DEXPI engineering-diagram reference cases

NeqSim carries three synthetic public reference cases as executable regression evidence for the coordinated professional PFD and DEXPI/P&ID campaigns. They establish a reproducible acceptance baseline and exercise the additive EngineeringDiagramDelivery facade without changing the legacy diagram exporters.

Case Source object Topology Material connections
DEXPI-REF-SIMPLE ProcessSystem Feed, valve, separator, compressor, cooler, and product boundaries 4
DEXPI-REF-BRANCHED ProcessSystem Separator with gas-compression and liquid-pumping branches 5
DEXPI-REF-MULTI-AREA ProcessModel Inlet, compression, export, and flare-boundary areas 7

The shared test fixture is EngineeringDiagramReferenceFixtures. It uses the SRK equation of state with the classic mixing rule, a synthetic methane/ethane/n-heptane fluid, kg/h material flow, bara pressure, and degree Celsius temperature specifications. The committed src/test/resources/dexpi/reference-cases/reference-case-manifest.json pins those assumptions, provenance, expected counts, engineering state, and qualification boundaries.

Executable evidence

EngineeringDiagramReferenceCasesTest creates two fresh instances of every case and checks:

EngineeringDiagramDeliveryTest reuses the same fixtures to check fail-closed single-area and multi-area delivery of controlled semantic JSON, assessed DEXPI, native SVG/PDF, content hashes, visual fingerprints, and structured projection losses.

DexpiVisualQualityAssessmentTest expands the fixed rendering benchmark with the same simple, branched, and per-area cases plus a mixer/heater/splitter/recycle/control-loop case and an isolated empty topology placeholder. Every sheet is exported only through NeqSim’s Proteus-compatible DEXPI writer and rendered from its actual graphical primitives. The suite checks:

The structural gate complements, but does not replace, full-sheet and readable-detail PNG visual inspection. It distinguishes renderer/layout defects from missing source-model engineering data.

The flow-arrow regression records routed-segment, source-arrow, and rendered-filled-arrow counts. Missing, duplicate, or renderer-dropped arrows produce stable findings instead of relying on visual memory. This is a directional-readability gate, not a standards-conformance determination.

The first full-sheet benchmark inspection found that instrument bubbles intersected full equipment data bars and sat outside the generated battery limit. The layout now reserves 55 mm above an equipment centre for instruments and expands the battery-limit envelope through the highest bubble; an exact coordinate regression protects both clearances. Re-rendered simple, branched, and recycle/control cases show separated bubbles and data bars inside the boundary.

A subsequent whole-sheet instrumentation review found generic measuring lines terminating at equipment centres, controller bubbles using field-location symbols, and synthesized controller signals ending in empty process space. Measurements now resolve to process segments or nozzles, their bubble groups follow the sensing-point lane outside equipment data bars, controller location symbols are explicit, and closed-loop actuation is emitted only for an attached final control element. The assessment records measurement, sensing-location, controller, complete-loop, incomplete-loop, and synthesized-proposal counts so these defects fail deterministically in future diagram work.

The compatibility metrics from the first topology review remain available alongside the stricter DEXPI-oriented metrics: transmitter/controller counts, resolved and missing measurement attachments, measuring-line validity, closed/incomplete loops, actuating-signal validity, and synthesized proposal counts. Missing source data is reported separately from renderer/layout failures; the exporter does not invent a nozzle, controller, valve, or operational intent to make a proposal appear complete.

The same fail-closed rule now applies to piping data. Hydraulic inside diameter remains labelled ID ... mm, never silently converted to DN/NPS. Missing nominal-size source data remains visible as SIZE?. The assessment records line-size provenance, missing-size states, detected size/property changes, reducers, and property breaks. A line-size transition without a fitting, a reducer without two distinct endpoint sizes and connection points, or a property change without a break marker is an error. Full-sheet inspection additionally checks that stream numbers, line designations, fitting labels, and equipment annotations remain legible and non-overlapping.

Recycle connectivity is projected when the recycle block exposes a configured outlet through the standard equipment outlet API and a downstream unit consumes that same stream identity. The benchmark asserts the directed recycle-to-mixer connection in the Proteus XML before rendering. The writer does not infer a return line from convergence state: an unconfigured outlet remains absent, and no fluid state or connection is invented.

Run the focused regression with the repository Maven wrapper:

./mvnw -Dtest=EngineeringDiagramReferenceCasesTest,EngineeringDiagramDeliveryTest,DexpiVisualQualityAssessmentTest test

Explicit limitations

The cases do not claim ISO 10628 conformance, DEXPI EV certification, or interoperability with a named commercial CAE product. Native DEXPI Process export remains area-scoped. The multi-area delivery is a deterministic NeqSim package of independently assessed native area files plus explicit plant-wide manifest evidence; it is not a native DEXPI whole-plant profile. Controlled document and native SVG/PDF projections are delivered beside DEXPI and remain separate semantic and graphical views.

The Proteus per-area files are compatibility sheets, not controlled drawings with paired off-page references and accountable revision approval. Generated P&ID content is a proposal until piping, valve, nozzle, instrument, safeguard, design-data, safety-lifecycle, and discipline reviews are completed by accountable engineers.