NeqSim 3.27.0

Status: prepared release candidate; publication pending. Prepared on 2026-10-31 from master 3c1c5dda4a58cbab78330612889e377b57f7793c, covering 232 commits after v3.26.0. Only changes merged into that snapshot are included. The version-preparation PR does not publish artifacts or create a tag.

Complete change list for this snapshot.

Review before upgrading

Numerical changes and model qualification

Rebaseline saved engineering cases against their measured data and balance checks. Results can change for wet-gas and trace-water flashes, CPA caloric properties, Huron–Vidal derivatives, compressor chart calculations, transport properties, reservoir depletion and two-fluid transient energy balances. Passing software tests does not qualify every fluid or operating range.

• Water-dew-point calculations now enforce gas/aqueous fugacity equilibrium; GERG-water temperature derivatives and CPA boundary/hydrate-amount paths have additional consistency checks (#4558, #4645, #4728, #4731). Water, ice and hydrate boundaries remain distinct calculations. Select the intended EOS, mixing rule and phase options; one boundary test is not a general inhibitor/oil/multiphase qualification.
• Compressor chart mode honors the selected polytropic method and resolves small wet-gas shaft duties more accurately. Compare pressure rise, head, efficiency and power when rerunning calibrated maps (#4721, #4735, #4739).
• EmissionsTracker uses methane GWP 28, consistent with the other IPCC AR5 emission calculators. Recorded CO2-equivalent totals may change (#4751).
• Unsupported PVT-regression targets and invalid heavy-TBP estimates fail explicitly. Revisit previously accepted input rather than suppressing the validation (#4540, #4570).

Saved models and public contracts

#4630 repairs loading of older .neqsim files, and thermodynamic deserialization reconciles inconsistent total-mole inventories. Validate representative prior-release files, equipment state, units, component inventories and rerun results before migrating an application. Same-version serialization round trips do not establish compatibility with every older object graph. A complete API and cross-version persistence assessment remains a release gate; this candidate does not claim universal backward compatibility.

When upgrading from before 3.26.0, also follow the 3.26.0 migration notes, particularly the JAMA-to-EJML interfaces, saved numerical-solver state and corrected E300 critical-volume units. Those changes already shipped in 3.26.0. Any newly identified breaking contract must follow governance and the API lifecycle policy, including the required NRC and major contract review; preparing the requested number does not waive those gates.

Runtime and packaging

The standard core artifact requires Java 17+, the separate Java 8 artifact requires Java 8+, and the MCP runner requires Java 21+. Java source in the repository remains Java 8 compatible. The Python development/report toolkit now requires Python 3.11+, following the Pillow security update; that requirement is separate from the core Java artifacts (#4542, #4713).

Main changes since 3.26.0

Thermodynamics, physical properties and PVT

• Conserve multiphase roots and phase amounts, safely recycle removed phase storage, equilibrate late aqueous seeds, refresh translated UMR-CPA state, and stop stalled PH iterations (#4502, #4526, #4533, #4629, #4649, #4658).
• Modernize PR-CPA and caloric consistency; expose higher and mixed site derivatives and an opt-in four-site polarity topology. Qualify SRK/PR Huron–Vidal derivatives and clone isolation, and reduce temporary derivative allocation. Measured MEG, nitrogen-water and CO2 assessments retain their documented limits; they do not qualify every parameter transfer (#4516, #4521, #4577, #4637, #4648, #4660, #4664, #4706, #4744, #4748).
• Validate interfacial-tension models, isolate transport clone state, correct diffusion volumes and liquid pair invariants, align Chung/CO2 models, and add bounded Letsou–Stiel viscosity and Watson latent-heat estimates (#4499, #4507, #4562, #4572, #4581, #4606, #4640, #4686, #4725).
• Preserve Pedersen plus-fraction mass; correct isothermal swelling, near-critical black-oil/PVTO conversion, and CVD/DLE regression path history. Add PVT characterization studies and reports, verified E300 handoff and calibrated EOS checkpoints (#4559, #4569, #4594, #4602, #4609, #4621, #4747).

Reservoir history and well-to-host systems

• Add generalized Rs/Rv material-balance history estimates, measured component history replay and fitting, depletion-path PVT, physical finite-aquifer coupling and joint Bayesian history scenarios (#4741, #4753, #4759). See the material-balance and aquifer uncertainty guides for the supported assumptions and uncertainty interpretation.
• Preserve layered well definitions through deep copies; make water-injection mass/PVT bases explicit; solve active multi-zone allocation and radial-Darcy coupling; add deviated tubing surveys, VFP lift curves and VFPPROD hydraulics (#4475, #4490, #4506, #4512, #4520, #4531, #4557, #4676).

Process equipment, control and optimization

• Add common compressor shafts, driver/site power limits, fixed-speed map operating-point classification, turboexpander nozzle/IGV/JT-bypass closure, fixed-geometry ejector off-design screening, analyser sample-system checks, and optional emulsion-separation studies (#4661, #4667, #4691, #4694, #4695, #4700, #4715).
• Add bounded transactional adaptive steps, sampled pressure/level transmitter chains, transactional valve actuators, embedded LCV/PCV tests, exact safety event timing, final-element fault response, and an opt-in run watchdog (#4650, #4665, #4672, #4685, #4693, #4702, #4738, #4752).
• Extend lifecycle constraint coverage, boundary refinement, non-additive relief portfolios, sampled two-dimensional envelopes, route-conditioned separator pressure and common-shaft speed actions (#4474, #4624, #4677, #4704, #4729, #4742, #4756).
• Add bounded refinery conversion and hydrogen-supply balances and correct electrolyzer water conservation/Faradaic modes and the alkaline efficiency reference (#4486, #4488, #4493, #4696, #4699).

Transient pipelines and safety: retain experimental limits

Two-fluid updates conserve phase-appearance sources, boundary/component energy, node thermodynamics and mechanical pressure at closed walls, with improved thermal recovery and stiff friction treatment (#4608, #4611, #4673, #4679, #4689, #4732, #4758). Review the canonical two-fluid evidence matrix. These individual repairs do not establish full rich-gas blowdown, terrain pooling and repressurization/evaporation qualification.

The experimental Safety Shutdown acoustic path adds exact valve-travel timing, conservative cell inventory, bounded forward pump curves and lift-check boundaries (#4708, #4714, #4720, #4724, #4734, #4746). Respect its documented fluid, wave-speed, pump-flow and boundary assumptions. This is not a general multiphase hydraulic-hammer or safety-system certification.

Live release-inventory studies now gate outlet withdrawal using current phase geometry (#4690); see the coupled inventory guide.

Structure-H hydrate data: preliminary qualification only

#4754 adds 24 digitized published experimental markers, provenance, reproducible system-correlation fits and checks. These coefficients describe complete methane/water/promoter systems; they are not component Langmuir/Kihara parameters, do not update COMP with qualified sH parameters, and do not enable structure H in hydrate flashes. Executable hydrate models still support sI and sII. See the structure-H qualification record.

Engineering reports, agents and development tools

Add process, PVT, flow-assurance, SURF, utilities and onshore/offshore layout study/report workflows, with executable documentation, equipment tables, drawings, capability discovery and PFD layout review (#4503, #4578, #4592, #4604, #4612, #4613, #4749, #4757). Plugin-only installs gain required report modules and explicit engine-version checks (#4643). Engineering-coverage CI retains validation while avoiding writes to protected master (#4763).

Planned distributions

Distribution	Version or expected asset	Runtime
Core Maven artifact	com.equinor.neqsim:neqsim:3.27.0	Java 17+
Core standalone library	neqsim-3.27.0.jar	Java 17+
Java 8 Maven artifact	com.equinor.neqsim:neqsim:3.27.0-Java8	Java 8+
Java 8 standalone library	neqsim-3.27.0-Java8.jar	Java 8+
MCP runner and checksum	neqsim-mcp-server-3.27.0-runner.jar and .jar.sha256	Java 21+
MCP container	ghcr.io/equinor/neqsim-mcp-server:3.27.0	Container runtime

The MCP module’s own Maven identity remains 1.0.0-SNAPSHOT; its core revision/neqsim.version resolve to 3.27.0. The runner filename and container tag use the core release version. The Python package and agent-plugin marketplace have separate publication processes.

The preparation restores Maven 3.9.16, the publishing baseline used by 3.25.0 and 3.26.0. Dependency update #4615 had reintroduced 3.10.0 while the Java 8 release profile still uses central-publishing 0.11.0. #4318 documents the unwanted local metadata in that combination’s publication bundle. Keep the standard profile at 0.12.0 and both profiles at waitUntil=published; qualify both bundles before advancing the wrapper. This preparation changes no library dependency or simulation source.

Release completion requirements

1. Review these notes against the final candidate and confirm API and saved-model compatibility, migration instructions and any required contract decisions.
2. Require configured CI and security gates on the final candidate: full core tests on supported configurations, slow-test shards, model specifications, formatting, documentation/Javadoc, coverage and packaged MCP protocol tests. Parent-commit results do not establish final-candidate success.
3. Build and inspect the Java 17/Java 8 libraries, sources/Javadoc, and Java 21 MCP runner. Verify embedded versions, class-file targets, Java 8 dependency pins and runtime behavior, and MCP checksums. Local POM validation alone is not package or protocol qualification.
4. Merge the reviewed version PR. The default-branch workflow can build assets and create a draft release. Include these notes and the generated change list in the release body. Preparation alone does not authorize publication.
5. After release-owner approval, publish and verify both Maven Central versions reach PUBLISHED, GitHub assets/checksums are downloadable, and the MCP container is available. Record the final tag, commit and workflow evidence.

Local and hosted validation evidence is recorded in the preparation PR. These notes describe candidate changes and required gates; the expected artifacts above are not asserted to be published or available.