Quantitative risk engine

AQRA

Advanced Quantitative Risk Analysis Software

One environment from release to risk contour, with the assumptions carried on the record

AQRA holds source term, dispersion, effect and risk summation on one case, so any number on any output can be walked back to the release that generated it. Where a chain is split across tools the assumptions that produced a contour are carried by hand between them, and a contour that reaches a drawing arrives without the basis that justifies it.

Flare and vapour plume dispersing over a petrochemical plant

Computational workflow

Each stage writes its selected correlation, its validity range and its assumption set into the case, and the next stage reads them from there rather than from a re entered copy.

Source term characterisation

Discharge modelling for sub cooled, saturated and two phase inventories. Choked and unchoked regimes, isenthalpic flash fraction, aerosol rainout and pool spreading with conductive and convective substrate heat transfer.

Atmospheric dispersion

Momentum dominated near field jet with air entrainment, gravity driven slumping where the cloud is negatively buoyant, and passive Gaussian far field. Pasquill Gifford stability with surface roughness and averaging time correction.

Consequence modelling

Jet and pool fire by solid flame radiation with view factor and atmospheric transmissivity, BLEVE fireball diameter and duration, flash fire envelope to the lower flammable limit, and vapour cloud overpressure by congestion and confinement assessment.

Vulnerability and harm

Probit transformation of thermal dose in kilowatt per square metre to the four thirds power second, side on and reflected overpressure, and toxic load integrated under Haber's rule with substance specific exponent.

Risk integration

Frequency weighted summation across release cases, hole size bands, wind sectors and stability classes. Location specific individual risk, individual risk per annum against occupancy, and cumulative societal risk.

Decision support

Occupied building exceedance assessment to API RP 752 and 753, blast resistant design basis derivation, and fire and gas detector coverage optimisation against the same scenario set.

What it computes

Release and dispersion

Phases: Gas, liquid, flashing liquid and two phase
Regimes: Momentum jet, dense gas slumping, passive Gaussian transition
Duration: Continuous, time varying and instantaneous
Atmosphere: Pasquill stability A to F, wind rose weighted, day and night split

Consequence and harm

Fire: Jet fire, pool fire, fireball, flash fire envelope
Explosion: Congestion based overpressure with obstructed region definition
Toxic: Dose integrated load against substance specific probits
Vulnerability: Probit functions for thermal, overpressure and toxic effect

Risk and application

Individual risk: Location based contours and IRPA with occupancy patterns
Societal risk: F N curve against jurisdictional criteria
Siting: Occupied building assessment to API RP 752, 753 and 756
Detection: Fire and gas coverage mapping from the same scenario set

Individual risk contours at 1E-05 to 1E-08 per year drawn over a plant plot plan with a control room outside the inner contour — Individual risk contours are computed across the whole scenario set and laid over the plot plan. A siting decision is read off the same field the assessment produced, not a redrawn approximation of it.

Societal risk F N curve plotted against intolerable and broadly acceptable criterion lines — Societal risk leaves the engine as a cumulative F N curve, plotted against the criterion lines that apply in your jurisdiction. The step in the tail is escalation, where one event outcome carries a fatality count far above the rest of the distribution.

Thermal radiation decay with escape impairment, piloted ignition and equipment damage thresholds marked — Thermal radiation falls with distance and atmospheric transmissivity. The engine reports the distance to each harm threshold. Those distances set escape routes, equipment protection and siting.

How we know the answers are right

Each correlation is stated with the data it was fitted to, the range it holds over and the conditions under which it stops holding.

Log log comparison of engine results against an independent reference with a tolerance band and outliers ringed — Every case in the validation corpus is plotted against the independent reference. Points inside the band are accepted. Ringed cases are opened and the physics examined before the result is accepted.

Benchmarked on real studies, not textbook cases

The engine is validated against an independent industry reference across a corpus of completed assessments carrying real inventories, real layouts and real weather sets. Textbook verification cases confirm the arithmetic. They do not test whether a model behaves on a plant.

Disagreement is investigated, not tuned away

Where our result and the reference diverge beyond tolerance, the case is opened and the physics examined. Sometimes the reference is closer to reality, sometimes we are, and sometimes both models are outside their stated range. The case is closed only once the cause of the divergence is understood.

The exceptions register is published to users

Substances and regimes where standard correlations are known to be weak are documented and shipped with the engine, including reactive monomers and several strongly non ideal materials. An engineer who knows where a model is weak can allow for it in the assessment.

What stays on the record

The release scenario, with the hole size basis and the isolation assumption behind it
The frequency dataset used, named and versioned, not merely referenced
Every correlation selected at every stage, with the range it is valid over
The weather set and occupancy pattern applied, with their source
Who ran the case, when, and what changed since the previous run
Which outputs were issued, so a drawing can be tied back to a case revision

Deployment, licensing and support

Installation: Installed to your own workstations or a managed virtual desktop. No inventory, layout or population data leaves your network boundary at any point.
Licensing: Named user or concurrent seat, with a floating pool where an analysis team shares capacity across offices.
Validation pack: Verification cases, benchmark comparison record and the model exceptions register, issued with the licence for inclusion in your own quality file.
Case portability: Self contained case files that remain openable across versions, so an assessment can be reopened and rerun years after issue.
Competency: Analyst training covering discharge regime selection, stability and roughness assignment, probit basis and assumption discipline, not menu navigation.
Technical support: Model and correlation queries answered by the engineers who wrote the engine and who run assessments on it.

What it is not

AQRA is a screening to assessment grade engine for the release, dispersion, consequence and risk chain. It is not a computational fluid dynamics package, and where a problem genuinely needs resolved three dimensional flow we will tell you so rather than stretch a correlation to cover it.

Put AQRA on your next assessment

We will come back with the scope, the timeline and what the licence and training would cost. Tell us the release cases you need assessed and the site the contours have to land on.

Arrange a walkthrough