NeqSim Process Extraction Skill

Convert unstructured engineering information into the canonical NeqSim JSON format accepted by ProcessSystem.fromJson() and ProcessSystem.fromJsonAndRun().

Core Principle

Extract structured data into a constrained JSON schema. Do NOT write NeqSim Java/Python code.

The JSON schema is finite and well-defined. ProcessSystem.fromJson() handles all
NeqSim API calls deterministically. Errors come back as structured, actionable messages.

P&ID operational workflow: When the source is a P&ID and the user asks
about a valve action, active train, isolation boundary, bypass, drain, vent,
or control-loop behavior, load neqsim-pid-process-operations. Extract both
the steady-state topology and the model delta needed to simulate the action.

Exception for route hydraulics: When the source is a STID/E3D/P&ID/stress-isometric
line-list table with serial pipe segments, use
neqsim.process.equipment.pipeline.routing.PipingRouteBuilder rather than the generic
JSON process builder. The route builder preserves line-list segment metadata, K-value
minor losses, elevations, and explicit connection topology.

Architecture decision (MANDATORY): Before assembling JSON, classify the process
complexity. Small/medium processes (≤ ~15 units, single recycle loop) use a single
ProcessSystem. Large processes (multiple plant areas, cross-area recycles, different
fluids) must be split into multiple ProcessSystem objects composed inside a
ProcessModule, or use pre-built ProcessModuleBaseClass implementations.
See Section 16 for the decision guide.

---

1. Target JSON Schema

Every extraction must produce JSON matching this format:

{
  "fluid": {
    "model": "SRK",
    "temperature": 323.15,
    "pressure": 65.0,
    "mixingRule": "classic",
    "multiPhaseCheck": false,
    "components": {
      "methane": 0.80,
      "ethane": 0.08,
      "propane": 0.05,
      "CO2": 0.03,
      "n-butane": 0.02,
      "nitrogen": 0.01,
      "n-pentane": 0.005,
      "n-hexane": 0.005
    }
  },
  "process": [
    {"type": "Stream", "name": "well stream", "properties": {"flowRate": [75000.0, "kg/hr"]}},
    {"type": "ThreePhaseSeparator", "name": "inlet separator", "inlet": "well stream"},
    {"type": "Compressor", "name": "export compressor", "inlet": "inlet separator.gasOut",
     "properties": {"outletPressure": 120.0, "isentropicEfficiency": 0.78}},
    {"type": "ThrottlingValve", "name": "letdown valve", "inlet": "inlet separator.oilOut",
     "properties": {"outletPressure": 15.0}}
  ],
  "autoRun": true
}

Field Reference

Multiple Fluids (Named)

For processes with different feed compositions, use the fluids map:

{
  "fluids": {
    "gas_feed": { "model": "SRK", "temperature": 323.15, "pressure": 80.0, "mixingRule": "classic", "components": {"methane": 0.90, "ethane": 0.05, "propane": 0.03, "n-butane": 0.02} },
    "water_feed": { "model": "CPA", "temperature": 293.15, "pressure": 80.0, "mixingRule": "CLASSIC_TX_CPA", "components": {"water": 0.999, "MEG": 0.001} }
  },
  "process": [
    {"type": "Stream", "name": "gas inlet", "fluidRef": "gas_feed", "properties": {"flowRate": [50000.0, "kg/hr"]}},
    {"type": "Stream", "name": "water inlet", "fluidRef": "water_feed", "properties": {"flowRate": [5000.0, "kg/hr"]}}
  ]
}

---

2. Equipment Type Mapping

Map natural language equipment names to NeqSim JSON type values. Use the longest matching keyword to avoid false matches.

Separation

Compression & Expansion

Heat Transfer

Valves

Pumps

Piping Routes

PipingRouteBuilder is not a JSON equipment type. It is a Java/Python-accessible builder for serial route hydraulics. Use it when the input table has from/to nodes, pipe lengths, sizes, elevations, fittings, valves, and K values. Extract the route rows first, then build the route model and export route.toJson() for traceability.

For P&ID valve-action studies, classify each valve before mapping it to NeqSim: control valves become ThrottlingValve equipment, isolation and shutdown valves become scenario switches or boundary states, check valves become directed route constraints, and BDV/PSV/vent valves become relief or blowdown paths.

Mixing & Splitting

Always model a manifold as Manifold, not Mixer/Splitter. Add all inlet streams with addStream(...), then route downstream from a split stream, not getMixedStream(). A single-destination gathering manifold sets one split (setSplitFactors([1.0])) and routes getSplitStream(0); a distributing manifold sets setSplitFactors([...]) (fractions summing to 1) and reads each outlet with getSplitStream(i). getMixedStream() is the internal commingled stream (before the split) — for inspection only. The Manifold also carries header / branch diameters for hydraulics and mechanical design.

Streams

Other Equipment

---

3. Stream Wiring (Dot-Notation)

Equipment is connected via dot-notation references in the inlet field.

Port Reference Table

Wiring Rules

1. First equipment MUST be a Stream — it gets the fluid from the fluid section
2. Every subsequent equipment MUST have an inlet reference pointing to a previously defined equipment
3. For separators, specify the port — gasOut, liquidOut, oilOut, waterOut
4. For single-outlet equipment — use "name.outlet" or just "name" (default resolves to outlet)
5. Do NOT create circular references — the JSON builder does not support recycle loops directly (add Recycle equipment for convergence)
6. Branching is supported — multiple equipment can reference different ports of the same separator
7. Mixer multi-inlet: use "inlets" (plural) — Mixers require "inlets": ["stream1", "stream2"] (array). Do NOT use "inlet" with an array — it will fail with "Array must have size 1"
8. Separator name without port returns null — resolveStreamReference("HP Sep") returns null. Always use "HP Sep.gasOut" or "HP Sep.liquidOut"

Branching Example

{"type": "ThreePhaseSeparator", "name": "inlet sep", "inlet": "feed"},
{"type": "Compressor", "name": "gas comp", "inlet": "inlet sep.gasOut", ...},
{"type": "ThrottlingValve", "name": "oil valve", "inlet": "inlet sep.oilOut", ...},
{"type": "Pump", "name": "water pump", "inlet": "inlet sep.waterOut", ...}

Mixer Multi-Inlet Example

{"type": "Stream", "name": "gas 1", "properties": {"flowRate": [10000.0, "kg/hr"]}},
{"type": "Stream", "name": "gas 2", "properties": {"flowRate": [5000.0, "kg/hr"]}},
{"type": "Mixer", "name": "gas mixer", "inlets": ["gas 1", "gas 2"]},
{"type": "Cooler", "name": "mixed cooler", "inlet": "gas mixer", "properties": {"outletTemperature": [25.0, "C"]}}

CRITICAL: Use "inlets" (plural key, with array value) for Mixer/multi-inlet equipment. Using "inlet" with an array value will fail.

---

4. Equipment Properties Reference

Stream

Compressor

Cooler / Heater

Property Unit Arrays: Equipment properties can be specified with units using the [value, "unit"] array format. This applies to any property that accepts a unit string, such as outletTemperature, flowRate, etc. The JSON builder uses Java reflection to find matching setter methods.

ThrottlingValve

Pump

Separator / ThreePhaseSeparator

No required properties. Operates at inlet conditions.

Splitter

DistillationColumn

HeatExchanger (Multi-Inlet)

HeatExchanger supports two inlets (hot and cold side):

{"type": "HeatExchanger", "name": "gas-gas HX",
 "inlets": ["hot stream", "cold stream"]}

The first inlet becomes the feed stream; the second is set via setFeedStream(1, stream).

AdiabaticPipe

Route-Level Piping Line Lists

When the source has a line-list or stress-isometric table, extract these fields before constructing the route:

Route extraction workflow:

1. Sort rows in hydraulic flow order from upstream to downstream.
2. Convert NPS/schedule to internal diameter before calling addSegment(...).
3. Convert every valve, bend, tee, reducer, strainer, and entry/exit loss to K.
4. For a route-only study, build the route with PipingRouteBuilder.build(feedStream)

and run the returned ProcessSystem.

1. For a full plant model, call route.addToProcessSystem(process, inletStream)

and pass the returned outlet stream to the downstream equipment. Use the overload with source-equipment metadata when the inlet is an upstream equipment outlet stream.

1. Save route.toJson() and pressure-drop results in the task folder.

Reference guide: docs/process/pipingroutebuilder.md.

---

5. Component Name Mapping

Map common aliases to NeqSim database names. The NeqSim name is case-sensitive.

Hydrocarbon Components

Non-Hydrocarbon Components

Chemical Additives

Aromatics

---

6. Unit Conversion Rules

All NeqSim JSON values must be in standard units. Convert before inserting into JSON.

Temperature

Pressure

Flow Rate

Flow rate in JSON uses the [value, "unit"] array format. Supported unit strings:

Composition: Weight% to Mole Fraction Conversion

NeqSim uses mole fractions (summing to 1.0) in the JSON components field. If the source provides weight% (wt%), mass fractions, or ppm-by-weight, convert as follows:

Formula:

For each component $i$ with weight fraction $wi$ and molar mass $Mi$:

$$xi = \frac{wi / Mi}{\sumj (wj / Mj)}$$

Common Molar Masses (g/mol):

Worked Example:

Input: 70 wt% methane, 20 wt% ethane, 10 wt% propane

Result JSON: {"methane": 0.837, "ethane": 0.128, "propane": 0.035}

ppm-by-weight: Convert ppmw to weight fraction first: $wi = \text{ppm}_w \times 10^{-6}$

Volume% (gas at standard conditions): Volume% ≈ mole% for ideal gas behavior. Use directly as mole fractions.

---

7. EOS Model Selection

Choose the thermodynamic model based on the fluid system:

Decision Rules

1. If water or glycol is present → use CPA with mixing rule "CLASSICTXCPA" and set multiPhaseCheck: true
2. If accuracy for gas density/Z-factor is critical → use GERG2008
3. If heavy oil (C20+) → use PR or SRK with "classic" mixing rule
4. Default / unknown → use SRK with "classic" mixing rule

---

8. Extraction Workflow

Follow this step-by-step process for every extraction:

Step 1: Identify the Source Type

• Text description — paragraph or bullet list describing a process
• Table / spreadsheet — heat & mass balance, operating data, well test
• PFD / sketch — process flow diagram (described or as image)
• Data sheet — equipment data sheet with design conditions
• Mixed — combination of above

Step 2: Extract Fluid Composition

1. Look for: mole fractions, mol%, weight%, component tables
2. Map component names to NeqSim names using the Component Name Mapping table
3. Normalize to mole fractions summing to 1.0
4. If weight% given, note it as an assumption (NeqSim uses mole fractions)
5. If no composition given, flag as missing and use a placeholder

Step 3: Extract Equipment List

1. Scan for equipment keywords using the Equipment Type Mapping table
2. Match the longest keyword first (e.g., "three-phase separator" before "separator")
3. Assign unique names/tags (use P&ID tags if provided, or generate descriptive names)
4. Record the NeqSim type for each

Step 4: Extract Stream Connectivity

1. Look for phrases indicating flow direction: "enters", "goes to", "feeds", "is routed to", "flows to", "passes through"
2. Identify which phase exits which equipment: "gas from the separator", "oil from the 3-phase separator", "compressed gas"
3. Build dot-notation references: "equipment_name.port"
4. Verify no orphan streams (every equipment except feed has an inlet)

Step 5: Extract Operating Conditions

1. Pressures: look for bara, barg, bar, psi, MPa, kPa, atm
2. Temperatures: look for °C, °F, K, degC, degF
3. Flow rates: look for kg/hr, t/h, MMSCFD, MSm3/d, Am3/hr
4. Convert all to NeqSim standard units (K, bara)
5. Assign to the correct equipment property

Step 6: Assemble JSON

1. Build the fluid section with model, T, P, mixing rule, and components
2. Build the process array in topological order (upstream before downstream)
3. First element MUST be a Stream with the feed fluid
4. Wire all equipment with inlet references
5. Set autoRun: true

Step 7: Validate and Report

1. Check composition sums to ~1.0 (within 0.01)
2. Check all stream references point to existing equipment
3. Check no circular references
4. Compute confidence score (see Confidence Scoring below)
5. List all assumptions made
6. List all missing information detected

---

9. Confidence Scoring

Score the extraction confidence on a 0.0–1.0 scale:

Confidence Bands

---

10. Assumption Defaults

When information is not specified, use these engineering defaults and always track them:

---

11. Process Templates

When the extracted topology matches a known pattern, use a template for better reliability.

Template: Gas Dew Point Control

Pattern: cooler → separator → compressor

{
  "fluid": { "model": "SRK", "temperature": "$FEED_T_K$", "pressure": "$FEED_P$", "mixingRule": "classic", "components": "$COMPOSITION$" },
  "process": [
    {"type": "Stream", "name": "feed", "properties": {"flowRate": ["$FLOW$", "kg/hr"]}},
    {"type": "Cooler", "name": "dew point cooler", "inlet": "feed", "properties": {"outTemperature": "$COOLER_T_K$"}},
    {"type": "Separator", "name": "cold separator", "inlet": "dew point cooler.outlet"},
    {"type": "Compressor", "name": "export compressor", "inlet": "cold separator.gasOut", "properties": {"outletPressure": "$EXPORT_P$", "isentropicEfficiency": 0.78}}
  ],
  "autoRun": true
}

Template: Two-Stage HP/LP Separation

Pattern: 3-phase sep → gas compression + oil letdown → LP sep → recompression

{
  "fluid": { "model": "SRK", "temperature": "$FEED_T_K$", "pressure": "$HP_P$", "mixingRule": "classic", "components": "$COMPOSITION$" },
  "process": [
    {"type": "Stream", "name": "well stream", "properties": {"flowRate": ["$FLOW$", "kg/hr"]}},
    {"type": "ThreePhaseSeparator", "name": "HP separator", "inlet": "well stream"},
    {"type": "Cooler", "name": "gas cooler", "inlet": "HP separator.gasOut", "properties": {"outTemperature": 308.15}},
    {"type": "Compressor", "name": "export compressor", "inlet": "gas cooler.outlet", "properties": {"outletPressure": "$EXPORT_P$", "isentropicEfficiency": 0.78}},
    {"type": "ThrottlingValve", "name": "HP-LP valve", "inlet": "HP separator.oilOut", "properties": {"outletPressure": "$LP_P$"}},
    {"type": "Separator", "name": "LP separator", "inlet": "HP-LP valve.outlet"},
    {"type": "Compressor", "name": "LP recompressor", "inlet": "LP separator.gasOut", "properties": {"outletPressure": "$HP_P$", "isentropicEfficiency": 0.75}}
  ],
  "autoRun": true
}

Template: Multi-Stage Compression with Intercooling

Pattern: compressor → cooler → scrubber → compressor → cooler → scrubber → ... (N stages)

Build dynamically with equal pressure ratio per stage:

• ratioperstage = (Pout / Pin) ^ (1/N)
• stageP[i] = Pin × ratioperstage^i
• Each stage: Compressor → Cooler (to intercooler temp) → Separator (scrub condensate)
• Last stage: Compressor only (no aftercooler/scrubber, unless specified)

Template: Gas Cooling and JT Expansion

Pattern: cooler → separator → JT valve → cold separator

{
  "fluid": { "model": "SRK", "temperature": "$FEED_T_K$", "pressure": "$FEED_P$", "mixingRule": "classic", "components": "$COMPOSITION$" },
  "process": [
    {"type": "Stream", "name": "feed", "properties": {"flowRate": ["$FLOW$", "kg/hr"]}},
    {"type": "Cooler", "name": "pre-cooler", "inlet": "feed", "properties": {"outTemperature": "$PRECOOL_T_K$"}},
    {"type": "Separator", "name": "inlet scrubber", "inlet": "pre-cooler.outlet"},
    {"type": "ThrottlingValve", "name": "JT valve", "inlet": "inlet scrubber.gasOut", "properties": {"outletPressure": "$JT_P$"}},
    {"type": "Separator", "name": "cold separator", "inlet": "JT valve.outlet"}
  ],
  "autoRun": true
}

Template: Simple Oil Stabilization

Pattern: 3-phase sep → valve → flash drum → valve → atmospheric flash

{
  "fluid": { "model": "SRK", "temperature": "$FEED_T_K$", "pressure": "$HP_P$", "mixingRule": "classic", "multiPhaseCheck": true, "components": "$COMPOSITION$" },
  "process": [
    {"type": "Stream", "name": "well fluid", "properties": {"flowRate": ["$FLOW$", "kg/hr"]}},
    {"type": "ThreePhaseSeparator", "name": "production separator", "inlet": "well fluid"},
    {"type": "ThrottlingValve", "name": "1st stage valve", "inlet": "production separator.oilOut", "properties": {"outletPressure": "$STAGE2_P$"}},
    {"type": "Separator", "name": "2nd stage separator", "inlet": "1st stage valve.outlet"},
    {"type": "ThrottlingValve", "name": "2nd stage valve", "inlet": "2nd stage separator.liquidOut", "properties": {"outletPressure": "$STAGE3_P$"}},
    {"type": "Separator", "name": "stabilizer", "inlet": "2nd stage valve.outlet"}
  ],
  "autoRun": true
}

Template: Subsea Tieback (Well → Pipeline → Platform)

Pattern: well stream → choke → pipeline → separator

{
  "fluid": { "model": "SRK", "temperature": "$WELLHEAD_T_K$", "pressure": "$WELLHEAD_P$", "mixingRule": "classic", "multiPhaseCheck": true, "components": "$COMPOSITION$" },
  "process": [
    {"type": "Stream", "name": "well stream", "properties": {"flowRate": ["$FLOW$", "kg/hr"]}},
    {"type": "ThrottlingValve", "name": "production choke", "inlet": "well stream", "properties": {"outletPressure": "$CHOKE_P$"}},
    {"type": "Heater", "name": "pipeline heat loss", "inlet": "production choke.outlet", "properties": {"outTemperature": "$ARRIVAL_T_K$"}},
    {"type": "ThreePhaseSeparator", "name": "inlet separator", "inlet": "pipeline heat loss.outlet"}
  ],
  "autoRun": true
}

Template: TEG Dehydration

Pattern: wet gas → TEG absorber ← lean TEG; dry gas out, rich TEG out

Note: SimpleTEGAbsorber requires two input streams added via addGasInStream() and addSolventInStream(). The JSON builder currently wires via the inlet field, so for TEG dehydration, build the lean TEG stream as a separate feed with a TEG+water fluid.

{
  "fluids": {
    "wet_gas": {
      "model": "SRK", "temperature": "$FEED_T_K$", "pressure": "$FEED_P$",
      "mixingRule": "classic",
      "components": "$GAS_COMPOSITION_WITH_WATER$"
    },
    "lean_teg": {
      "model": "CPA", "temperature": "$TEG_T_K$", "pressure": "$FEED_P$",
      "mixingRule": "CLASSIC_TX_CPA",
      "components": {"TEG": 0.99, "water": 0.01}
    }
  },
  "process": [
    {"type": "Stream", "name": "wet gas feed", "fluidRef": "wet_gas", "properties": {"flowRate": ["$GAS_FLOW$", "kg/hr"]}},
    {"type": "Stream", "name": "lean TEG", "fluidRef": "lean_teg", "properties": {"flowRate": ["$TEG_FLOW$", "kg/hr"]}},
    {"type": "SimpleTEGAbsorber", "name": "TEG absorber", "inlet": "wet gas feed",
      "properties": {"numberOfStages": "$STAGES$", "stageEfficiency": 0.5}},
    {"type": "Heater", "name": "TEG reboiler sim", "inlet": "TEG absorber.liquidOut",
      "properties": {"outTemperature": "$REBOILER_T_K$"}}
  ],
  "autoRun": true
}

Typical defaults: 3–5 stages, stage efficiency 0.5, lean TEG flow 5–10× water to remove, reboiler at ~200°C (473 K). TEG purity: 99–99.5 wt%.

Template: NGL Recovery (Turbo-Expander + Demethanizer)

Pattern: gas inlet → cooler → expander → demethanizer column; overhead = sales gas, bottoms = NGL

{
  "fluid": { "model": "SRK", "temperature": "$FEED_T_K$", "pressure": "$FEED_P$", "mixingRule": "classic", "components": "$COMPOSITION$" },
  "process": [
    {"type": "Stream", "name": "inlet gas", "properties": {"flowRate": ["$FLOW$", "kg/hr"]}},
    {"type": "Cooler", "name": "gas chiller", "inlet": "inlet gas", "properties": {"outTemperature": "$CHILLER_T_K$"}},
    {"type": "Separator", "name": "cold separator", "inlet": "gas chiller.outlet"},
    {"type": "Expander", "name": "turbo-expander", "inlet": "cold separator.gasOut", "properties": {"outletPressure": "$EXPANDER_P$", "isentropicEfficiency": 0.85}},
    {"type": "ThrottlingValve", "name": "liquid JT valve", "inlet": "cold separator.liquidOut", "properties": {"outletPressure": "$EXPANDER_P$"}},
    {"type": "Mixer", "name": "column feed mixer", "inlets": ["turbo-expander.outlet", "liquid JT valve.outlet"]},
    {"type": "Separator", "name": "demethanizer sim", "inlet": "column feed mixer.outlet"},
    {"type": "Compressor", "name": "residue compressor", "inlet": "demethanizer sim.gasOut", "properties": {"outletPressure": "$SALES_P$", "isentropicEfficiency": 0.78}}
  ],
  "autoRun": true
}

Note: For a rigorous demethanizer, replace the Separator with a Column (DistillationColumn). The simplified version uses a cold separator as a proxy. Typical expander outlet: 15–25 bara, efficiency 0.82–0.88, chiller to –30°C to –40°C.

Template: Acid Gas Removal (Amine Sweetening)

Pattern: sour gas → amine absorber ← lean amine; sweet gas out, rich amine to regenerator

Note: Uses SimpleTEGAbsorber which works for generic absorption. For amine-specific thermodynamics, the CPA EOS with MDEA is recommended.

{
  "fluids": {
    "sour_gas": {
      "model": "CPA", "temperature": "$FEED_T_K$", "pressure": "$FEED_P$",
      "mixingRule": "CLASSIC_TX_CPA",
      "components": "$SOUR_GAS_COMPOSITION$"
    },
    "lean_amine": {
      "model": "CPA", "temperature": "$AMINE_T_K$", "pressure": "$FEED_P$",
      "mixingRule": "CLASSIC_TX_CPA",
      "components": {"MDEA": 0.40, "water": 0.60}
    }
  },
  "process": [
    {"type": "Stream", "name": "sour gas feed", "fluidRef": "sour_gas", "properties": {"flowRate": ["$GAS_FLOW$", "kg/hr"]}},
    {"type": "Stream", "name": "lean amine", "fluidRef": "lean_amine", "properties": {"flowRate": ["$AMINE_FLOW$", "kg/hr"]}},
    {"type": "SimpleTEGAbsorber", "name": "amine absorber", "inlet": "sour gas feed",
      "properties": {"numberOfStages": "$STAGES$", "stageEfficiency": 0.5}},
    {"type": "Heater", "name": "amine regenerator sim", "inlet": "amine absorber.liquidOut",
      "properties": {"outTemperature": "$REGEN_T_K$"}}
  ],
  "autoRun": true
}

Typical defaults: 10–20 stages, MDEA 40–50 wt%, amine circulation rate 50–100 L/kg acid gas, regenerator at 120–130°C. Use CPA EOS with mixing rule "CLASSICTXCPA" for polar systems.

Template: Produced Water Treatment (Degassing)

Pattern: produced water → heater → 3-phase separator → water stripper or flash drum

{
  "fluid": { "model": "CPA", "temperature": "$FEED_T_K$", "pressure": "$FEED_P$", "mixingRule": "CLASSIC_TX_CPA", "multiPhaseCheck": true,
    "components": "$WATER_OIL_GAS_COMPOSITION$" },
  "process": [
    {"type": "Stream", "name": "produced water", "properties": {"flowRate": ["$FLOW$", "kg/hr"]}},
    {"type": "Heater", "name": "water heater", "inlet": "produced water", "properties": {"outTemperature": "$HEATER_T_K$"}},
    {"type": "ThreePhaseSeparator", "name": "water degasser", "inlet": "water heater.outlet"},
    {"type": "ThrottlingValve", "name": "flash valve", "inlet": "water degasser.waterOut", "properties": {"outletPressure": "$FLASH_P$"}},
    {"type": "Separator", "name": "atmospheric flash", "inlet": "flash valve.outlet"}
  ],
  "autoRun": true
}

Typical defaults: Produced water at 60–80°C, degassing at 1–3 bara. Use CPA EOS with mixing rule "CLASSICTXCPA" when water is a major component. Composition: primarily water (>95 mol%) with dissolved methane, CO2, and trace hydrocarbons.

---

12. Worked Examples

Example 1: Simple Text Description

Input:

"Feed gas at 80 bara and 40°C enters a cooler to 15°C. The cooled stream goes to a separator. Gas from the separator is compressed to 120 bara."

Extraction:

Output JSON:

{
  "fluid": {
    "model": "SRK", "temperature": 313.15, "pressure": 80.0,
    "mixingRule": "classic",
    "components": {"methane": 0.90, "ethane": 0.05, "propane": 0.03, "n-butane": 0.02}
  },
  "process": [
    {"type": "Stream", "name": "feed gas", "properties": {"flowRate": [50000.0, "kg/hr"]}},
    {"type": "Cooler", "name": "gas cooler", "inlet": "feed gas", "properties": {"outTemperature": 288.15}},
    {"type": "Separator", "name": "scrubber", "inlet": "gas cooler.outlet"},
    {"type": "Compressor", "name": "export compressor", "inlet": "scrubber.gasOut", "properties": {"outletPressure": 120.0, "isentropicEfficiency": 0.75}}
  ],
  "autoRun": true
}

Report:

• Confidence: 0.50 (temperature ✓, pressure ✓, topology ✓, but no composition, no flow rate)
• Assumptions: Placeholder composition (90% CH4), default flow rate 50000 kg/hr, default compressor efficiency 0.75, SRK EOS
• Missing: Feed composition, feed flow rate

---

Example 2: Process with Composition

Input:

"The well stream arrives at 65 bara and 80°C with 75000 kg/hr. Composition: 80% methane, 8% ethane, 5% propane, 3% CO2, 2% n-butane, 1% N2, 0.5% n-pentane, 0.5% n-hexane. It enters a 3-phase separator. Gas goes to a compressor at 120 bara. Oil goes through a letdown valve to 15 bara."

Output JSON:

{
  "fluid": {
    "model": "SRK", "temperature": 353.15, "pressure": 65.0,
    "mixingRule": "classic",
    "components": {
      "methane": 0.80, "ethane": 0.08, "propane": 0.05,
      "CO2": 0.03, "n-butane": 0.02, "nitrogen": 0.01,
      "n-pentane": 0.005, "n-hexane": 0.005
    }
  },
  "process": [
    {"type": "Stream", "name": "well stream", "properties": {"flowRate": [75000.0, "kg/hr"]}},
    {"type": "ThreePhaseSeparator", "name": "inlet separator", "inlet": "well stream"},
    {"type": "Compressor", "name": "gas compressor", "inlet": "inlet separator.gasOut", "properties": {"outletPressure": 120.0, "isentropicEfficiency": 0.75}},
    {"type": "ThrottlingValve", "name": "letdown valve", "inlet": "inlet separator.oilOut", "properties": {"outletPressure": 15.0}}
  ],
  "autoRun": true
}

Report:

• Confidence: 0.85
• Assumptions: SRK EOS (not specified), default compressor efficiency 0.75
• Missing: Compressor efficiency, separator operating temperature

---

Example 3: From a Heat & Mass Balance Table

Input Table:

Equipment List: HP separator → export compressor (95 barg) → aftercooler (35°C) → scrubber

Extraction:

• T = 60°C → 333.15 K; P = 48 barg → 49.01325 bara
• Flow = 120 t/h → 120000 kg/hr
• Composition: map aliases, normalize remainder to nC4
• Export P = 95 barg → 96.01325 bara

Output JSON:

{
  "fluid": {
    "model": "SRK", "temperature": 333.15, "pressure": 49.01325,
    "mixingRule": "classic",
    "components": {
      "methane": 0.825, "ethane": 0.072, "propane": 0.041,
      "CO2": 0.023, "n-butane": 0.039
    }
  },
  "process": [
    {"type": "Stream", "name": "feed", "properties": {"flowRate": [120000.0, "kg/hr"]}},
    {"type": "ThreePhaseSeparator", "name": "HP separator", "inlet": "feed"},
    {"type": "Compressor", "name": "export compressor", "inlet": "HP separator.gasOut", "properties": {"outletPressure": 96.01325, "isentropicEfficiency": 0.75}},
    {"type": "Cooler", "name": "aftercooler", "inlet": "export compressor.outlet", "properties": {"outTemperature": 308.15}},
    {"type": "Separator", "name": "export scrubber", "inlet": "aftercooler.outlet"}
  ],
  "autoRun": true
}

Report:

• Confidence: 0.90 (composition ✓, conditions ✓, flow ✓, equipment clear)
• Assumptions: SRK EOS, default compressor efficiency, remaining 3.9 mol% assigned to n-butane
• Missing: Compressor efficiency, individual heavy component split

---

Example 4: From Equipment Tag List (ProcessPilot style)

Input:

Separators: 20VA001 (HP, 50 barg, 60°C), 20VB001 (LP, 14 barg)
Compressors: 23KA600 (discharge 60 barg, eff 0.78)
Valves: 20XV001 (outlet 15 barg)
Topology: feed→20VA001, 20VA001.gasOut→23KA600, 20VA001.oilOut→20XV001, 20XV001.outlet→20VB001
Feed: 100000 kg/hr, 60°C, 50 barg

Output JSON:

{
  "fluid": {
    "model": "SRK", "temperature": 333.15, "pressure": 51.01325,
    "mixingRule": "classic",
    "components": {"methane": 0.90, "ethane": 0.05, "propane": 0.03, "n-butane": 0.02}
  },
  "process": [
    {"type": "Stream", "name": "feed", "properties": {"flowRate": [100000.0, "kg/hr"]}},
    {"type": "ThreePhaseSeparator", "name": "20VA001", "inlet": "feed"},
    {"type": "Compressor", "name": "23KA600", "inlet": "20VA001.gasOut", "properties": {"outletPressure": 61.01325, "isentropicEfficiency": 0.78}},
    {"type": "ThrottlingValve", "name": "20XV001", "inlet": "20VA001.oilOut", "properties": {"outletPressure": 16.01325}},
    {"type": "Separator", "name": "20VB001", "inlet": "20XV001.outlet"}
  ],
  "autoRun": true
}

---

Example 5: Multi-Stage Compression (from a design report)

Input:

"Compress LP gas from 5 bara to 150 bara in 3 stages with intercooling to 30°C. Gas composition: 88% methane, 6% ethane, 3% propane, 2% CO2, 1% nitrogen. Feed is 25000 kg/hr at 25°C."

Extraction:

• 3 stages, ratio = (150/5)^(1/3) = 3.107
• Stage 1: 5 → 15.5 bara
• Stage 2: 15.5 → 48.2 bara
• Stage 3: 48.2 → 150 bara

Output JSON:

{
  "fluid": {
    "model": "SRK", "temperature": 298.15, "pressure": 5.0,
    "mixingRule": "classic",
    "components": {"methane": 0.88, "ethane": 0.06, "propane": 0.03, "CO2": 0.02, "nitrogen": 0.01}
  },
  "process": [
    {"type": "Stream", "name": "feed gas", "properties": {"flowRate": [25000.0, "kg/hr"]}},
    {"type": "Compressor", "name": "compressor stage 1", "inlet": "feed gas", "properties": {"outletPressure": 15.5, "isentropicEfficiency": 0.78}},
    {"type": "Cooler", "name": "aftercooler 1", "inlet": "compressor stage 1.outlet", "properties": {"outTemperature": 303.15}},
    {"type": "Separator", "name": "scrubber 1", "inlet": "aftercooler 1.outlet"},
    {"type": "Compressor", "name": "compressor stage 2", "inlet": "scrubber 1.gasOut", "properties": {"outletPressure": 48.2, "isentropicEfficiency": 0.78}},
    {"type": "Cooler", "name": "aftercooler 2", "inlet": "compressor stage 2.outlet", "properties": {"outTemperature": 303.15}},
    {"type": "Separator", "name": "scrubber 2", "inlet": "aftercooler 2.outlet"},
    {"type": "Compressor", "name": "compressor stage 3", "inlet": "scrubber 2.gasOut", "properties": {"outletPressure": 150.0, "isentropicEfficiency": 0.78}}
  ],
  "autoRun": true
}

---

13. Output Format

Every extraction MUST produce three things:

1. The NeqSim JSON

The complete JSON object ready for ProcessSystem.fromJsonAndRun().

2. Extraction Report

Present as a structured summary:

EXTRACTION REPORT
─────────────────
Source type:  [text / table / PFD / data sheet / mixed]
Confidence:  [0.XX] — [High / Medium / Low / Very Low]
Equipment:   [N] units extracted
Streams:     [N] connections wired

Assumptions Used:
  - [assumption 1]
  - [assumption 2]
  ...

Missing Information:
  - [missing item 1]
  - [missing item 2]
  ...

Warnings:
  - [warning 1]
  ...

3. Simulation Results (after running)

SIMULATION RESULTS
──────────────────
Status: [SUCCESS / ERROR]

Equipment        T (°C)    P (bara)   Flow (kg/hr)   Notes
─────────       ──────    ────────   ────────────   ─────
Feed              80.0      65.0      75000          —
HP Separator      80.0      65.0      —              Gas + Oil split
Gas Compressor   145.2     120.0      62000          Power: 2450 kW
Oil Valve         55.3      15.0      13000          —

---

14. Error Handling

If ProcessSystem.fromJsonAndRun() returns errors, interpret them:

Tolerant Error Handling

The JSON builder uses tolerant error handling for stream wiring: when a stream reference cannot be resolved (e.g., the upstream unit was skipped), the equipment is removed from the process rather than failing the entire build. These show up as warnings (not errors) in SimulationResult. Similarly, if process.run() throws an exception, it is caught and returned as a warning — the process is still returned in its partially-run state.

This means result.isSuccess() can be true even when result.hasWarnings() is also true. Always check warnings to identify any equipment that was skipped:

result = ProcessSystem.fromJson(json_str)
if result.hasWarnings():
    for w in result.getWarnings():
        print(f"WARNING [{w.getCode()}]: {w.getMessage()}")

---

15. Operating Data Bridge Patterns

Patterns for converting operating data from external sources (Excel, CSV, historians) into NeqSim JSON.

From Excel / CSV Tabular Data

When operating data arrives as a table (common in FEED reports, well tests, plant data):

Step 1 — Identify columns:

Step 2 — Build JSON from rows:

Excel Row:                          NeqSim JSON Unit:
┌─────────────────────────────┐    ┌──────────────────────────────────┐
│ Tag: 23KA601                │ →  │ "name": "23KA601"                │
│ Type: Centrifugal compressor│ →  │ "type": "Compressor"             │
│ Suction P: 48 barg          │ →  │ (inlet stream pressure)          │
│ Discharge P: 95 barg        │ →  │ "outletPressure": 96.01325       │
│ Efficiency: 78%             │ →  │ "isentropicEfficiency": 0.78     │
│ Inlet: from 20VA001 gas     │ →  │ "inlet": "20VA001.gasOut"        │
└─────────────────────────────┘    └──────────────────────────────────┘

Step 3 — Handle multi-row compositions:

If composition is spread across rows (one row per component):

Component    mol%
methane      82.5
ethane        7.2
propane       4.1
CO2           2.3
remainder     3.9  ← assign to n-butane or split to nC4/nC5

Convert to: {"methane": 0.825, "ethane": 0.072, "propane": 0.041, "CO2": 0.023, "n-butane": 0.039}

From Plant Historian Data (PI / IP.21)

When operating data comes from realtime historians:

Tag Mapping Pattern:

Historian Tag            → NeqSim Parameter
─────────────           ─────────────────
PT-20001.PV (bara)      → feed stream pressure
TT-20001.PV (°C)        → feed stream temperature (+273.15)
FT-20001.PV (kg/hr)     → feed stream flowRate
AT-20001-CH4.PV (mol%)  → fluid component "methane" (/100)
AT-20001-C2H6.PV (mol%) → fluid component "ethane" (/100)
PT-23001.PV (bara)      → compressor outletPressure

JSON with Historian Placeholder Tags:

Use "$TAG:tagname$" placeholders that a data bridge fills at runtime:

{
  "fluid": {
    "model": "SRK",
    "temperature": "$TAG:TT-20001.PV+273.15$",
    "pressure": "$TAG:PT-20001.PV$",
    "mixingRule": "classic",
    "components": {
      "methane": "$TAG:AT-20001-CH4.PV/100$",
      "ethane": "$TAG:AT-20001-C2H6.PV/100$"
    }
  },
  "process": [
    {"type": "Stream", "name": "feed", "properties": {"flowRate": ["$TAG:FT-20001.PV$", "kg/hr"]}},
    {"type": "Compressor", "name": "23KA001", "inlet": "feed",
      "properties": {"outletPressure": "$TAG:PT-23001.PV$"}}
  ]
}

Python Data Bridge Example (pandas):

import pandas as pd
import json

# Load operating data from Excel
df = pd.read_excel("operating_data.xlsx", sheet_name="Well Test")

# Build fluid composition from columns
composition = {}
component_map = {"CH4": "methane", "C2H6": "ethane", "C3H8": "propane",
                 "CO2": "CO2", "N2": "nitrogen", "H2S": "H2S"}
for col, neqsim_name in component_map.items():
    if col in df.columns:
        val = float(df[col].iloc[0])
        if val > 1.0:  # Likely mol% not fraction
            val /= 100.0
        composition[neqsim_name] = round(val, 6)

# Normalize
total = sum(composition.values())
composition = {k: round(v / total, 6) for k, v in composition.items()}

# Build JSON
neqsim_json = {
    "fluid": {
        "model": "SRK",
        "temperature": float(df["Temperature_C"].iloc[0]) + 273.15,
        "pressure": float(df["Pressure_bara"].iloc[0]),
        "mixingRule": "classic",
        "components": composition
    },
    "process": [
        {"type": "Stream", "name": "feed",
         "properties": {"flowRate": [float(df["FlowRate_kghr"].iloc[0]), "kg/hr"]}}
        # ... add equipment from equipment sheet
    ],
    "autoRun": True
}

Handling Common Data Issues

---

16. ProcessSystem vs ProcessModule Architecture

NeqSim has three levels of process model organization. The extraction agent MUST choose the right level before assembling JSON or Python code.

Decision Guide

Decision rules (apply in order):

1. Can the entire process be expressed as a single linear/branching chain? → Single ProcessSystem via JSON.
2. Does the process have ≤ 1 recycle loop with ≤ 15 total units? → Single ProcessSystem via JSON (recycles supported via Recycle equipment type).
3. Does the process have distinct plant areas (e.g., separation train + compression + dehydration + export) with different fluids or cross-area recycles? → Split into multiple ProcessSystem objects inside a ProcessModule.
4. Does a pre-built module exist (separation, TEG dehydration, CO2 removal, adsorption)? → Use the corresponding ProcessModuleBaseClass subclass.
5. Does the process require > 50 units or nested recycles across areas? → Use nested ProcessModule containing other ProcessModules.

Architecture A: Single ProcessSystem (JSON-compatible)

When to use: Most extraction scenarios. The JSON builder creates a single ProcessSystem with equipment wired via stream references.

┌──────────────────────────────────────────┐
│           ProcessSystem                  │
│                                          │
│  Stream → Cooler → Separator → Compressor│
│                      │                   │
│                      └→ Valve → LP Sep   │
└──────────────────────────────────────────┘

Limitations of single ProcessSystem:

• All equipment shares one feed fluid definition
• One convergence loop (adjusters + recycles solved together)
• No sub-module encapsulation
• Gets unwieldy above ~15-20 units

Architecture B: ProcessModuleBaseClass (Reusable Black-Box Modules)

When to use: The source describes a standard process subsystem that matches one of NeqSim's pre-built modules. Use the module directly rather than rebuilding from individual equipment.

Available pre-built modules:

Pattern — module as equipment in ProcessSystem:

# A module IS a ProcessEquipmentInterface — add it to a ProcessSystem
from neqsim import jneqsim

SepModule = jneqsim.process.processmodel.processmodules.SeparationTrainModule

sep_module = SepModule("separation train")
sep_module.addInputStream("feed stream", feed_stream)
sep_module.setSpecification("pressure1", 65.0)  # HP sep pressure
sep_module.setSpecification("pressure2", 25.0)  # MP sep pressure
sep_module.setSpecification("pressure3", 5.0)   # LP sep pressure

process = jneqsim.process.processmodel.ProcessSystem()
process.add(feed_stream)
process.add(sep_module)  # Module treated as 1 equipment
process.run()

# Get output streams from module
gas_out = sep_module.getOutputStream("gas exit stream")
oil_out = sep_module.getOutputStream("oil exit stream")

Key API:

• addInputStream(String name, StreamInterface stream) — wire input by port name
• getOutputStream(String name) — get output by port name
• getOperations() → ProcessSystem — access the internal process
• getUnit(String name) — access individual equipment inside the module
• initializeModule() — called automatically when added to a ProcessSystem

Architecture C: ProcessModule (Multi-System Composition)

When to use: The source describes a large facility with multiple distinct process areas that need separate convergence or different thermodynamic models.

┌─────────────────────────────────────────────────────────┐
│                    ProcessModule                        │
│                                                         │
│  ┌──────────────┐   ┌──────────────┐   ┌─────────────┐ │
│  │ ProcessSystem │──►│ ProcessSystem │──►│ProcessSystem│ │
│  │  Separation   │   │  Compression │   │ Dehydration │ │
│  └──────────────┘   └──────────────┘   └─────────────┘ │
│         │                                    │          │
│         └──────── recycle stream ─────────────┘          │
└─────────────────────────────────────────────────────────┘

Pattern — composing multiple ProcessSystems:

from neqsim import jneqsim

ProcessSystem = jneqsim.process.processmodel.ProcessSystem
ProcessModule = jneqsim.process.processmodel.ProcessModule

# Build each plant area as a separate ProcessSystem
sep_system = ProcessSystem()
sep_system.add(feed_stream)
sep_system.add(hp_separator)
sep_system.add(lp_separator)

comp_system = ProcessSystem()
comp_system.add(gas_from_sep)      # shared stream object links the systems
comp_system.add(compressor_1)
comp_system.add(aftercooler_1)
comp_system.add(compressor_2)

dehyd_system = ProcessSystem()
dehyd_system.add(compressed_gas)   # shared stream object
dehyd_system.add(teg_absorber)
dehyd_system.add(dry_gas_stream)

# Compose into a ProcessModule
plant = ProcessModule("Gas Processing Plant")
plant.add(sep_system)
plant.add(comp_system)
plant.add(dehyd_system)
plant.run()  # Runs sub-systems in order; handles cross-system recycles

# Access any equipment across all sub-systems
comp = plant.getUnit("compressor 1")
print(f"Power: {comp.getPower('kW'):.0f} kW")

Key API for ProcessModule:

Cross-system wiring mechanism: Systems connect through shared stream objects. When equipment in system A produces an outlet stream, that same Java object is used as the inlet to equipment in system B. The ProcessModelGraphBuilder automatically detects these cross-system links.

Recycle convergence: If recycles span multiple sub-systems, ProcessModule.run() flattens all units and iterates until all recycles converge (max 100 iterations).

Extraction Workflow Integration

When extracting from a source document, add a Step 0 before the normal extraction:

Step 0: Classify Process Complexity

1. Count the total number of equipment items mentioned
2. Identify distinct plant areas (separation, compression, dehydration, etc.)
3. Check for cross-area recycles (e.g., regenerated solvent returning to absorber)
4. Check if any area matches a pre-built module
5. Choose architecture:

Important: When Architecture B or C is chosen, the extraction agent MUST:

• Still extract all parameters using the same skill rules (units, component names, etc.)
• Produce Python code instead of (or in addition to) JSON
• Clearly state why JSON alone is insufficient
• Use ProcessSystem.fromJsonAndRun() for individual sub-systems where possible,

then compose them in Python

Hybrid Approach (JSON + Python Composition)

For large processes, use JSON for each sub-system and compose in Python:

import json
from neqsim import jneqsim

ProcessSystem = jneqsim.process.processmodel.ProcessSystem
ProcessModule = jneqsim.process.processmodel.ProcessModule

# Each plant area as JSON → ProcessSystem
sep_json = json.dumps({"fluid": {...}, "process": [...]})
comp_json = json.dumps({"fluid": {...}, "process": [...]})

sep_result = ProcessSystem.fromJson(sep_json)   # build but don't run yet
sep_process = sep_result.getProcessSystem()

comp_result = ProcessSystem.fromJson(comp_json)
comp_process = comp_result.getProcessSystem()

# Wire cross-system streams
# Get output from separation and set as input to compression
sep_gas = sep_process.getUnit("HP separator").getGasOutStream()
comp_feed = comp_process.getUnit("comp feed")  # The Stream in comp_json
# Replace comp feed's fluid with sep gas's fluid
comp_feed.setThermoSystem(sep_gas.getThermoSystem().clone())

# Compose and run
plant = ProcessModule("plant")
plant.add(sep_process)
plant.add(comp_process)
plant.run()

When NOT to Use Modules

• Simple processes — don't over-architect a 4-unit process with modules
• Learning / demos — use single ProcessSystem for clarity
• JSON-only requirement — modules require Python/Java code
• No cross-area interaction — independent systems can just be separate fromJsonAndRun() calls

---

17. Word Document (.docx) Input Support

Process descriptions often arrive as Word documents. This section covers extracting simulation parameters from .docx files using python-docx.

Required Library

import docx  # pip install python-docx

Document Loading Pattern

import os, docx

DOC_PATHS = [
    os.path.join(os.path.dirname(globals().get("__vsc_ipynb_file__", "")),
                 "process_description.docx"),
    r"C:\Users\...\process_description.docx",
    "process_description.docx",  # fallback: current directory
]

doc = None
for path in DOC_PATHS:
    if path and os.path.exists(path):
        doc = docx.Document(path)
        break
if doc is None:
    raise FileNotFoundError("Could not find document")

Extracting Paragraphs and Tables

paragraphs = [p.text.strip() for p in doc.paragraphs if p.text.strip()]
tables_data = []
for table in doc.tables:
    rows = [[cell.text.strip() for cell in row.cells] for row in table.rows]
    tables_data.append(rows)

Composition Extraction from Tables

Map document component names to NeqSim names using a lookup table:

NEQSIM_COMPONENT_MAP = {
    "Nitrogen": "nitrogen", "N2": "nitrogen",
    "CO2": "CO2", "Carbon Dioxide": "CO2",
    "Methane": "methane", "C1": "methane",
    "Ethane": "ethane", "C2": "ethane",
    "Propane": "propane", "C3": "propane",
    "i-Butane": "i-butane", "iC4": "i-butane",
    "n-Butane": "n-butane", "nC4": "n-butane",
    "i-Pentane": "i-pentane", "iC5": "i-pentane",
    "n-Pentane": "n-pentane", "nC5": "n-pentane",
    "C6+": "n-hexane",  # Simplified plus-fraction
    "n-Hexane": "n-hexane", "C6": "n-hexane",
    "H2S": "H2S", "Water": "water", "H2O": "water",
}

# Find the composition table (look for "Component" + "Mole fraction" headers)
composition = {}
for table_rows in tables_data:
    header = [h.lower() for h in table_rows[0]]
    if "component" in header and any("mole" in h or "fraction" in h for h in header):
        comp_col = header.index("component")
        frac_col = next(i for i, h in enumerate(header) if "mole" in h or "fraction" in h)
        for row in table_rows[1:]:
            doc_name = row[comp_col].strip()
            neqsim_name = NEQSIM_COMPONENT_MAP.get(doc_name)
            if neqsim_name:
                composition[neqsim_name] = float(row[frac_col])
        break

# Normalize to sum = 1.0
total = sum(composition.values())
composition = {k: v/total for k, v in composition.items()}

Operating Condition Extraction from Narrative

Extract temperatures and pressures from the prose text using pattern matching:

import re

def extract_from_paragraphs(paragraphs, patterns):
    """Search paragraphs for key-value patterns."""
    results = {}
    for p in paragraphs:
        for key, regex in patterns.items():
            m = re.search(regex, p, re.IGNORECASE)
            if m and key not in results:
                results[key] = float(m.group(1))
    return results

# Typical patterns found in process descriptions
patterns = {
    "feed_pressure":  r"(?:inlet|feed)\s+(?:pressure|P)\s*[:\s]+(\d+\.?\d*)\s*bar",
    "feed_temperature": r"(?:inlet|feed)\s+(?:temperature|T)\s*[:\s]+(\d+\.?\d*)\s*°?C",
    "hp_sep_pressure": r"HP\s+(?:separator|sep).*?(\d+\.?\d*)\s*bar",
    "export_pressure": r"export.*?(\d+\.?\d*)\s*bar",
}

For complex documents, also extract from tables with "Parameter"/"Value" structure:

for table_rows in tables_data:
    header = [h.lower() for h in table_rows[0]]
    if "parameter" in header and any("value" in h for h in header):
        param_col = header.index("parameter")
        val_col = next(i for i, h in enumerate(header) if "value" in h)
        for row in table_rows[1:]:
            param = row[param_col].lower()
            try:
                val = float(re.sub(r'[^\d.]', '', row[val_col]))
            except ValueError:
                continue
            if "pressure" in param:
                results["feed_pressure"] = val
            elif "temperature" in param:
                results["feed_temperature"] = val

Architecture Decision After Extraction

After extracting all parameters, count equipment and decide architecture:

# Count equipment mentioned in document
equipment_count = sum(1 for p in paragraphs
    if any(kw in p.lower() for kw in
           ["separator", "compressor", "cooler", "heater", "valve",
            "column", "pump", "mixer", "exchanger", "drum"]))

if equipment_count <= 8:
    print("Architecture A: Single ProcessSystem")
elif equipment_count <= 20:
    print("Architecture A or B: Single system or JSON builder")
else:
    print("Architecture C: ProcessModule with multiple sub-systems")

Tested Example

See examples/notebooks/processextractionfrom_document.ipynb for a full end-to-end example that demonstrates the Free Text → JSON → NeqSim Model pipeline:

1. Loads a Word document describing a gas+condensate processing facility
2. Extracts composition from a table (10 components)
3. Extracts feed and operating conditions from tables and narrative
4. Generates a NeqSim JSON file (processfromdocument.json) as portable intermediate
5. Builds process from JSON via ProcessSystem.fromJson() (core gas train, 10 units)
6. Adds a scrubber liquid recycle using the hybrid approach (clone + Recycle object)
7. Builds a full plant via ProcessModule with 3 sub-systems (20 units)
8. Validates results (8/8 + 11/11 checks pass) and generates 3 figures

18. Recycle Stream Handling

When Recycles Are Needed

Recycles appear whenever a downstream liquid or gas stream must be returned to an upstream unit:

• Scrubber liquid recycle: Knockout drum condensate → HP separator
• Anti-surge recycle: Compressor discharge → compressor suction via cooler
• Solvent loop: Regenerated solvent → top of absorber column
• Reflux: Condenser liquid → top tray of distillation column

JSON Builder Limitation

The JSON builder (ProcessSystem.fromJson()) does not support recycle wiring natively. The Recycle class requires setOutletStream() which is not handled by the reflection-based inlet wiring in JsonProcessBuilder.

Solution: Hybrid approach — use JSON for the main process, then add recycles via Python/Java code after building from JSON.

Recycle Pattern (Clone + Recycle Object)

# 1. Clone a stream to create the "tear" (initial guess) stream
recycle_stream = source_stream.clone("Recycle Stream Name")
recycle_stream.setFlowRate(1e-6, "kg/hr")  # Tiny initial guess
process.add(recycle_stream)

# 2. Add the clone as an extra inlet to the receiving equipment
#    Use addStream() for separators/mixers
hp_separator.addStream(recycle_stream)

# 3. Build the rest of the process normally
# ... (knockout scrubber, pump, etc.)

# 4. Pump the actual recycle liquid back to upstream pressure
pump = ns.Pump("Recycle Pump", scrubber.getLiquidOutStream())
pump.setOutletPressure(upstream_pressure)
process.add(pump)

# 5. Create Recycle object to converge the loop
recycle = ns.Recycle("Recycle Name")
recycle.addStream(pump.getOutletStream())      # Actual downstream output
recycle.setOutletStream(recycle_stream)         # Clone tear stream
recycle.setTolerance(1e-2)                      # Flow/temp/comp tolerance
process.add(recycle)

# ProcessSystem.run() iterates until recycle converges
process.run()

Recycle API Reference

JSON Schema Extension for Recycles

When the extraction identifies a recycle, include it in the JSON process array as a comment/marker that signals hybrid wiring is needed:

{
  "type": "Recycle",
  "name": "Scrubber Liquid Recycle",
  "inlet": "Scrubber Pump",
  "properties": {
    "outletStream": "Recycle Stream Name",
    "tolerance": 0.01
  },
  "_note": "HYBRID: Requires Python wiring — JSON builder cannot set outletStream"
}

The agent should detect this marker and generate additional Python code for recycle wiring after the fromJson() call.

19. Three-Step Extraction Pipeline

The recommended workflow for converting free text to a running simulation:

┌─────────────┐     ┌──────────────┐     ┌──────────────────┐
│  Free Text   │ ──→ │  NeqSim JSON  │ ──→ │  Running Process  │
│  (.docx/txt) │     │  (.json file) │     │  (ProcessSystem)  │
└─────────────┘     └──────────────┘     └──────────────────┘
   Step 1: Parse       Step 2: Structure     Step 3: Build & Run

Step 1: Parse — Extract Parameters

• Use python-docx for .docx, regex for plain text
• Extract: composition, feed conditions, equipment list, operating conditions
• Map component names using NEQSIMCOMPONENTMAP (Section 5)
• Normalize mole fractions to sum to 1.0

Step 2: Structure — Generate JSON

• Build the JSON dict following the schema in Section 1
• Use dot-notation for stream wiring (Section 3)
• Save to a .json file for portability and version control
• The JSON file is the single source of truth for the process

Step 3: Build & Run — Load JSON into NeqSim

# Load JSON and build process
import jpype
ProcessSystem = jpype.JClass("neqsim.process.processmodel.ProcessSystem")
result = ProcessSystem.fromJson(json_str)

if result.isError():
    for err in result.getErrors():
        print(f"[{err.getCode()}] {err.getMessage()}")
else:
    process = result.getProcessSystem()
    process.run()

    # Access equipment by name
    separator = process.getUnit("V-101 HP Separator")

    # Access streams via dot-notation
    gas_stream = process.resolveStreamReference("V-101 HP Separator.gasOut")

API Reference for JSON Pipeline

Stream Reference Dot-Notation

20. Compressor Performance Curves & Anti-Surge Handling

When to Add Compressor Curves

Add performance curves when the document mentions any of:

• Compressor performance maps, characteristic curves, or operating envelopes
• Surge protection, anti-surge control, or recycle valve
• Compressor turndown requirements or minimum flow
• Polytropic or isentropic head values
• Multiple speed operations or variable speed drive (VSD)

CompressorChartGenerator — Auto-Generate from Design Point

CompressorChartGenerator creates a performance map (head, efficiency, surge/choke curves) scaled from the compressor's computed design point. The compressor must be run at least once first so the generator can read getPolytropicFluidHead(), getInletStream().getFlowRate("m3/hr"), and getSpeed().

import jpype
CompressorChartGenerator = jpype.JClass(
    "neqsim.process.equipment.compressor.CompressorChartGenerator"
)

# 1. Configure compressor
comp = ns.Compressor("K-401", inlet_stream)
comp.setUsePolytropicCalc(True)
comp.setOutletPressure(125.0)
comp.setPolytropicEfficiency(0.75)

# 2. Run process once to establish design point
process.add(comp)
process.run()  # Compressor now has head, flow, speed

# 3. Generate chart from design point
chart_gen = CompressorChartGenerator(comp)
comp.setCompressorChart(chart_gen.generateCompressorChart("mid range"))
comp.setCompressorChartType("interpolate and extrapolate")

# 4. Re-run with chart enabled
process.run()

Generation options:

Loading Manufacturer Curves from JSON

If the document includes actual performance data (from OEM data sheets):

# From a JSON file
comp.loadCompressorChartFromJson("path/to/chart.json")

# From a JSON string (e.g., extracted from document)
comp.loadCompressorChartFromJsonString(json_string)

# Export chart as JSON
chart_json = comp.getCompressorChartAsJson()

Chart JSON format:

{
  "compressorName": "K-401 Export Compressor",
  "headUnit": "kJ/kg",
  "maxDesignPower_kW": 8500.0,
  "referenceConditions": {
    "molecularWeight": 18.5,
    "temperature_K": 303.15,
    "pressure_bara": 65.0,
    "compressibilityZ": 0.92
  },
  "speedCurves": [
    {
      "speed_rpm": 3000.0,
      "flow_m3h": [3300, 4700, 6700],
      "head_kJkg": [88.8, 74.0, 37.0],
      "polytropicEfficiency_pct": [67.5, 75.0, 63.75]
    }
  ],
  "surgeFlow": [3300],
  "surgeHead": [88.8],
  "chokeFlow": [6700],
  "chokeHead": [37.0]
}

Anti-Surge Loop Topology

The typical NCS anti-surge pattern (used on offshore platforms):

                    ┌──────────────────────────────────────────┐
                    │              Anti-Surge Loop              │
                    ▼                                          │
Gas ──→ [Mixer] ──→ [Cooler] ──→ [Compressor w/Chart] ──→ [Splitter]──→ Main Export
                                                                │
                                                          [Calculator]
                                                                │
                                                     [Anti-Surge Valve] ──→ [Recycle]
                                                                               │
                                                                          back to Mixer

Equipment roles:

Complete Anti-Surge Code Pattern

import jpype
CompressorChartGenerator = jpype.JClass(
    "neqsim.process.equipment.compressor.CompressorChartGenerator"
)
Calculator = jpype.JClass("neqsim.process.equipment.util.Calculator")

# 1. Clone for anti-surge recycle
as_recycle_stream = gas_stream.clone("AS Recycle Stream")
as_recycle_stream.setFlowRate(1e-6, "kg/hr")
process.add(as_recycle_stream)

# 2. Suction mixer
suction_mixer = ns.Mixer("Suction Mixer")
suction_mixer.addStream(gas_stream)
suction_mixer.addStream(as_recycle_stream)
process.add(suction_mixer)

# 3. Suction cooler (optional, removes recycle heat)
suction_cooler = ns.Cooler("Suction Cooler", suction_mixer.getOutletStream())
suction_cooler.setOutTemperature(273.15 + 30.0)
process.add(suction_cooler)

# 4. Compressor with polytropic calculation
comp = ns.Compressor("Export Compressor", suction_cooler.getOutletStream())
comp.setUsePolytropicCalc(True)
comp.setOutletPressure(125.0)
comp.setPolytropicEfficiency(0.75)
process.add(comp)

# 5. Discharge splitter (stream 0 = product, stream 1 = recycle)
splitter = ns.Splitter("Discharge Splitter", comp.getOutletStream(), 2)
splitter.setFlowRates(jpype.JArray(jpype.JDouble)([-1.0, 1e-6]), "kg/hr")
process.add(splitter)

# 6. Calculator: auto-adjusts splitter based on compressor state
calc = Calculator("Anti-Surge Calculator")
calc.addInputVariable(comp)
calc.setOutputVariable(splitter)
process.add(calc)

# 7. Anti-surge valve
valve = ns.ThrottlingValve("Anti-Surge Valve", splitter.getSplitStream(1))
valve.setOutletPressure(65.0)  # Match suction pressure
process.add(valve)

# 8. Recycle converges the loop
recycle = ns.Recycle("Anti-Surge Recycle")
recycle.addStream(valve.getOutletStream())
recycle.setOutletStream(as_recycle_stream)
recycle.setTolerance(1e-2)
process.add(recycle)

# 9. Export cooler on main product
cooler = ns.Cooler("Export Cooler", splitter.getSplitStream(0))
cooler.setOutTemperature(273.15 + 40.0)
process.add(cooler)

# Phase 1: Run without chart (establish design point)
process.run()

# Phase 2: Generate chart and re-run
chart_gen = CompressorChartGenerator(comp)
comp.setCompressorChart(chart_gen.generateCompressorChart("mid range"))
comp.setCompressorChartType("interpolate and extrapolate")
process.run()

# Check results
recycle_flow = splitter.getSplitStream(1).getFlowRate("kg/hr")
if recycle_flow > 1.0:
    print(f"Anti-surge active: {recycle_flow:.0f} kg/hr recycled")
else:
    print("Operating above surge line — no recycle needed")

Built-in AntiSurge Object (Simpler Alternative)

For simpler checks without the full recycle topology:

# Access the compressor's built-in anti-surge object
anti_surge = comp.getAntiSurge()
anti_surge.setActive(True)
anti_surge.setSurgeControlFactor(1.05)  # 5% margin above surge line

# After running:
is_surge = comp.isSurge()         # True if operating below surge line
is_stonewall = comp.isStoneWall() # True if at choke
head = comp.getPolytropicFluidHead()  # kJ/kg

AntiSurge control strategies:

JSON Builder Limitation

Compressor curves and anti-surge loops cannot be configured via the JSON builder alone. Like recycles (Section 18), they require the hybrid approach: JSON for the main process topology, then Python/Java code to add:

1. Performance chart (CompressorChartGenerator or loadFromJson)
2. Anti-surge loop equipment (Mixer, Splitter, Calculator, Valve, Recycle)

Extraction Agent Detection Rules

When extracting from documents, flag compressors for anti-surge setup when:

• Document mentions "anti-surge", "surge control", "recycle valve", or "minimum flow"
• Compressor has turndown requirements (< 70% of design flow operation expected)
• Multiple compressors in series (cascade surge risk)
• Variable feed conditions (composition or flow rate swings)
• Export/pipeline compressors (typically have anti-surge systems)

---

21. Tested Pitfalls & Known Limitations

Systematic testing of the extraction workflow (see examples/notebooks/testextractionworkflow.ipynb) identified these pitfalls with fixes:

Component Name Mapping (CRITICAL)

NeqSim uses specific component names in its database. Verbose names from documents must be mapped to NeqSim names before generating JSON. The fromJson() builder does NOT do automatic alias resolution.

Rule: Always map verbose component names to the short NeqSim database names listed in Section 5 before generating the JSON.

CPA Mixing Rule Name (CRITICAL)

The mixing rule for CPA must be the enum name "CLASSICTXCPA", not the numeric string "10". The setMixingRule(String) method calls EosMixingRuleType.byName() which expects uppercase enum names.

Mixer Multi-Inlet Syntax (CRITICAL)

Mixers require "inlets" (plural key) with an array value. Using "inlet" with an array causes "Array must have size 1" error.

Separator Stream Resolution

resolveStreamReference("HP Sep") returns null for separators — you MUST specify the port. For non-separator equipment (Cooler, Compressor, etc.), bare names work fine.

Empty Composition Silently Passes

An empty "components": {} does not error at build time but produces a system with no components. Always validate len(components) > 0 before generating JSON.

Unsupported Equipment Types

The following equipment types are NOT supported by fromJson():

• Absorber — use Separator chain or hybrid approach
• DistillationColumn — use hybrid approach (build column in Python/Java)
• Reactor — use hybrid approach with GibbsReactor in Python/Java
• Filter — not in JSON builder

When these are detected in a document, flag them as requiring hybrid approach and build the surrounding process in JSON, leaving these for manual wiring.

Error Message Actionability

The fromJson() error messages are structured and actionable:

• [FLUID_ERROR] → Check component names and mixing rule
• [UNIT_ERROR] → Check equipment type name and properties
• [STREAMNOTFOUND] → Check inlet reference matches a previously defined unit
• [NO_FLUID] → The fluid failed to create, cascading to dependent units

Test Results Summary (14 tests)

See examples/notebooks/testextractionworkflow.ipynb for the full test suite.