions from the uncertainty analysis can be used from which the risk ratio at different percentiles can be obtained. The main contribution to the risk to occupants when travel distances are increased in residential buildings is due to the increased number of apartments. By adding apartments in the corridor, the probability of a fire to occur on that level, as well as vi Extended travel distance in residential apartment buildings – A comparative risk model the consequence for all scenarios when smoke leaks into the corridor are increased. That means the relative risk nearly increases as the square of the factor of apartment increase. The difference in travel time due to the extended travel distance was shown to not have any measurable impact on the risk level for corridor lengths between 7 m and 25 m. If corridors would be increased significantly longer than that, the travel time could have an impact on the risk level. But even then it is likely other factors such as number of apartments, provisions for fire brigade intervention and risk for queue in the stair would have a larger impact on the risk level than the actual travel time. The model can show how the risk is changed when different fire safety measures are provided or removed from a building solution. The impact of adding or removing fire safety measures is different depending on the building solution. For example the effect on the relative risk when providing a sprinkler system is not the same if it is provided in a building that has a smoke detection system installed as in a building that does not have smoke detection system installed. The impact of changing building features in a building with a corridor length of 9 m, a maximum travel distance of 6 m and smoke alarms installed in each apartment is shown in the table below. Change of fire safety measure Impact on the risk level 18 m additional travel distance for all no impact apartments Doubling the number of apartments in the + 328 % corridor Providing a heat detection system - 77 % Providing a smoke detection system - 77 % Providing a sprinkler system - 89 % Removing fire ratings between the + 29 % (presuming smoke seals are apartments and the corridor provided to the apartment doors) Removing self-closers from apartment + 7 % (presuming smoke seals are doors provided to the apartment doors) Providing smoke seals to apartment doors - 22 % Removing smoke alarms from apartments + 14 % Providing a mechanical corridor - 88 % ventilation system (including smoke detection to activate system) The analysis showed that deterministic evaluations are very depending on what scenario is evaluated, and even if the consequences have been lowered in the Alternative Solution, the increased probability for a fire to occur could generate a vii 0 Summary higher average risk. If the analysis is limited to the consequences of fires the result of it could show that the safety level has been improved while in reality the risk to occupants within the building actually has been increased due to the higher probability of fire. One way of decreasing the level of uncertainty is to conduct comparative assessments. Using comparative assessments means that assumptions and limitations, even if they are wrong, applies to both compared solutions and therefore don’t affect the results in any direction. However a comparative assessment does not assure its validity. The developed model showed that the scenario chosen for evaluation is unquestionably essential to the result. But even more importantly, the development of the model also revealed that deterministic ASET/RSET analyses may sometimes not be appropriate for evaluating fire safety in extended corridors. The conclusion from this work is that evaluating fire safety in residential buildings should always be a probabilistic risk process considering both the likelihood and consequences of several credible scenarios when safety systems are both working and failing. To conduct a quantitative risk analysis for every residential apartment building project which includes deviations from the prescriptive requirements of the building code is not feasible in the economic environment they are developed in. The model in this thesis was therefore developed to show how the process can be standardised and made more efficient. However, the model can still be improved to achieve more accurate results. The model was also shown to be largely depending on choices of tenability criteria, evaluated fire scenarios etc. which means the results generated by the model must always be interpreted by the user and cannot be relied on solely without engineering judgement evaluating the reasonableness of the findings. viii Extended travel distance in residential apartment buildings – A comparative risk model Terminology and Definitions Definitions Alternative Solution - A building solution that satisfies the Performance Requirements by other means than complying with prescriptive requirements. Available Safe Egress Time – The time from fire ignition until escape no longer is possible for a specific scenario. Deemed-to-Satisfy – Prescriptive provisions in the BCA. Detection scenario – A given set of circumstances which generates a specific detection time. Deterministic – An approach where the input parameters for the scenario are pre- determined, which will result in a single outcome. Effective height – The height to the floor of the topmost storey from the floor of the lowest storey providing direct egress to a road or open space. Event tree - A logic diagram which presents the probabilities for different outcomes of an initial event depending on the following sequence of events. Fire Resistance Level – The grading period in minutes determined for structural adequacy, integrity and insulation. Fire scenario – A fire scenario is, in this thesis, a specific design fire with all parameters affecting the fire determined. Flashover - The rapid transition to a state of total surface involvement in a fire of combustible material within an enclosure. F/N Curve (Frequency – Number curve) – A statistical measure to present societal risk. The curve shows the frequency of accidents causing x or more fatalities per year. Fractional Effective Dose – The cumulative effect of exposure to multiple narcotic gases. Individual risk – the risk to which any particular occupant is subjected at on the location defined by the scenario. If an occupant is inside a building, he or she will be subjected to a risk in terms of the hazard frequency (Frantzich, 1998). Monte Carlo analysis – A method to perform random sampling to compute predictions of mathematical models. ix 0 Terminology and Definitions Occupant Warning System – Any system that provide occupants with a warning signal of fire hazard. Probabilistic – An approach where both the probabilities and the consequences of different scenarios are evaluated. Response scenario – A given set of circumstances which determines how quickly the occupants in the building decide to evacuate. Is depending on how the occupants become aware of a fire. Required Safe Egress Time – The time required from when the fire is initiated until the occupant has evacuated. The sum of the detection time, response time and movement time. Societal risk – The risk to the society for example characterised as expected number of fatalities per year. Sole Occupancy Unit – Apartment. Abbreviations AOF Apartment Of Fire origin AFAC Australian Fire Authority Council ALARP As Low As Reasonably Practicable ASET Available Safe Egress Time BCA Building Code of Australia CO Carbon Monoxide DtS Deemed-to-Satisfy (prescriptive provision) ERL Expected Risk to Life FED Fractional Effective Dose FRL Fire Resistance Level IEC International Electrotechnical Commission IFEG International Fire Engineering Guidelines IR Individual Risk PDF Probability Density Function x