Verification

Each test checks one part of pyFDS-Evac against a value computed without the code under test: a hand calculation, a published table, or FDS’s own output. Everything here is verification (does the code solve its equations correctly?), not validation against observed evacuations.

A test works at one of three levels:

Level	What runs	What it catches
Equation	one model function, no simulation	a wrong formula or constant
Coupled	a full pyFDS-Evac run on a synthetic field, no FDS	wrong wiring: the field is not sampled at the agent, or the result does not reach the agent
FDS case	a full run on a small FDS case (`assets/`)	wrong reading of FDS output: slice height, units, species names

The tests

Click a test for its page: what it checks, the equation, the setup, the expected value, and the simulated result. Tests without a page yet link to their test file.

Component	Test	Level	What it checks	Status
Gas FED	CO dose in a uniform room	FDS case	FED reaches 1 at the hand-calculated time; the fraction of agents incapacitated follows the log-normal threshold (opt-in probabilistic mode)	passes: dose and deterministic stop; the probabilistic stop pooled over 10 seeds (1000 agents) stays in the 95 % band (D = 0.023 against 0.043)
Gas FED	FDS `FED_FIC` case	Equation	FED and FIC in the four zones of FDS’s own verification case	passes
Gas FED	S1 corridor	Coupled	an agent in constant CO stops within one FED update of t*	passes
Gas FED	ISO 20414 Test 19: incapacitation by toxic gases	FDS case	four gas mixtures separating the CO, CO₂ and O₂ terms: each occupant stops at the next update after the hand-calculated time, as FDS’s own FED device does	passes for CO, CO₂, O₂; HCN, NOₓ, irritants not yet (#257)
Heat FED	Heat dose in a uniform room	FDS case	heat FED at 100, 150 and 200 °C: every agent stops at the update where the hand sum on the FDS temperature reaches 1	passes with `--heat-clothing unclothed`; checks the pipeline, not Eq. 63.44 itself; the default ISO Eq. (9) is an expected row only (#307)
Heat FED	S6 heat	Coupled	heat incapacitation at t*, cause recorded as `heat`	passes; checks the wiring of the default ISO Eq. (9), not the law itself
Heat FED	SFPE tables	Equation	Eq. 63.44 against Table 63.20 (convective rows): within the assumed ±0.5 min rounding of each whole-minute entry, shorter at 100–140 °C; never longer than Table 63.17’s reported dry-air times; the default ISO Eq. (9) recorded at 1.8–3.0 times Table 63.20	passes (#219)
Heat FED	Heat endpoints	Coupled	`--heat-endpoint`: on the temperature history of SFPE Table 63.21 the tolerance dose reaches 1 in the fourth minute; at 150 °C the fatal dose crosses at the Eq. 63.47 time; endpoint, 205 °C and non-finite flag and unknown humidity reach the outputs; with `--heat-clothing unclothed` the FED history and manifest match a baseline from main (apart from `heat_clothing`); without an endpoint the run crosses at ISO Eq. (9)	passes
Heat FED	Total flux (test)	Coupled	`--heat-fed-method total-flux` against the hand formula of Eqs. 63.49 and 63.43: dense smoke at 300 °C (tolerance, ε 0.9, h 8) crosses at 5.8 s; clear air at 100 °C (tolerance, ε 0.05, h 8) at 190 s on convection alone; smoke at 200 °C (fatal, ε 0.5, h 8) has a radiant term of 1.17 kW/m², below the 2.5 kW/m² threshold; outputs carry `heat_flux_kw_m2` and the method	passes (#223)
Heat FED	Hot layer (test)	Coupled	`--heat-regime layer`: head at 120 °C below a 300 °C layer (φ 0.5, ε_L 0.9; layer term 2.52 kW/m², just above the threshold) crosses at the hand time; distinct from the smoke regime and from double counting; layer column in the outputs	passes; synthetic fields only; no FDS case runs the layer regime yet (#308); the #224 radiometer decks are FDS-only reference data
Heat FED	INTEGRATED INTENSITY (test)	Coupled + FDS case	radiant-only crossing at the hand time (q = 3.98 kW/m², fatal after 159.7 s) and its scaling with f; a committed FDS 6.10.1 case read at the head height against FDS’s own devices; f = 1/4 against the FDS skin gauge in the #224 uniform room	passes; radiometer part skipped without the sciebo data (#221)
Heat FED	Radiant threshold	Equation	ISO 2.5 kW/m² on the radiant term of each source (gas, layer, U): 2.4 counts as zero, 2.5 counts, a negative term counts as zero, NaN still voids the dose; convection unchanged; `heat_flux_kw_m2` stays physical; manifest records the threshold	passes
Heat FED	Radiant and flame-pass references	Equation	Eq. 63.45 against Table 63.21; Eq. 63.43 against Table 63.20’s radiant rows; dose of an agent walking past a flame	reference only; they call no pyFDS-Evac code (#223)
Heat FED	Heat radiometer reference decks	FDS case	FDS only, no pyFDS-Evac run: where a skin gauge’s incident flux q sits between U/4 and U under a hot layer, in a uniform room and beside a burner	reference data; no run reads it; the uniform room is the expected value of the f = 1/4 check above
Smoke speed	S2 corridor	Coupled	applied speed factor equals the Frantzich–Nilsson and Fridolf laws	passes
Smoke speed	ISO 20414 Test 18: walking speed in smoke	FDS case	egress time in the ISO corridor at K = 0.5–10 1/m and from FDS: equals the clear-air time scaled by 1/f(K)	passes
Irritants	FIC vs FED speed	Equation	on a short egress FIC slows agents at once while FED stays far below 1	passes
Pre-movement	Distributions	Equation	sampled moments of the gamma, log-normal, Weibull and uniform presets; the constant 10 s default is not checked	passes (#309)
Slice sampling	Nearest FDS node	Equation	node- and cell-centred slices read at the right index	passes
Slice sampling	Slice height	FDS case	the gas FED reads the horizontal slice nearest 1.6 m in a room with CO in three layers	passes
Slice sampling	Sign legibility slice height	FDS case	fdsvismap reads the horizontal extinction slice nearest 1.6 m, never a vertical slice listed first, and the same slice as the smoke sampler	passes
Slice sampling	Gradient field	Coupled	the field is sampled at the agent’s current position	missing (#24)
Sign legibility	Unit tests	Equation	reading distance, 30 m cap, viewing angle	passes
Sign legibility	S3 visibility gating	Coupled	a sign is acquired only when legible through smoke	missing (#22)
Cognitive map	Familiarity: full map vs discovered map	Coupled	what each agent knows, learns and walks, against sight lines computed from the plan	fails (tour, convergence): criteria 1–3 and 5 pass; 4 agents skip CP1, which criterion 4 does not allow for; the discovery egress time is not grid-converged (#168, #250)
Cognitive map	S5 staff vs visitor	Coupled	the cognitive map is used in the run loop	missing (#23)
Routing	S4 T-junction	Coupled	smoke on one branch makes every agent switch exit	passes
Routing	Golden rerouting decisions	Equation	route costs and decisions unchanged (regression, not verification)	passes

Every field is read from the horizontal slice nearest --smoke-slice-height (1.6 m). The FDS-case checks of that rule, per quantity: gas, the slice-height test above; sign legibility, the sign-legibility slice-height test; temperature and U, the committed case of the INTEGRATED INTENSITY test; extinction for walking speed, only the weak check of ISO Test 18, whose slices differ by very little (#259).

Movement itself (walking, collision avoidance, flow through doors) is JuPedSim’s and is not retested here; see Status against ISO 20414.

Running the tests

uv run pytest tests/verification -m "not slow"   # fast suite
uv run pytest tests/verification                  # incl. ensemble checks

Seven rows of the table link tests outside tests/verification, including the CI tests behind ISO Tests 18 and 19. Run them with:

uv run pytest tests/test_heat_radiant_threshold.py tests/test_fic_vs_fed_speed.py \
    tests/test_slice_node_index.py tests/test_visibility.py \
    tests/test_rerouting_golden.py tests/test_smoke_speed.py \
    tests/test_iso_table21_coupled.py tests/test_iso_table22_coupled.py

uv run pytest runs everything. Tests that need the external data store (for example the radiometer decks, $HEAT_RADIOMETER_DATA) are skipped when it is absent.

The FDS cases need their FDS output, which is not in the repository; each test page says where to get it or how to rerun FDS.

Last updated on September 30, 2026