Tutorial: build a propellant feed line

Build a pressure-fed liquid-oxygen line from an empty canvas (tank, feed line, main valve, and injector orifice discharging against chamber pressure) with CoolProp oxygen, run it steady, check it by hand, and see where the pressure goes. Run the two cases in Getting started first. Reference material is in manual §1.2 (network definitions), §3 (data structure), and §6 (interface).

Concepts you need

OpenFLUME represents a system as a directed graph (manual §1.2):

• Nodes are control volumes holding pressure, temperature, and density; mass and energy are conserved there.
  • An internal node has unknown P and T that the solver finds.
  • A boundary node has P and T imposed and acts as an infinite reservoir. Every network needs at least one, because it fixes the pressure level of the whole system.
• Branches connect two nodes and carry one unknown, the mass flow. The component sets the momentum relation ΔP(ṁ): pipe friction, valve, orifice, pump curve, and so on.
• Sign convention: each branch is drawn from → to. Positive ṁ flows in that direction; negative ṁ runs backward. ΔP ≡ P_from − P_to (manual §1.4).

Spatial resolution is up to you. One pipe branch has one flow rate and no interior profile; to resolve a gradient along a line, split it into segments (manual §3.13, Split…). A single-phase liquid feed line doesn't need that.

In steady mode an internal node's volume is ignored and its P and T serve only as the starting guess. A transient run needs a positive volume on every internal node (manual §3.2).

The system

LOX tank ──pipe──▶ valve inlet ──valve──▶ injector manifold ──orifice──▶ chamber
(boundary)        (internal)              (internal)                    (boundary)
 2.0 MPa, 90 K                                                           1.0 MPa

Element	Kind	Parameters (SI)
tank — LOX tank	boundary node	P = 2 000 000 Pa, T = 90 K
n1 — Valve inlet	internal node	initial guess P = 1 900 000 Pa, T = 90 K
n2 — Injector manifold	internal node	initial guess P = 1 800 000 Pa, T = 90 K
chamber	boundary node	P = 1 000 000 Pa, T = 90 K
line — Feed line	pipe	length 3 m, diameter 0.02 m, roughness 1.5×10⁻⁶ m
valve — Main valve	valve	area 3×10⁻⁴ m², Cd = 0.9, position 1
inj — Injector	orifice	area 3×10⁻⁵ m², Cd = 0.7

The numbers are illustrative and describe no particular engine.

Why the chamber is a boundary node. The chamber is a fixed back-pressure; its 90 K temperature is used only if flow reverses into the line. Liquid oxygen cannot mix into a hot-gas node, because OpenFLUME has no general mixing of unlike fluids at a junction other than a CEA reacting junction. For combustion, see the LOX/RP-1 thruster examples and combustion.md.

Step 1: start a new model and choose the fluid

1. Click New and type a name into the network-name field, such as LOX feed line.
2. Set Units ▾ to SI so the values you type match the table above. Any preset works; the stored model is always SI (manual §1.3).
3. Open the Setup tab.
   • Solver: leave Mode at Steady with the defaults (tolerance 1×10⁻⁶, 200 iterations, relaxation 0.9).
   • Fluids: in the default fluid's card, set the model to Real fluid (CoolProp), then search the catalogue for Oxygen, one of the curated favorites at the top of the list.

Which fluid model to use. The fluid catalogue lists 124 CoolProp HEOS fluids. Oxygen, Nitrogen, Hydrogen, ParaHydrogen, Methane, and Helium are among the nine favorites. RP-1 is not in the catalogue: CoolProp has no RP-1 equation of state. For a hydrocarbon fuel line, use an incompressible fluid with custom ρ, μ, and cp, or a catalogue hydrocarbon as a surrogate if you can justify it.

For a first water model, use Incompressible with the Water preset, which needs no CoolProp.

The CoolProp WebAssembly module (about 6.5 MB) loads the first time you select a real fluid (manual §4.2). If it fails to start, the status pill shows Fluid init failed.

Step 2: place the nodes

From the left rail, under Fluid nodes:

1. Click Boundary node, select it, and in the Properties panel set Label LOX tank, Pressure 2000000, and Temperature 90.
2. Add two Internal nodes, Valve inlet and Injector manifold, with pressure guesses 1900000 and 1800000 and temperature 90. These only seed the Newton iteration, but a sensible guess is the best help for a hard real-fluid case (manual §3.2).
3. Add a second Boundary node labelled Chamber, at 1000000 Pa and 90 K.

Drag the nodes into a left-to-right row; canvas positions don't affect the solution.

A real-fluid node's State variable can be switched from Temperature to Vapour quality. That isn't needed here: at 2 MPa and 90 K, oxygen is a subcooled liquid.

Step 3: connect the branches

Drag from the tank's connection handle to Valve inlet. On release, the Connect with chooser opens; pick Pipe. The new branch is selected with default parameters; set Length 3, Diameter 0.02, and Roughness1.5e-6, and leave Friction on Correlation (Swamee–Jain).

Repeat for the other two branches:

• Valve inlet → Injector manifold: Valve, area 3e-4, Cd 0.9, position 1 (fully open; position runs from 0 to 1).
• Injector manifold → Chamber: Orifice, area 3e-5, Cd 0.7.

To make several connections of one type, arm the component in the palette, then click a source node and a target node. The tool stays armed for the next connection (manual §6.4).

The status pill should now read Ready to solve. If it shows N issue(s) to fix, click it; each listed issue selects the offending element.

Or load it as text

The same model as a .fn file, generated with the project's serializer (serializeText in src/substrate/textProjection.ts), parses back without errors. Save it as lox-feed.fn and Load it, or open Text view, replace the contents, and click Apply.

// Fluid Network config v2
network "LOX feed line" {
settings: {"mode":"steady","tolerance":0.000001,"maxIterations":200,"relaxation":0.9}
fluid: {"model":"realFluid","params":{"fluidName":"Oxygen"}}
node "tank" boundary @ (0, 0) data: {"pressure":2000000,"temperature":90,"label":"LOX tank"}
node "n1" internal @ (200, 0) data: {"pressure":1900000,"temperature":90,"label":"Valve inlet"}
node "n2" internal @ (400, 0) data: {"pressure":1800000,"temperature":90,"label":"Injector manifold"}
node "chamber" boundary @ (600, 0) data: {"pressure":1000000,"temperature":90,"label":"Chamber"}
branch "line": "tank" -> "n1" pipe data: {"label":"Feed line","length":3,"diameter":0.02,"roughness":0.0000015}
branch "valve": "n1" -> "n2" valve data: {"label":"Main valve","area":0.0003,"cd":0.9,"position":1}
branch "inj": "n2" -> "chamber" orifice data: {"label":"Injector","area":0.00003,"cd":0.7}
}

The format has one record per line, in SI. Each record's data object holds every field that isn't already on the record line (manual Appendix C).

Step 4: run and check

Click Run. A single-phase liquid line converges quickly.

Hand check. The injector is far smaller than the line or the valve, so it should take almost all of the 1 MPa available. The liquid orifice law is ṁ = Cd·A·√(2ρΔP) (manual §4.3.2; the expansibility factor is ≈ 1 for a liquid). CoolProp gives ρ ≈ 1146 kg/m³ for oxygen at 2 MPa and 90 K. If the injector took the full 1 MPa:

ṁ ≤ 0.7 × 3e-5 × √(2 × 1146 × 1.0e6) ≈ 1.005 kg/s

The solved flow should come out slightly below this, because the line and the valve take a small share of the drop. When this tutorial was written, the model gave ṁ ≈ 0.996 kg/s with an injector ΔP of about 0.98 MPa; use those as a cross-check and confirm them on your own run.

Verify each component the same way:

• Valve: ΔP = ṁ² / (2ρ(Cd·A·p)²) (manual §4.3.6). With Cd·A·p = 2.7×10⁻⁴ m², that comes to a few kPa.
• Pipe: Darcy–Weisbach with a Swamee–Jain friction factor (manual §4.3.1). At about 2.8 m/s in a 20 mm line, the drop is on the order of 10 kPa.
• Mass balance: all three branches must carry the same ṁ.

Step 5: interpret the results

• Per element: click any branch or node. The Results block in the Properties panel lists every solved quantity: ṁ, ΔP, velocity, Reynolds and Mach numbers, etc. for branches; P, T, density, enthalpy, phase, etc. for real-fluid nodes.
• Where the pressure goes: on the Results tab, set the plot's x axis to Station along path, then add the node pressures. Off the time axis, each channel becomes a point along the network, so the plot is the hydraulic grade line and shows where pressure is lost (manual §6.8).
• Findings: under the plot, OpenFLUME lists deterministic readings of the result when there is something to report, such as the component dominating the pressure drop, branches running backward, or near-sonic flow. Here it should point at the injector.
• On the diagram: set Color by to Pressure drop or Pressure.

Step 6: change something

• Throttle the valve. Set the valve position to 0.3 and rerun. Flow barely changes: the valve's ΔP rises to tens of kPa, still small next to the injector's, as expected in an injector-dominated feed system. Sweep the position from 0.05 to 1 (manual §6.9) to see where the valve takes control.
• Raise tank pressure to 3 MPa. Flow should rise roughly as √(ΔP_injector).
• Compare runs: + Compare run… overlays earlier runs on one plot. Simulation variants keep several designs in one file (manual §6.12).
• Make it transient. Set Mode to Transient in Setup → Solver, give both internal nodes a positive Volume, and drive the tank Pressure Schedule or the valve positionSchedule. For a fast valve closure, lumped fluid inertia on pipes captures bulk surge, but OpenFLUME has no acoustic wave propagation, so it cannot predict water-hammer wave timing or peak pressure (manual §1.7).

Same model from Python

The .fn file you saved runs unchanged from the Python package (python.md):

import openflume as of

net = of.load("lox-feed.fn")
assert net.validate() == []
res = net.solve()
print(res.converged, res.branch("inj")["mdot"])        # kg/s
print(res.nodes_frame())                              # pandas, if installed

net.branch("valve")["component"]["position"] = 0.3     # edit and re-solve
print(net.solve().branch("inj")["mdot"])

Scope notes for feed systems

• Cavitation is checked only at a cavitatingVenturi closure, not on valves, orifices, pumps, or low-pressure liquid nodes. If you throttle a liquid line close to saturation, check local pressure against vapor pressure yourself.
• A pump is a pressure-rise-versus-volumetric-flow curve only, with no shaft-work term, so pump heating of the fluid is not captured.
• Two-phase flow uses the homogeneous equilibrium model (HEM), which matters if your line flashes or is still chilling down; see the chilldown tutorial.