Oil-water emulsion separation studies

Use @emulsion-separation and the model in neqsim.process.equipment.separator.emulsion. Read the study guide for exact equations, units, APIs, numerical checks and applicability before running a calculation. Read the current Javadocs for overloads and setup order.

Build from the existing separator

1. Start from a solved ThreePhaseSeparator with oil and aqueous phases. Record EOS, mixing rule, fluid characterization, absolute pressure and temperature. Reuse named process objects rather than reconstructing an unrelated fluid from guessed component fractions.
2. Create EmulsionSeparationStudy with the separator and an explicit effective separation area in m2. The study copies state; never initialize properties or alter phase inventories on the live separator to run the study.
3. Pass dispersed-water flow in m3/s to run(...). If the input is a flux in m/s, multiply it by the declared effective area first. Do not equate total produced-water flow to emulsified-water flow without declaring that assumption. The four-argument run(...) takes initial dispersed-water volume divided by area in m; it does not take total emulsion-layer thickness. Derive that conserved state from measured water inventory or an explicit inversion of the model's height relation. Never pass measured layer thickness directly.
4. State whether the effective area is measured, drawing-derived or estimated. For a horizontal vessel, use horizontal interface width times effective settling length when that approximation is appropriate. The vessel's circular cross section is a different area. Retain the constant-area assumption and the physical bounds of the available liquid space.

DenseEmulsionLayerModel provides the dense-layer closure. This first study tracks a separate diagnostic control volume; it does not modify the live separator's stored mass, outlet streams, level controllers or calculated entrainment. Ordinary process runs do not execute the study. Do not describe study results as coupled outlet-quality predictions.

Handle missing measurements without blocking screening

Screen every online oil-in-water or water-cut analyser for validity before using it (fraction of samples identical to the previous sample, a single repeated value, zero or negative readings, comm-fault sentinels such as 307 or 499.8). On Johan Sverdrup P1 in Oct 2026 all three OIW analysers were frozen or zero while the water was visibly oily; report such tags as unusable and fall back to bottle tests rather than computing with them.

Start an explicitly estimated case when droplet and coalescence measurements are absent. Use a droplet radius of 1.35e-3 m and a coalescence time of 60 s as the documented starting assumptions, retaining ESTIMATED_UNCALIBRATED. The radius comes from the paper's model-fluid experiment; neither value is a universal crude-oil property.

Use guessKinetics(radiusM, contactTimeSeconds) for explicit screening guesses or calibratedKinetics(radiusM, coalescenceRatePerM2Second) for supplied omega. Both return new immutable studies; retain the returned object. Give a supplied calibrated coefficient precedence over the estimated time. The explicit-property constructor accepts measured interfacial tension together with oil and water densities. The separator constructor instead uses its copied fluid's property models. Record the chosen basis and limitations. State radius versus diameter explicitly and let the current model API perform the coefficient conversion.

Vary radius and coalescence time by a factor of three around the starting assumptions and show which scenarios change the predicted operating regime. Call the result a sensitivity envelope, not a statistical confidence interval or P10/P90 range. Do not tune a coefficient merely to obtain the desired separator capacity.

Troubleshoot with distinct hypotheses

Compare a verified baseline with independently initialized scenarios for incoming dispersed-water flux, droplet size and coalescence kinetics. Preserve the distinction between measured operating changes and assumed mechanisms. Report layer growth or drainage, critical flux, steady-state existence, available-height limits and the model control-volume residual.

Use time-aligned rate, pressure, temperature, level and outlet-quality data when available. Identify whether the observations distinguish greater load, finer droplets and slower coalescence. A single water-in-oil measurement usually cannot identify a unique mechanism. Hand other equipment hypotheses to root-cause analysis, and chemical selection, dose-response evidence and compatibility to production chemistry.

For oil-in-water symptoms, use the existing separator performance model and check the phase roles. The present dense-layer closure is for water-in-oil; do not promise a validated inverse-emulsion model. Likewise, changes in temperature or brine chemistry require justified property/kinetic inputs, not a generic rule that they improve separation.

Evidence and limitations

The closure is based on Zlobin et al., AIChE Journal (2026), DOI 10.1002/aic.70429. Their experiments use millimeter-sized, nearly monodisperse water-in-oil model emulsions. Raw data are identified at DOI 10.5281/zenodo.17463071. Document which equations and data were actually reproduced. Field crude validation, polydisperse population dynamics and product-specific chemistry remain separate qualification tasks.

Do not claim this implementation resolves interfacial aging, adsorption, asphaltene film rheology, droplet breakup or electric-field effects. These can motivate later closures but are not implied by the coalescence parameter. Do not infer a demulsifier dose without experimental dose-response evidence.

Keep returned warnings, rejected cases and missing inputs visible. Record parameter provenance and the tested source version. Verify conservation within the study, timestep sensitivity and source-separator immutability. Report engineering validation separately from regression-test execution. Package routing is navigation only; inspect the deployed MCP registry before claiming any operation can be invoked remotely.

Related workflows

• Process modeling: solved, reusable flowsheets and named equipment.
• Production chemistry: chemical compatibility and measured dose-response models.
• Root-cause analysis: competing equipment hypotheses and historian evidence.
• Emulsion viscosity: existing viscosity and phase-inversion correlations, distinct from a dynamic emulsion-layer inventory.