Practicalities and Limitations of Coupling FDS with Evacuation Software

Transcription

1 ABSTRACT Practicalities and Limitations of Coupling FDS with Evacuation Software Daniel Rådemar, WSP, Sweden Daniel Blixt, WSP, Sweden Brecht Debrouwere, WSP, Sweden Björn Grybäck Melin, WSP, Sweden Andrew Purchase, WSP, Sweden Computational Fluid Dynamics (CFD) allows the available safe egress time (ASET) to be calculated for a design by modelling the fire and smoke movement. An assessment of evacuation movement then needs to be made to estimate the required safe egress time (RSET). Many evacuation software exist, but very few allow a direct coupling of the CFD and occupant movement data. This coupling is advantageous as it allows for visualization and quantification of tenability conditions. However, coupling two different software types is not always straightforward and is still a developing feature in evacuation software. The emphasis of this paper is on the coupling of evacuation software with Fire Dynamics Simulator (FDS). Pathfinder, STEPS and Evac software were considered. An example geometry based on an underground rail station was used to compare the setup, limitations, advantages and outcomes using the different evacuation software. The Fractional Effective Dose (FED) was used to calculate the RSET for this investigation. This paper provides a reference for other practitioners of specific issues to consider when setting up their own models. It also demonstrates some of the competencies required when coupling FDS with evacuation software that may not be obvious to less experienced practitioners. BACKGROUND The process of designing building fire protection varies from country to country. In Sweden, performance based design (PBD) methods can be used when deemed-to-satisfy building code (BBR, [1]) designs are not applied. PBD methods are often desirable due to economic, architectural or practical considerations, but often require more detailed calculations to be undertaken with advanced numerical tools such as CFD and evacuation modelling. In Sweden, BBRAD 3 [2] gives the framework for how PBD methods should be carried out to comply with BBR. For typical building types this includes guidance around the acceptance criteria, design fire scenarios and evacuation parameters such as pre-movement times and movement speeds. However, the building code assumes the calculations are performed correctly and gives no guidance on how the calculations should be executed. Fortunately, there is an industry guideline on how to perform CFD calculations which has been prepared by BIV [3], the Swedish chapter of SFPE. This guideline provides recommendations regarding the CFD model setup and self-checks that should be performed by the practitioner. However, there is not an equivalent guideline for evacuation calculations that exists in Sweden, and no guidance on how to couple the results from a CFD model to an evacuation model. An additional issue also exists with the application of BBR and BBRAD 3 to non-standard building types such as transport infrastructure. For example, tunnels are not classified as buildings but as large structures. This implies that BBR and BBRAD 3 are not applicable. This puts greater responsibility on the practitioner to develop inputs that are suitable for their particular project. In these situations, it could also mean that acceptance criteria based on simple tenability limits (e.g. visibility) are too onerous and some alternative criteria such as toxicity exposure is adopted. However, criteria based on exposure requires a dosage to be accumulated on agents which is difficult to do unless there is coupling between the CFD and evacuation models. This is because the agent s dosage must be traced in time and space. So the situation exists where there is little guidance around how evacuation calculations should be undertaken and even less about how CFD and evacuation software should be coupled to enable calculation of dosage based tenability. This situation is further compounded by the different evacuation software and the variation in methodologies that these software use. While FDS has been largely adopted as the industry standard for fire and smoke modelling, it is difficult to say if the same exists for

2 evacuation software. This could result in different outcomes for similar situations, and there is an uncertainty if this is real or an artefact of the different modelling methodology and how it is applied by the practitioner. While it would be interesting to compare outcomes for different evacuation software, it is rarely the case that this is possible within time and cost constraints on actual projects. The purpose of this paper is therefore to couple CFD output from FDS to different evacuation software and compare the outcome. The intent is not to provide a guideline on how this should be undertaken, but instead show how this can be achieved for different software and some of the pitfalls in doing this coupling. In doing this, the intent is to also demonstrate some of the competencies required when undertaking this type of coupling. Pathfinder, STEPS and Evac were used for the comparison as these software allow for FDS coupling and were available to the authors. METHODOLOGY To undertake the comparison of different evacuation software a four-step methodology was used: 1. Verify coupling of FDS with each evacuation software for a simple test model; 2. Develop a complex and interesting scenario to study and model this in FDS; 3. Couple the output from the FDS model with each evacuation software; and 4. Compare the outputs from each evacuation software. The total evacuation time and FED are used as metrics for comparison. The calculation of the FED is as described in the Pathfinder [4] and Evac user manuals [5]. As defined by ISO [6], an accumulated FED of 1.0 corresponds to a log-normal distribution of responses, with statistically 50% of the population expected to experience compromised tenability. A threshold criterion of accumulated FED > 0.3 is a somewhat typical design criteria and correlates to approximately 11% of the population being statistically susceptible to compromised tenability. It should be noted that the geometries used in this study are fictitious and not based on any existing or planned project. All scenarios and inputs are nominal for comparative purpose only, and although informed by project experience, are not related to any project. The geometry and scenarios used were intended to provide reduced tenability to produce interesting results. The authors provide no comment on the suitability of the acceptance criteria or acceptability of outcomes as these are merely for comparative purposes. Verifying FDS and Evacuation Coupling The FDS coupling methodology and calculation of the FED varies between Pathfinder, STEPS and Evac, and this is discussed later in this paper. Given the variation in methodologies, a simple test case was used to confirm the software setups and compare outcomes without the complexities of a detailed evacuation scenario. It should be noted that this test case was not intended to validate the models but merely to verify that the coupling with FDS and the FED calculation methodology was correct. The test case consisted of a simple 400 m 2 room with a 5 MW fast t-squared growth rate fire located in the corner. A total of 300 agents were evenly distributed around the room. Two scenarios were modelled including (1) agents exposed to fire and smoke for 600 seconds without any movement and (2) same as scenario 1, but agents move out via an unconstrained path after 500 seconds. The first scenario is intended to test FED accumulation by removing any differences in the way each software models agent movement. This allowed direct comparison of the FED calculation in Pathfinder and STEPS. However, Evac describes the accumulated FED when passing through a door or other mesh boundary. The second scenario allowed direct comparison of all three evacuation software and the unconstrained exit paths removes differences regarding movement and behavior at doors. Figure 1 shows the results from the two scenarios for the test case geometry. The FED distribution is sorted from highest to lowest for the 300 agents modelled. For both scenarios there is good agreement between the three models. Although Evac could not easily output FED for the first scenario, this software can display the FED visually for each occupant and this qualitatively showed good agreement with STEPS and Pathfinder. These outcomes gave confidence that the coupling between FDS and each evacuation software was correct and that the calculation of the FED was consistent between the software.

3 FDS Model and Evacuation Scenario Figure 1 Test Case Results An underground metro station was selected as a complex and interesting geometry. CFD modeling was undertaken with FDS version [7]. The modelling setup was generally consistent with the Swedish BIV guidelines for CFD modelling [3]. The fire was modeled as a fast t-squared growth rate that was kept constant at 10 MW after reaching this peak heat release rate. A soot yield of 0.1 kg/kg and a CO yield of 0.1 kg/kg were used, along with a heat of combustion of 20 MJ/kg. The FDS model was setup to save the output required for each evacuation software, with all output recorded at 10 second intervals. Figure 2 shows the CFD geometry. The tunnel stubs closest to the fire were modelled with a nominal applied pressure to simulate a residual train piston effect that disturbs the smoke layer. All other boundaries were modelled as open boundaries. No smoke exhaust was operating during the evacuation phase and there were limited smoke barriers to prevent smoke spreading into the concourse levels. East Concourse Fire Location Train 1 West Concourse Train 2 Figure 2 FDS Model Geometry (Ceiling Removed) An evacuation scenario was developed and implemented into each evacuation software. This scenario considered the following placement of agents (evenly distributed) and their pre-movement times: agents on the platform that start evacuating 1 minute after train 1 arrives agents on train 1 that arrives at the platform 2 minutes after the fire starts agents on train 2 that arrives at the platform 4 minutes after the fire starts on train 1 Both sides of the platform were equipped with stairs and escalators. The escalator on the eastern side closest to the fire train (Train 1) were assumed to be unavailable due to the proximity of the fire, and the adjacent escalator was running upwards towards the east concourse. On the western side one of the escalators was assumed to be unavailable due to maintenance, while the other one was running upwards. An elevator was provided on both ends of the platform to cater for the needs of the mobility impaired. Movement characteristics for the horizontal planes and stairs/escalators were guided by BBRAD 3 [2], which also requires that 1% of the agents are mobility impaired with reduced movement speeds.

4 Pathfinder Software Pathfinder is an agent-based movement simulator developed by Thunderhead Engineering [4]. The software uses a continuous model, where the simulated movement of the agents is calculated over a triangulated mesh. Pathfinder version was used for this study. PLOT3D data of the volumetric fractions of CO, CO 2 and O 2 must be output by FDS for Pathfinder to calculate the accumulated FED for each agent. Slice files can also be imported into Pathfinder for visualization. Neither the slice files nor PLOT3D data affects agent movement for the version of Pathfinder used. It is understood this feature may be included in future releases of Pathfinder. STEPS Software STEPS is an agent-based pedestrian movement tool developed by Mott Macdonald [8]. STEPS version 5.4 is a grid-based model that consists of pre-defined square cells that enable the movement of agents to be simulated. Only one agent can occupy a grid-cell at any time. The impact of reduced visibility on walking speed can be accounted for by using a correlation of extinction coefficient and walking speed. In STEPS, dosages can be accumulated based on FDS slice file data. This means that FDS must output slice file data for the volumetric fractions of CO, CO 2 and O 2 to calculate the FED, and the extinction coefficient to enable movement speed to be reduced with visibility. These slice files must be at the correct height for assessing tenability and on each movement plane. For this study a spreadsheet was used to calculate the FED from the output dosages. While this may appear to be a disadvantage over the other software, it does provide more flexibility in the calculation of the FED to include other components (e.g. HCN), as well as the possibility to calculate other exposure data such as temperature. Evac Software Evac is an evacuation simulation module for FDS developed and maintained by VTT [5]. The main feature of the EVAC model is the integration with FDS. The combined fire and evacuation software allows the interaction of fire and smoke on evacuating agents to be modelled. The geometry is common between FDS and Evac, although specific additions need to be made for an Evac model. In Evac each agent is treated as a separate entity and their movement is then simulated on 2D planes defined by the user. Interpersonal actions and reactions are defined in different input parameters. The forces acting on the agents consist of both physical forces, such as contact forces and psychological forces exerted by the environment and other agents. The social force model used for the movement algorithm is defined in the Evac manual [5]. The calculation of the FED in Evac is integrated with the FDS model and requires less user defined output and manual manipulation than for the other software used. RESULTS The total evacuation time (Figure 3) and FED (Figure 4) are used to compare evacuation software outcomes. It should be understood that the outcomes discussed are specific to the scenarios modelled. The first conclusion is that outcomes for Pathfinder and STEPS converge if no walking speed reduction factor is used in STEPS. A second conclusion is that the walking speed reduction has a significant influence on the total evacuation time of the agents. Little difference was observed between the walking speed reduction due to the pre-defined irritant and non-irritant smoke correlations used in STEPS. The Evac models show some interesting conclusions. With a walking speed reduction (default setup) there is a significant variation to the other software, however, when this is removed, there is much better convergence with the other models that also have no walking speed reduction. For the Evac model with a walking speed reduction there are agents with a total evacuation time of zero. This means that these agents were unable to leave the domain because their FED exposure reached unity. These agents also have the effect of hindering or blocking other agents during their evacuation. Another phenomenon observed with Evac is grouping behaviour which is an artefact of the social model used. The total evacuation time is somewhat meaningless without a tenability criteria for comparison. For this study, the FED was used to compare outcomes. Assuming a nominal threshold criteria of FED = 0.3, the outcomes are significantly different. With Pathfinder and STEPS (without walking speed reduction) the acceptance criteria would be satisfied for all agents. This, however, is not the case for the other

5 software variations which would have varying number of agents that exceed the threshold criteria. Outcomes are worst for the Evac software using the default social model. The interesting result here is that outcomes and acceptance of a design solution could be very much dependent on the model inputs and also the software used. The FDS coupling methodology also affects the FED outcome and this is discussed in the following section. Figure 3 Total Evacuation Time OTHER CONSIDERATIONS Figure 4 FED Prediction The intent of this paper is not to suggest that one evacuation software is better than the other as each has its own strengths and weaknesses in terms of the evacuation dynamics and coupling with FDS. Table 1 summarizes some possible advantages and limitations of each evacuation software for the purposes of this study. The limitations highlight some of the potential issues when it comes to coupling the software with FDS and that should be looked for when interpreting outcomes. The results presented above have been produced to rectify or post-process out these issues where possible. However, some of these issues are a inherit limitation of the software. For example, Pathfinder uses PLOT3D data which enables the FED to be accumulated in 3D space. However, STEPS uses slice data which means accumulating FED during vertical travel is much more difficult to calculate. Like any modelling effort, a level of competence is required from the practitioner when it comes to undertaking evacuation calculations, and even more so when it comes to coupling evacuation software with FDS. The FDS model must be setup with the evacuation model in mind, and special measures taken to ensure that FDS specific considerations (e.g. geometry simplifications, mesh locations) do not adversely impact the evacuation calculation. If these considerations are unavoidable, the practitioner must undertake post-processing and self-checks to ensure the outcomes are not an artefact of the model setup and coupling with FDS. Some of the limitations noted in Table 1 would be obvious in the results. Others are subtle and require a through probing of the output. This can become time-consuming with potentially thousands of agents in a model, so the practitioner needs to develop methods and potentially their own programs to process these large datasets to find and rectify potential issues.

6 Table 1 Advantages and limitations of each evacuation software Advantages Limitations Pathfinder Intuitive user interface with the option to import FDS geometry. Good visualization and support for complex geometries and evacuation scenarios. PLOT3D data enables FED to be calculated in simple and complex geometries rather than being limited to slice planes. No current support for complex agent interactions and behaviors (e.g. smoke/movement correlations). PLOT3D files require more data storage than slice files. If FDS/Pathfinder geometries are not well coordinated then agent FEDs can be distorted. If an agent moves into a region without PLOT3D data then the O2 concentration goes to zero resulting in an erroneously high FED. A separate CSV file is output for every occupant when calculating the FED resulting in many files. Scripts (e.g. Python) were used to process this data. STEPS Good visualization and support for complex geometries and evacuation scenarios. Ability to have user defined correlations for walking speed reduction. Allows flexibility in calculation of dosage based tenability criteria such as FED. Time-consuming to setup slice planes in complex geometries. Time-consuming to process dosage data for many agents with spreadsheets. Scripts (e.g. Python) are more efficient. Difficult to accumulate dosage on non-horizontal planes (e.g. stairs). This can distort the calculation. If FDS/STEPS geometries are not well coordinated then agent FEDs can be distorted. Similar to Pathfinder. Dosages across FDS mesh/slice boundaries can be distorted (counted twice). Need to consider in FDS model setup. Dosages can be distorted if a large output time-step is used (e.g. when people enter/leave the domain). Evac Seamless integration with FDS. Relatively short setup time as the EVAC model is largely the same as the FDS model. Distinctive/ complex group behavior when compared to the simpler models in Pathfinder and STEPS. No license costs and easily automated/scripted. Poor visualization compared to commercial software packages. No graphical user interface for model development. Limited support for complex environments (e.g. elevators, moving trains). Limited technical support, mainly forums. CONCLUSIONS This paper has compared different evacuation software and the capability of coupling these with FDS. For the scenario considered, there was significant variations in the outcome. Applied to an actual project, this could mean the difference between a design being acceptable or not. This variation was in part due to the different modelling and FDS coupling methodologies used by each software package. When it comes to coupling FDS with evacuation software, care must be taken to ensure that dosage based values are calculated correctly. Misalignment in FDS and evacuation software geometries and other issues can results in erroneous FED values. The user must take care when setting up the geometry and look for these issues in the post-processing of data and self-checking of model outcomes. REFERENCES 1. Boverket, "Building regulations, Safety in case of fire, BFS 2011:26 with changes including BFS 2015:3, BBR 24" Boverket, "Building regulations, General recommendations on analytical design of fire safety strategy, BBRAD 3" Boverket, Karlskrona, BIV, "Stöd för tillämpning av CFD-modeller," Thunderhead Engineering, "Pathfinder User Manual - Pathfinder 2017," VTT Technical Research Centre of Finland, "Fire Dynamics Simulator with Evacuation: FDS+Evac, Technical Reference and User s Guide," International Organization for Standardization, "ISO Life-threating components of fire Guidelines for the estimation of time to compromised tenability in fires," K. McGrattan, S. Hostikka, R. McDermott, J. Floyd, C. Weinschenk and K. Overholt, "IST Special Publication 1019: Fire Dynamics Simulator User s Guide, 6th ed.," Mott MacDonald, "STEPS User Manual," 2017.