What it is and which question it answers
Evacuation modelling represents a space, its population and selected movement rules to investigate how an exit or relocation to a place of safety might unfold. A simulation runs that representation under specific assumptions. It supports comparison of alternatives, identification of congestion and examination of the assumptions that dominate the results. It does not reproduce real human behaviour infallibly or certify, on its own, that a building is safe.
Frame a question before choosing software: what changes when workstations are rearranged, how two floors interact on a staircase, or what happens when a planned route becomes unavailable. The result should describe an identifiable population and scenario. An attractive animation without input data or justification of the method provides little evidence for selecting preventive measures. The level of detail should follow the decision being considered, rather than the visual sophistication of the software.
Sources: [1]
Geometry, occupancy and study boundaries
A model needs geometry checked against the actual premises: routes, doors, stairs, obstacles, connections and destinations. Usable width matters, rather than just the dimension on an earlier drawing. Door conditions and areas that cannot be used must also be specified. Simplifying a building is legitimate when the omissions and their possible influence on the conclusion are explained.
The population should distinguish shifts, visitors, contractors, initial distribution and familiarity with the premises. On an industrial estate, the study may extend to shared exits and the assembly area, addressing the interactions that actually exist. Record the model boundary explicitly: ending at the factory door does not demonstrate that the subsequent outdoor route has been resolved or that everyone has been accounted for. A locally successful exit can still feed a problem beyond the modelled area.
Pedestrian movement and route choice
Software may represent individuals, aggregate flows or combine approaches. That choice affects spatial resolution, interactions and interpretation of results. Check how the system handles queues, crossing movements, merging into stairways and flow constraints, as well as the units and scale of imported data. A feature the program does not represent must not be reported as having been analysed.
Sending every occupant to the nearest exit is an assumption, not a prediction of behaviour. Familiarity, instructions and route availability can change the choice. Historic NIST stair studies show variation between observed situations; they do not provide one walking speed suitable for every workforce. Selected values need a documented origin and relevance to the population and scenario. Copying a default setting from a previous project can obscure a significant difference in the new building.
Time before movement also matters
Evacuation does not necessarily start when a symbol appears on a screen. Event detection, alarm transmission, recognition and decisions before movement can occupy a substantial part of the process. Seeking information, helping another person or waiting for instructions are behaviours a model may simplify or represent using distributions. Explain which approach was used and the evidence supporting it.
Fire engineering distinguishes required safe egress time, RSET, from available safe egress time, ASET, before conditions become unacceptable. Comparing them requires compatible criteria, a common time origin, consistent scenarios and justified margins. Pedestrian movement software does not, by itself, establish smoke development or tenability. That comparison belongs within a specialist assessment, rather than an isolated reading of the final second of an animation. Time components and their definitions should remain visible so that omissions or double counting can be detected.
Accessibility and assistance needs
Representing occupant diversity means more than reducing the average speed of a percentage of figures. Needs may affect alarm perception, comprehension of messages, turning space, changes of level, wayfinding and assistance. Consultation with users helps reveal assumptions that a drawing or database cannot capture. Accessible communication and the physical route are interconnected parts of the strategy.
Evacuation strategies and planned resources must be compatible with the building and applicable requirements. If the model cannot represent a wheelchair, an assistance interaction or a wait in a protected area, that limitation requires complementary analysis. Do not assign capabilities automatically from age or diagnosis. Conclusions should identify which needs were studied, using which resources, and what remains unresolved before the strategy can be accepted. A numerical output cannot compensate for an assistance arrangement that does not exist in practice.
Verifying the model and validating its use
Verification examines whether the implementation correctly executes its stated rules and calculations. Validation examines how adequately the results represent relevant phenomena for an intended use. These are related but different checks: a calculation may be programmed correctly while failing to represent a particular situation with sufficient fidelity. Supplier marketing documentation does not replace checking the case under examination.
The study lead should review available tests, their relevance to the phenomena in the project and declared limitations. Next, check construction of the model: geometry, population, connections, assignments and boundary conditions. Preserve simple checks capable of exposing disconnected doors, impossible paths or incorrect units. Adjusting parameters until a drill is reproduced does not establish validity for every future configuration. Keep the evidence used for comparison separate from the scope of the conclusions, and document any discrepancy that remains unexplained.
Sources: [1]
Scenarios, uncertainty and sensitivity
A single favourable scenario may conceal important dependencies. Selection should respond to risk and the study objective: different occupancy distributions, justified changes in movement initiation or route unavailability where relevant. It is unnecessary to simulate every imaginable combination, but explain why the chosen alternatives are considered sufficient and which situations remain outside the analysis.
Where a model contains random components, multiple runs reveal variation; the appropriate number depends on the method and stability of results, not a universal figure. Examine local congestion and unfavourable outcomes as well as the final average. Sensitivity analysis identifies inputs that change conclusions. Displaying many decimal places does not remove uncertainty in estimated occupancy or poorly understood behaviour. Report ranges and important dependencies clearly enough for the decision maker to understand what the model does and does not support.
Sources: [1]
Applied example: two buildings sharing an exit
In a hypothetical example, two businesses share a courtyard leading to an assembly point. Each building initially achieves an acceptable result in its separate study, but both release occupants onto the same outdoor segment. The team decides to examine coincident flows and visitors during a busy working day. The case and its inputs are illustrative, rather than findings from a particular installation.
Geometry is checked and alternatives for layout and organisation are compared, including planned assistance and actual site constraints. The decision does not rely solely on reducing a total time: it must also avoid creating congestion elsewhere or transferring risk to a vehicle entrance. The selected option is translated into instructions, assigned responsibilities and site checks. The model supports the reasoning behind the decision; implementation must be checked physically. Any remaining uncertainty should travel with the recommendation into the action plan.
Relationship with occupational safety and drills
In occupational safety and health, this technique helps examine changes in layout, occupancy and emergency organisation. Article 20 of Spain’s Occupational Risk Prevention Act requires emergency analysis, appropriate personnel and resources, and checks within its terms; it does not require computer simulation in every workplace. The applicable fire and self-protection framework depends on use, characteristics and relevant legislation, including territorial requirements. Spanish references retain their jurisdiction in this English article.
A drill allows observation of real organisational arrangements; a model explores scenarios that should not be physically created. They complement each other when observable facts, incidents and assumptions are compared without exposing people to danger merely to collect data. A study does not automatically authorise fewer exits, removal of safeguards or greater occupancy. Alternative solutions require the relevant justification and procedures, together with training and effective implementation.
Since October 2026, Spain’s national self-protection framework refers to the new guideline published by Order INT/1068/2026. The agreement ends application of RD 393/2007 from publication on 10 October and takes effect the following day. Its adaptation arrangements distinguish plans already in force from other cases and must be checked alongside regional and sector-specific requirements.
Documentation, decisions and useful indicators
A reproducible study identifies its objective, software version, technical lead, date, original drawings, population, parameters, sources, scenarios and limitations. It preserves the files needed to repeat the analysis and links every conclusion to a measure, an owner and a verification step. When site layout or occupancy arrangements change, assess whether the study still represents the existing situation.
Suggested management indicators include the share of assumptions checked on site, resolved accessibility issues and verified improvement actions. These are monitoring proposals, not legal thresholds or universal predictors of casualties. Evacuation time is one output among several and should be communicated with its scenario and uncertainty. Preventive value appears when findings become usable routes, understandable instructions and an organisation capable of implementing them. Keeping the record current makes later reviews more reliable than relying on an unexplained animation saved years earlier.
Frequently asked questions
Does it replace an evacuation plan?
No. It informs selected decisions; the plan organises resources, responsibilities and instructions and must be implemented.
Does a fast result demonstrate safety?
No. Scenarios, safety conditions, accessibility, uncertainty and model limitations must also be considered.
How many runs are needed?
That depends on the model, its random components and result stability. No universal number suits every study.
Can a drill be used to check a model?
Yes, using observable data and documented conditions. Reproducing one drill does not automatically validate other scenarios.
