# Case studies

> Case studies of numerical simulation applied to safety: dispersion, egress, fire resistance, explosions, reliability.

A few engagements, chosen across different services and sectors. For each: the question, the method adopted and how the result was used.

Explosives storage under the Seveso Directive · Published study

## Integrated risk analysis of an explosives storage facility

**Question**
Verify the safety of an existing establishment, with fifteen storage magazines and two processing buildings, and its compatibility with the surrounding land use.

**Limit of the simplified method**
Quantity-distance tables assume free-field conditions: they ignore the reflections, channelling and shielding produced by the site geometry and the earth barricades.

**Model**
Overpressure with Viper::Blast over nested domains; fragment projection with a Monte Carlo simulation, 10,000 trajectories per scenario; dispersion of toxic products with FDS; human reliability with the SHERPA method for the processing buildings.

**Outcome**
At the higher thresholds CFD and empirical relations agree within a few metres. At 0.03 bar a representative magazine reaches 212 m against 544 m in free field: the barricades more than halve the area of reversible injuries. The toxic impact beyond the site boundary is negligible.

**Use of the result**
Damage zones for the Safety Report and the technical review; sizing of earth barricades and storage limits.

[Paper · Chemical Engineering Transactions 126 (2026) 169–174](https://doi.org/10.3303/CET26126029)

CFDRISKHRA

Software [Explosions](https://www.nvmeria.it/en/software/#g-explo) · [Consequences and risk](https://www.nvmeria.it/en/software/#g-risk) · [Reliability and human factors](https://www.nvmeria.it/en/software/#g-rel)

Explosives plant under the Seveso Directive · New manual operations

## Human reliability with the SHERPA method for new explosives operations

**Question**
Estimate the frequency of the explosions associated with new manual operations in three rooms, repackaging of explosives and assembly of propelling charges, for the Preliminary Safety Report.

**Limit of the simplified method**
In manual operations the initiating event is an operator error: component failure databases provide no frequency, and a generic value does not distinguish one task from another.

**Model**
SHERPA diagrams for each room, with the error probability of every task step; errors multiplied by annual work volumes (about 250,000 manual weighings of black powder), by the fraction that brings explosive to the floor, by the ignition probability (10% for loose black powder, 2% for propellant) and by barrier unavailability. Overpressure with Viper::Blast on the site terrain, fragment projection and dispersion of toxic products with FDS.

**Outcome**
Seven explosion events, from 25 to 1,000 kg, with frequencies between 1.8·10⁻⁷ and 1.3·10⁻⁵ occurrences/year: six exceed the credibility threshold of 10⁻⁶. For 1,000 kg of propellant the 0.03 bar threshold reaches 275 m. Compartmentation limits the domino effect: loose black powder stays confined to a 50 kg room.

**Use of the result**
Accident scenarios, frequencies and damage areas for the Preliminary Safety Report; layout of operations across the rooms as a prevention measure.

HRACFDRISK

Software [Reliability and human factors](https://www.nvmeria.it/en/software/#g-rel) · [Explosions](https://www.nvmeria.it/en/software/#g-explo)

Utility-scale battery energy storage · NFPA 855

## Integrated safety analysis of BESS installations

**Question**
Demonstrate, through the Hazard Mitigation Analysis required by NFPA 855, that the credible failure modes of about 240 lithium-ion containers are mitigated.

**Limit of the simplified method**
Separation distances do not distinguish between fire, off-gas explosion and toxic dispersion, which are separate damage vectors.

**Model**
HAZID and bow-tie; SIL verification of 27 instrumented functions to IEC 61508 and 61511, and LOPA; fire with FDS, off-gas explosion with Viper::Blast, toxic gases with EFFECTS.

**Outcome**
On the adjacent container, 6 m away, the inner wall stays below 35 °C. In front of the explosion vent 15.5–17.2 kPa are reached; beyond 15 m the overpressure falls below 4.4 kPa. All scenarios remain below 10⁻⁶ occurrences/year.

**Use of the result**
A Hazard Mitigation Analysis compliant with NFPA 855 and the Italian fire service guidelines; operating rules for emergency teams.

CFDSILRISK

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire) · [Explosions](https://www.nvmeria.it/en/software/#g-explo) · [Consequences and risk](https://www.nvmeria.it/en/software/#g-risk) · [Reliability and human factors](https://www.nvmeria.it/en/software/#g-rel)

Hydrogen production, storage and refuelling for trains

## Quantitative risk analysis of a hydrogen facility

**Question**
Assess the frequencies and damage areas of a new facility with electrolysis, storage up to 700 barg in 120 vessels and train refuelling, in an urban area next to the railway line.

**Limit of the simplified method**
Generic databases do not reflect the failure rates of components in hydrogen service, and hydrocarbon correlations underestimate the reactivity of the mixture and the possible transition to detonation.

**Model**
HAZID on an ISO 17776 checklist; parts count of 1,718 components and 745 m of piping over nine sections, five hole sizes up to full bore; QRA with HyRAM+ 5.1.1, event trees for dispersion, jet fire and explosion (Baker-Strehlow-Tang); compressor relief valve openings with Phast.

**Outcome**
Damage distances for radiation and overpressure within a few tens of metres for most scenarios; flammable cloud up to about 87 m (½ LFL) for the worst credible release; some flash fires at maximum pressure critical towards the surrounding areas.

**Use of the result**
Risk reduction measures (IEC 61508 detection and isolation, Fire & Gas mapping, walls resistant to jet fire and explosion) and HAZOP, LOPA and SIL studies for detailed design.

RISKH₂

Software [Consequences and risk](https://www.nvmeria.it/en/software/#g-risk)

Natural gas transmission · Landfall in a microtunnel

## Rupture of a gas pipeline inside a landfall microtunnel

**Question**
Establish whether, after a catastrophic pipeline rupture, the microtunnel stays intact, whether gas can migrate through the soil to the surface, and whether water and gas can be expelled through the access shaft.

**Limit of the simplified method**
Gas and water undergo extreme deformation in a flooded tunnel surrounded by saturated sand: a simplified model cannot describe the interaction between jet, water, concrete and soil.

**Model**
Eulerian code SPEED 3D on two half-symmetry models of 11.2 and 10 million cells; concrete with the Holmquist-Johnson-Cook damage model, seawater with a Steinberg equation of state, saturated sand with a P-alpha porous model; rupture of 20% of the diameter, at 145 and 78 barg.

**Outcome**
Damage to the tunnel remains local. Gas reaches the surface, with an estimated crater of 5–10 m at 145 barg and 3–6 m at 78 barg, and an underground cavity forms that could collapse. Water is discharged through the access shaft.

**Use of the result**
A documented and traceable picture of the effects for the pipeline operator.

EULER

Software [Explosions](https://www.nvmeria.it/en/software/#g-explo)

Chemical process plant · After an explosion

## Flammable vapour or dust: which explosion fits the damage

**Question**
After an explosion in a production unit, quantify the overpressures of two hypotheses, a hydrocarbon vapour cloud and dust accumulated inside the process piping, and compare them with damage thresholds of 20–50 kPa.

**Limit of the simplified method**
Empirical correlations do not represent the actual congestion and confinement of the unit, which determine which hypothesis can explain the damage.

**Model**
exploCFD, which sets the initial conditions at ignition analytically (CSC correlation for vapours, Nagy correlation for dust) and computes their evolution with the SU² solver; parametric study of 624 combinations of obstacles, blockage and cloud height; dust Kst from laboratory analysis.

**Outcome**
The actual vapour cloud produces at most 8.5 kPa: reaching 20 kPa would take 115 kg of fuel. The dust explosion reaches about 750 kPa at 27 ms, consistent with extensive damage.

**Use of the result**
A physics-based discrimination between the accident hypotheses.

CFDATEX

Software [Explosions](https://www.nvmeria.it/en/software/#g-explo)

Petrochemical plant · Critical plant section

## Risk of a plant section in three configurations

**Question**
Compare the individual and societal risk of a critical plant section in its current configuration, in the planned one and with additional control measures.

**Limit of the simplified method**
Multi-energy explosion models require congestion and confinement to be estimated beforehand: two similar configurations may turn out indistinguishable.

**Model**
Releases and consequences with Phast; overpressure computed on the three-dimensional geometry of the section with 3D Explosions; individual (LSIR) and societal risk with Safeti, compared with the acceptability criteria for workers and the public.

**Outcome**
In the current configuration the worker F/N curve exceeds the acceptability limit up to about 5 fatalities, at 3·10⁻³ per year for one fatality. In the planned configuration it falls within the ALARP region; the additional measures lower it further, by up to 40% for 2–3 fatalities.

**Use of the result**
Control measures to add to those originally planned, reducing the risk within the ALARP region.

RISKEXPLO

Software [Explosions](https://www.nvmeria.it/en/software/#g-explo) · [Consequences and risk](https://www.nvmeria.it/en/software/#g-risk)

Pharmaceutical plant · New ethanol distillation unit

## Fire safety engineering for a distillation unit

**Question**
Demonstrate a level of safety equivalent to the R/EI 120 compartmentation up to +14.94 m required by the fire service, which was hard to build because of wind, seismic loads and existing foundations.

**Limit of the simplified method**
The prescriptive solution fixes the compartmentation without considering the radiation the fire actually produces on the targets.

**Model**
Fire safety engineering to Ministerial Decree of 9 May 2007; five ethanol pool fires simulated with FDS, 16 meshes and over 885,000 cells, 5 m/s wind towards the nearest target, radiation sensors on the escape route and the tanks.

**Outcome**
Maximum radiation of 3.3 kW/m² against the 12.6 kW/m² domino threshold, and below 1 kW/m² on the escape route. A 3 m high EI 120 wall protecting the stairwell is enough.

**Use of the result**
A design variation report submitted to the fire service.

CFDFSE

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire)

Food plant · Vegetable oil filtration unit

## Fire resistance with coupled CFD and thermo-structural analyses

**Question**
Demonstrate performance level III for steel structures required to achieve class 240, which prescriptively meant extensive passive protection.

**Limit of the simplified method**
The nominal curve takes no account of the actual fire or of the suppression system.

**Model**
Six vegetable oil pool fire scenarios simulated with FDS, up to 1.4 million 20 cm cells, with and without the foam system; thermal transient and global non-linear structural analysis with SAFIR, to EN 1993-1-2.

**Outcome**
With the high-availability foam system the steel stays below 400 °C in every scenario, with no collapse mechanism. Without it, 500–600 °C are exceeded in about five minutes and a beam fails after 330–420 s.

**Use of the result**
Performance level III without passive protection, with the fire suppression system as the equivalent measure.

CFDFEM

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire) · [Structures](https://www.nvmeria.it/en/software/#g-struct)

Complex underground station

## Emergency ventilation of an underground station

**Question**
Size the fans and define the ventilation control logic for normal operation and for emergencies.

**Limit of the simplified method**
A one-dimensional model describes the network of tunnels and stations, but not the local smoke and visibility conditions along the escape routes.

**Model**
IDA Tunnel for network ventilation; FDS on geometry from a 3D scan, with boundary conditions computed by IDA Tunnel; Pathfinder for egress.

**Outcome**
Tenability conditions in the areas of interest (visibility, temperature, FED, radiation) compared with the calculated evacuation times.

**Use of the result**
Fan sizing, emergency logic and a fire safety management system supported by bow-tie analysis.

CFD1DCROWD

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire) · [Egress and pedestrian flows](https://www.nvmeria.it/en/software/#g-crowd)

Metro interchange station between two lines · Emergency ventilation

## Emergency ventilation of an interchange station with a network model

**Question**
Verify whether, by replacing the fans in the existing ventilation chambers, the system provides at least 8 air changes per hour in the station during a fire (Italian Ministerial Decree of 11 January 1988) and 15 m visibility towards the exits in the tunnels (Ministerial Decree of 21 October 2015).

**Limit of the simplified method**
Calculating air changes from the nominal fan flow rates ignores the piston effect of trains, the draught of the fire and the season: in a network of connected tunnels and stations the actual flows redistribute.

**Model**
One-dimensional model of both lines in IDA Tunnel, with tunnels, shafts, ventilation chambers, moving trains and reference weather data; a train fire peaking at 16 MW; platform scenarios in winter and summer and tunnel scenarios in winter, each with its own ventilation strategy. Fans of 100,000 and 150,000 m³/h replacing the existing 80,000 and 60,000 m³/h units.

**Outcome**
On the platforms air changes stay above 13 per hour on one line and above 20 on the other throughout the fire, against the 8 required. In the tunnels visibility drops to zero only between the fire and the ventilation chamber: in all eight scenarios the smoke does not reach the platform.

**Use of the result**
Sizing of the new fans and ventilation strategies for each fire location; boundary conditions for the FDS simulations of the mezzanines.

1DFSE

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire)

Road transport of dangerous goods · Motorway section

## Risk from dangerous goods transport on a motorway section

**Question**
Assess, through a quantitative risk analysis, the risk associated with dangerous goods traffic along a motorway section.

**Limit of the simplified method**
Quantitative risk analysis was developed for fixed installations; in transport the sources move along the route, and road carries about 60% of dangerous goods in Europe and over 90% in Italy.

**Model**
Transport risk analysis (TRA), the extension to dangerous goods transport of the process-industry QRA methodology; risk integration with RISKCURVES.

**Outcome**
Over about 43 km and about 34,000 dangerous goods transits, two thirds of them diesel and petrol, individual risk along the road axis stays around 10⁻⁹ per year and falls below 10⁻¹² within about 55 m. The F/N curve lies entirely below the reference criterion.

**Use of the result**
A quantitative basis for assessing the risk of the section and the reduction measures.

RISKTRA

Software [Consequences and risk](https://www.nvmeria.it/en/software/#g-risk)

Food production plant · Change to the means of escape

## New protected escape routes in a food production plant

**Question**
Verify that a new layout of the protected escape routes, required by construction and hygiene constraints at an advanced stage of works, performs comparably to the approved fire safety design.

**Limit of the simplified method**
The approved design relied on an ASET/RSET comparison: any change to the geometry of the routes requires the comparison to be repeated with the same method.

**Model**
Five fire scenarios simulated with FDS, three from the approved design and two additional ones, with visibility and temperature at 1.8 m above the floor; evacuation computed with Pathfinder on the new geometry, with simulations extended to 300 s.

**Outcome**
RSET between 110 and 133 s across the five scenarios. ASET > RSET is satisfied in all of them, with the minimum 30 s margin required by the Ministerial Decree of 3 August 2015.

**Use of the result**
Technical opinion supporting the change to the fire safety design, with self-closing fire-rated doors at the positions indicated.

CFDCROWD

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire) · [Egress and pedestrian flows](https://www.nvmeria.it/en/software/#g-crowd)

School building · Cross-laminated timber structure

## EI fire resistance of CLT floor slabs

**Question**
Verify the EI 120 fire resistance of the slabs as actually installed, with build-ups not covered by tabulated certifications or direct tests.

**Limit of the simplified method**
Tables do not describe build-ups combining timber, screeds, insulation and ceilings in the as-built configuration.

**Model**
Two-dimensional finite element thermal analysis with SAFIR, ISO 834 curve for 120 minutes to Eurocode 5; nine scenarios, including no ceilings and insulation degrading above 100 °C.

**Outcome**
The unexposed face reaches at most 86 °C with gases above 1,000 °C, and charring stays within the first centimetres of the panel. EI 120 is met with wide margins in every scenario.

**Use of the result**
Compliance of the as-built configuration and a technical basis for the fire resistance certificates.

FEM

Software [Structures](https://www.nvmeria.it/en/software/#g-struct)

Industrial warehouses over 1,000 m²

## Fully developed fire in industrial warehouses

**Question**
Verify occupant safety and the fire resistance of the structures of several warehouses, and the outdoor areas reached by a fully developed fire.

**Limit of the simplified method**
A rule-based check does not show how far radiation and heat reach outside the building.

**Model**
Engineering approach to Ministerial Decree of 9 May 2007; post-flashover fire on the maximum fire loads, with Fluidyn’s Panache and Panfire codes.

**Outcome**
Outdoor areas exposed to radiation and residual heat for each warehouse, with the possible secondary effects.

**Use of the result**
Verification of the fire resistance class and of interference with the surrounding areas.

FSESEMI-CFD

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire)

Glass industry · Storage of finished product

## Fire in pallets of finished product stored under tensile structures

**Question**
Define the conditions for storing pallets of glass bottles, with layer pads and polyethylene film, under tensile structures, up to six pallets high, and the distances to keep from internal buildings and the site boundary.

**Limit of the simplified method**
No reliable heat release curve is available in the literature for mixed-composition pallets; tabulated distances ignore the boundary wall and the wind.

**Model**
Laboratory tests on one pallet and on two stacked pallets; fire spread over a 16×16×6 pallet island with one pallet ignited every 60 s; thermal radiation computed with Fluidyn-PANFIRE at 2 and 5 m above ground, without wind and with a 5 m/s wind.

**Outcome**
Peak heat release of 312 kW for one pallet and 427 kW for two stacked pallets. Without wind the 2.7 m boundary wall keeps the outside below 2.5 kW/m²; with a 5 m/s wind it exceeds 5 kW/m². Staying below 12.6 kW/m² at 5 m above ground requires 12.5 m.

**Use of the result**
A 13 m distance from internal structures and 10 m from the boundary, with a wall at least 2.5 m high; a 37 l/min·m water curtain to NFPA 13; a map of the areas usable for storage.

TESTFSE

Software [Fire and smoke](https://www.nvmeria.it/en/software/#g-fire)

Vegetation fire · Power distribution network

## Vegetation fire and electrical equipment: reconstructing the origin

**Question**
Establish whether a vegetation fire could have been ignited by a piece of power distribution equipment that was involved in it.

**Limit of the simplified method**
The equipment fault and the fire coincide in time, but that does not show where the fire started: its spread has to be reconstructed on the actual terrain, fuels and wind.

**Model**
Helicopter-borne LiDAR survey of the area; perimeter and severity from multispectral satellite images, sequence of satellite-detected thermal anomalies; wind from several weather stations. Alternative ignition hypotheses simulated with Wildfire Analyst.

**Outcome**
The wind-driven fire developed along a narrow, elongated path. With ignition at the equipment the front would have had to advance hundreds of metres against the wind, and the simulation does not reproduce the perimeter; with ignition at the upwind end the simulated perimeter matches the surveyed one, as the damage to the vegetation confirms.

**Use of the result**
Technical advice on the origin of the fire.

WFIRELIDAR

Software [Wildland fire](https://www.nvmeria.it/en/software/#g-wild)

Wildland fire · Overhead power line

## Wildland fire and an overhead line fault: cause or consequence

**Question**
Verify whether a fault on an overhead power line could have ignited the wildland fire that swept the surrounding area.

**Limit of the simplified method**
A fault inside the burned area does not prove that the fire started there: in strong wind a conductor break may be a consequence, not the cause.

**Model**
Ground and helicopter-borne LiDAR survey; fuels and severity from satellite images. Backward simulation from the final perimeter with Wildfire Analyst to locate the compatible ignition points, then forward scenarios with wind adjusted for terrain and vegetation. Energy check of the conductor break caused by a wind-borne branch.

**Outcome**
Only one ignition point, away from the line, reproduces the actual perimeter, and the time at which the fire reaches the line is consistent with the reports of the event. The recorded wind was strong enough for a branch to cut the conductor.

**Use of the result**
Technical advice on the causal link between the fault and the fire.

WFIRELIDAR

Software [Wildland fire](https://www.nvmeria.it/en/software/#g-wild)

Package of three process furnaces

## Availability of a furnace system

**Question**
Verify that mean availability exceeds 98% over a five-year mission (43,800 hours), as required by the client.

**Limit of the simplified method**
Availability depends on the combination of random failures and spurious trips of the safety functions, which a component-by-component calculation does not capture.

**Model**
Reliability block diagrams in RAMP with Monte Carlo simulation over 30,000 missions; spurious trip rates of the instrumented functions taken from the SIL verification in exSILentia.

**Outcome**
Mean availability of 99.1%, within the acceptance criterion at 90% confidence: configuration approved.

**Use of the result**
A maintenance strategy based on the criticality of each component.

RAMSIL

Software [Reliability and human factors](https://www.nvmeria.it/en/software/#g-rel)

Electrolysis plant with hydrogen storage · Gasholder and header

## Fault trees for the hydrogen gasholder and header

**Question**
Quantify the frequency of two critical events in an electrolysis plant: an explosive atmosphere in the hydrogen gasholder and an unisolated release from the header.

**Limit of the simplified method**
A qualitative review of the barriers does not show which combinations of failures lead to the event or how much each component weighs; with voting logic and redundant instruments a hand calculation is not practical.

**Model**
Logos, software developed by TECSA for fault tree analysis; AND, OR and 2-out-of-4 voting gates; 19 and 17 basic events across level and pressure transmitters, shutdown valves, limit switches, compressors and DCS cards, with rates from OREDA, exida, HSE and Lees and annual proof tests. Minimal cut sets and importance measures (criticality and RAW).

**Outcome**
Both events stay many orders of magnitude below the credibility threshold of 10⁻⁶ occurrences/year. In the gasholder all dominant cut sets contain the pair of level transmitters; in the header 42% of the frequency comes from a line leak combined with failure to trip the electrolysis.

**Use of the result**
Components on which to focus testing and maintenance, and redundancies to check against common cause failures.

FTARISK

Software [Reliability and human factors](https://www.nvmeria.it/en/software/#g-rel)

Refinery under the Seveso Directive · Congested process area

## Fire & Gas mapping of a process area

**Question**
Define quantitatively the performance required of flame, flammable gas and H₂S detection, and verify whether the existing layout achieves it.

**Limit of the simplified method**
Placing detectors close to the sources does not tell what fraction of events is detected: obstructions and shadowing depend on the actual plant geometry.

**Model**
Site-specific performance targets agreed with the client: fires of 90 to 640 kW to be detected within 10 s with target coverage of 80% and 70%; gas clouds of 5, 7 or 10 m depending on confinement and congestion (HSE OTO 93-002). Three-dimensional coverage with HazMap3D over nine fire zones, for single (1ooN) and confirmed (2ooN) alarm.

**Outcome**
A network of 160 multi-spectrum IR flame detectors in 2ooN voting, in an area that had none; 14 open-path and 56 point detectors for flammable gas; 132 new H₂S detectors and 23 relocated. For hydrogen, fixed detection is not recommended in favour of ignition control and ventilation.

**Use of the result**
Gap analysis and upgrade recommendations for the Fire & Gas system, with traceable targets for the operator and the authority.

F&GRISK

Software [Consequences and risk](https://www.nvmeria.it/en/software/#g-risk)

---
Original page: https://www.nvmeria.it/en/case-studies/
