DNV Phast and Safeti 9.31: Consequence Analysis or QRA?

A thermal-radiation contour answers a consequence question; it does not, by itself, tell an operator how often people are exposed to that hazard. That distinction is the starting point for choosing between DNV Phast and Safeti. This research-based technical assessment uses official documentation checked on November 7, 2026 UTC; we did not independently install or test either product.

DNV’s product FAQ identifies Phast as consequence-modeling software and Safeti as consequence-and-risk software. Both remain current DNV offerings. The older DNV GL name in some project specifications refers to the same corporate lineage: DNV’s history records the return to the DNV name in 2021, and its governance page identifies the foundation-owned group structure. The official release notes mark Phast and Safeti 9.31, dated August 21, 2026, as the latest version. The earlier 9.11 label is therefore not the current baseline verified for this assessment.

Begin with a defensible release scenario

DNV documents Phast for discharge, dispersion, fire, explosion and toxic-effect modeling following loss of containment from equipment such as vessels, tanks, pipework and pipelines. The FAQ distinguishes instantaneous and continuous releases, constant or time-varying discharge, atmospheric-tank spills and user-defined sources. The engineering task is to select a scenario that represents the equipment and operating condition, rather than simply choose the largest contour from an arbitrary input set.

For a hypothetical pressurized line leak, a study brief should identify the material, operating pressure and temperature, release geometry and duration, and the environmental conditions to evaluate. The responsible engineer also needs to justify the selected source and consequence model. DNV’s documentation establishes that these classes of scenario can be modeled; it does not establish that a particular hole size, duration or weather case represents your facility. Proposed mitigation measures should be compared using a consistent and documented scenario basis.

Phast presents results in graphs, a 3D viewer and Word/Excel reports according to the product page. The FAQ also documents GIS shape-file import and consequence/risk-contour export. These outputs can support layout and emergency-planning discussions, but a plotted effects distance is not a guarantee of safety outside that line. The acceptance study should explain the effect threshold and assumptions associated with every contour used in a decision.

Safeti adds the likelihood and exposure question

Safeti combines consequence and risk modeling and supplies individual and societal risk metrics, including location-specific individual risk contours, FN curves and potential loss of life. DNV describes risk-ranking points for identifying contributors at a location and sensitivity studies for evaluating risk-reduction measures. This is the relevant step when the deliverable requires a QRA rather than the consequence of one specified event.

The study team must therefore govern the likelihood and exposure assumptions as well as the consequence inputs. For example, changing the event-frequency basis or the population assumed at a receptor changes the risk question being asked even when a physical release scenario is unchanged. A useful procurement demonstration should trace one important risk contributor back to those assumptions, then show how an agreed mitigation changes the reported metric. Producing an FN curve alone does not establish an ALARP demonstration or regulatory acceptance.

Decide when the surroundings need a different model

DNV’s free-field guidance describes an open environment without obstacles or undulating terrain affecting the modeled phenomena, and recommends CFD where those effects matter. The Phast CFD documentation separates pool-fire, jet-fire and dispersion add-ons, powered by KFX. It specifies, for example, that the pool-fire add-on assumes immediate ignition of a pool fixed in size. That assumption matters if the real question involves a changing spill or delayed ignition; the method must fit the scenario being claimed.

The same documentation describes the geometry-sensitive outputs and treats those CFD capabilities as separately purchased or leased extensions. Multi-Component and Extended Explosions are also listed separately on the product pages. Obtain the exact module scope before interpreting a base-license quote as coverage for every mixture, explosion or congested-geometry study.

Safeti’s product page documents import of CFD results through .asc files from version 9.1 onward. The 9.31 release notes are a material part of evaluating that workflow: they describe result-affecting fixes involving CFD hazard-grid alignment and toxic-probit unit handling in earlier versions. For an existing study, record the actual build, review the relevant known issues and fixes, and have the study owner decide what must be rerun. An upgrade should not silently change the documented basis of an issued assessment.

Deployment and licensing need an exact configuration

The published requirements describe 64-bit Windows installations and give different hardware guidance for small scenario sets, larger studies, CFD and Safeti QRA. For example, the FAQ lists 32–64 GB RAM for CFD and large Safeti studies, while Excel is required only for the corresponding import/export functions. Use the full current requirements with the proposed workload rather than selecting a machine from the smallest example.

There is an important qualification in the same FAQ: it offers a preinstalled virtual-machine route but also lists Citrix, Hyper-V, VirtualBox and VMware as unsupported virtualization tools. The CFD commercial page likewise mentions preinstalled virtual machines. Resolve that distinction with DNV in writing for the exact hosted service or infrastructure being proposed; the commercial option does not establish support for an arbitrary in-house virtual desktop. Phast Online is described separately as a web-based set of purpose-built consequence applications, not as proof that all desktop Phast and Safeti functions run in a browser.

DNV documents standalone, dongle and network licensing, alongside subscription/lease and perpetual options. The quoted product, duration, concurrent-user limits, modules and support entitlement need to match the project. The Knowledge Centre supplies downloads, release information, manuals and support resources. Its catalogue also describes a Phast API webinar, but this assessment did not verify a current API contract covering the buyer’s intended automation. Confirm that scope rather than assuming Safeti and every add-on share the same automation surface.

Compare the required deliverable and model assumptions

Gexcon EFFECTS is a relevant consequence-modeling alternative, with documented fire-radiation, explosion-overpressure and toxic-dispersion outputs. RISKCURVES addresses QRA, including individual-risk contours, FN curves and risk-ranking outputs. Compare Phast with the consequence task and Safeti with the QRA task using the same scenario set, material data, likelihood basis and report requirements. The reviewed evidence does not justify declaring one package universally more accurate or faster.

Our CHARM consequence-modeling article provides related reading for a wider screening list. When the downstream decision concerns detector placement, the Detect3D article covers that adjacent workflow. A consequence or risk result and a detector-coverage assessment answer different questions; no automatic connector between these products was verified here.

Phast is worth shortlisting when the deliverable centers on defensible consequence scenarios; Safeti becomes relevant when those consequences must be combined into a quantified risk study. Confidence is high in that documented distinction and the current release identification, but limited for site-specific predictions and deployment support beyond the published configuration. Before commitment, ask DNV to demonstrate the governing scenario and, where applicable, its contribution to risk in the exact build and module set being offered, with the assumptions and result-affecting release changes visible to the approving engineer.