Public-data rocket CAD.
A foundation for connected CFD.

What I built, and what comes next

I built an editable, component-based rocket assembly in Blender to create a public foundation for a connected thermal-fluid study. The reference is the publicly described Falcon 9 architecture: an exterior aerodynamic envelope, two stages of propellant tanks, representative tank internals, feed hardware, recovery appendages, and engine assemblies. Python controls the geometry, identifiers, presentation scenes, camera motion, material opacity, and export checks.

The completed deliverables are the CAD concept, geometry inspection records, and a 42.4-second assembly and component-tour film. External aerodynamic CFD, aerothermal loading, transient tank pressurization, diffuser flow, and a mission gas-budget model are the planned next stages. This film contains CAD geometry and presentation animation; its colors and transparency are not calculated pressure, temperature, velocity, liquid fraction, or stress.

This is my independent Khani Solutions portfolio project, built from public references and documented assumptions. It is grouped with the cryogenic methods in my Blue Origin experience because the engineering questions are related. It is not Blue Origin hardware or an employer deliverable, and it is not manufacturer CAD or a validated Falcon 9 performance model. No employer geometry, operating histories, or proprietary test data are used.

From public references to a parameterized assembly

The SpaceX Falcon User’s Guide anchors the reference architecture: a standard 70 m overall vehicle, a 3.66 m stage diameter, LOX/RP-1 propulsion, nine first-stage engines arranged around a central engine, and one vacuum engine on the second stage. It describes a common dome, a LOX transfer tube through the fuel tank, heated-helium pressurization, grid fins, landing legs, and nitrogen attitude-control hardware. Those public statements establish arrangement and scale, rather than a complete dimensioned manufacturing model.

I use metres throughout the Blender master. A project coordinate system places the longitudinal axis along +Z, with axial stations defining tank barrels, domes, interstage, payload adapter, fairing, and propulsion sections. This is a convenient modeling convention; coordinate transforms must be documented when comparing it with a published launch-vehicle frame or a solver frame.

Tank meridians, barrels, domes, shell thicknesses, local penetrations, and hardware envelopes are independently specified where public drawings do not resolve them. For example, the concept uses a 3.50 m tank bore, a 22 m first-stage LOX barrel, and a 12 m first-stage fuel barrel. These are study dimensions, not measured Falcon tank dimensions or published capacities. Detailed wall thickness, support locations, diffuser geometry, bottle positions, material properties, and thermal interfaces remain assumed inputs.

Each body has a stable component identifier and a group describing its role: exterior, tank module, plumbing, propulsion, payload, appendages, thermal protection, or analysis domain. The component register carries the material description, assumed or published evidence, geometric role, dimensions, and thickness where specified. This is the link between a visible component, a future mesh boundary, and a parameter in a system model.

The presentation scene uses split shells so that the internal assembly can be seen without destroying the master geometry. At the opening, the separated halves turn 90° about the rocket axis to expose their half profiles from the fixed camera. They return to their original orientation as they assemble. These cuts, rotations, exaggerated gaps, and assembly movements are explanatory devices; they are not a proposed manufacturing split or a flight separation sequence.

A future external-flow surface will be prepared from the assembled outer envelope, with the intended flight configuration and small-feature treatment declared. Tank CFD instead needs the enclosed fluid space, internal obstructions, and thermal wall regions. The shaded presentation mesh will not be passed directly to a solver as though it were already a qualified computational geometry.

Tank interiors and representative hardware

The concept includes common-dome and transfer-line features, annular baffles, anti-vortex elements, feed pickups, vent and pressurant lines, representative instrumentation, and perforated diffuser bodies. Engine assemblies include schematic bells, chambers, feed tubes, turbopump envelopes, and support features. Their inclusion shows where component-level questions belong in the architecture; it does not establish OEM internal geometry, combustion behavior, cooling-channel performance, or flight loads.

The baffle and diffuser openings are geometric cuts rather than painted dots. In the stored acceptance checks, example baffles have genus 13 and example perforated diffuser bodies genus 25, with zero flagged nonmanifold edges. These topology checks help distinguish real connected passages from a visual texture. They do not establish adequate cells through each passage, pressure-loss accuracy, or a physically correct jet model.

Fluid-domain construction subtracts immersed solid hardware from the gross tank envelope. One retained second-stage LOX example has a gross enclosed candidate volume of 55.7140 m³ and a net candidate volume of 55.4755 m³, a 0.23856 m³ hardware displacement. Those are volumes of this independently assumed geometry, not Falcon propellant-capacity measurements. The candidate still needs fluid connectivity, inlet/outlet, wall-region, clearance, and volume-mesh checks before a pressurization calculation.

LOX is the cryogenic liquid in this reference architecture; RP-1 is rocket-grade kerosene. A LOX pressurization model therefore has different phase-equilibrium and thermal requirements from an RP-1 fuel-tank model. A separate insulation/MLI concept can be introduced for thermal trades, but blanket construction, vacuum performance, seam losses, contact paths, and ground wetting or condensation must be specified independently. The optional MLI concept is not represented as confirmed Falcon hardware and is omitted from this film.

What the video shows, and how it was produced

The final film is an upright, fixed-angle tour. Main modules first join into one rocket. The exterior becomes transparent, and the camera translates and changes its orthographic scale to visit the fairing, upper-stage tanks, vacuum engine, grid fins, first-stage LOX tank, RP-1 tank, landing legs, and nine-engine cluster in top-to-bottom order. One component name appears at a time.

During each stationary close-up, the featured component group is fully opaque. Other bodies fade to 2.5% opacity or less, so a tank or engine can be identified without competing with the surrounding assembly. The pauses are doubled relative to the preceding revision. The camera then returns to a full view, the shell becomes solid, and the assembly opens back to its initial pose.

Python defines the motion analytically with smooth interpolation between camera and assembly states. Projected component bounds determine framing and margins; the camera orientation remains fixed and the rocket axis stays vertical in the image. The editable Blender scene retains transform, camera-scale, and material-opacity animation. A separate master preserves the original geometry and component metadata.

Rendering used Blender 5.0.1 and Eevee, followed by FFmpeg H.264 encoding at 1080 × 1080 pixels, 30 frames/s, with 1,272 frames and no audio or subtitle stream. Identical stationary poses reuse one rendered image, reducing flicker during the longer holds. The saved animation and the rendered sequence are checked separately. The decoded first and last MP4 frames are pixel-identical, and the file decoded cleanly across three repeated cycles.

The transparency is a display control. It must not be read as fluid content, a cutaway generated by a solver, or an actual material’s optical behavior. Blue and bronze distinguish modeled component groups; they are not field legends.

Planned external aerodynamics and thermal protection study

The external-flow study will use a dedicated assembled aerodynamic surface and a declared freestream operating point: Mach number, Reynolds number or density/velocity/length, static pressure and temperature, incidence, wall state, and appendage configuration. Subsonic, transonic, supersonic, and high-Mach cases will be treated as distinct regimes. An incompressible model is not an appropriate high-Mach substitute.

The initial compressible formulation will solve conservation of mass, momentum, and total energy with a documented gas-property model and turbulence closure. A steady RANS calculation is a possible first screening step; separated, unsteady wakes or shock motion may require a transient formulation. At higher thermal severity, the assumptions behind a calorically perfect gas, turbulence, transition, and near-wall treatment must be revisited.

The reference area, axes, force direction, and moment origin will be recorded for every case. Surface pressure, shocks, separation, skin friction, component forces, and moments will be inspected alongside residuals and force histories. Grid studies will address the fairing shoulder, interstage transitions, protrusions, grid fins, and wake; wall resolution and domain sensitivity will be selected for the closure being used.

Aerodynamic load convergence does not establish heat-flux accuracy. Thermal-protection work adds its own wall-temperature, transition, turbulence, property, and heating validation requirements. The planned thermal interface is a spatially and temporally resolved outer-wall heat-flux history, followed by transient wall/insulation conduction or conjugate heat transfer. Conductivity, thickness, thermal contact resistance, emissivity, radiation environment, and inner-wall boundary conditions become explicit inputs. A rendered TPS panel is not a calculated protection thickness.

Planned multiphase tank pressurization

The tank study starts with a defined fill fraction, initial liquid and ullage temperatures, pressure, gas composition, wall temperatures, pressurant supply state, vent logic, and acceleration history. For LOX with helium pressurization, the gas region contains a noncondensable pressurant and propellant vapor. Helium cooling and oxygen evaporation/condensation need distinct treatment; a single incompressible “air” phase would not capture their respective contributions to pressure.

The proposed 3D model combines a liquid/gas interface description with compressible ullage properties, energy transport, species transport, and conjugate wall heat transfer. Phase-change closure, saturation properties, surface tension, contact angle, interface resolution, and turbulence treatment must be selected and tested. Phase-change coefficients will not be accepted simply because they produce a smooth pressure trace.

Ground and orbital conditions will be separate cases. Ground pressurization begins with a gravity-dominated liquid pool and thermally stratified ullage. Low-acceleration or coast cases can be controlled by capillarity, wetting, residual acceleration, slosh, and jet/interface interaction. Removing gravity from a ground solution does not automatically create the correct orbital initial interface or wetted-wall state. NASA’s ZBOT research provides a candidate microgravity benchmark; matching the experiment’s fluid, geometry, acceleration and thermal conditions is a separate task from extrapolating to LOX.

Outputs will include P(t), pressure rise and decay, ullage temperature/species fields, interface position, liquid temperature, wall heat storage, phase-change mass transfer, integrated vent flow, and cumulative pressurant use. Species mass and total energy balances will be checked over the complete transient, including source terms and boundary fluxes.

The NASA collapse-factor review describes how heat transfer to cold walls and liquid, together with mass-transfer effects, can increase pressurant demand. For this study, the ideal reference will be declared with the same initial inventory, target pressure, ullage-volume history, pressurant species and supply-temperature convention. Mixed-gas partial pressures and the equation of state must be handled consistently. “Collapse factor” here is a gas-demand metric, not shell buckling or the pressure-drop coefficient of a diffuser. No collapse-factor value has been calculated for this concept yet.

Diffuser hardware: from resolved jets to a tank closure

The diffuser is the component-level bridge between the supply line and the ullage. Its parameter set will include inlet diameter, plenum dimensions, number and diameter of holes, open-area ratio, wall thickness, discharge directions, stand-off to the liquid interface, and supply pressure/temperature. The first comparison will change one geometric variable at a time while holding the tank state and gas supply condition fixed.

A resolved flow calculation will examine manifold maldistribution, per-hole mass flow, jet velocity and momentum, compressible/choked discharge where applicable, pressure loss, wall interaction, and whether the jets impinge on the liquid. For equal-area holes, the coefficient of variation σ(ṁ_i)/mean(ṁ_i) is one useful measure of discharge uniformity; the definition must change if the holes deliberately have different areas or target flows.

For an appropriate low-compressibility reference, K = Δp/(½ρU_ref²) can summarize pressure loss. Compressible discharge instead needs a consistent total/static-pressure convention, gas state, and mass-flow relation rather than an indiscriminate constant-K treatment. Pressure loss alone is not the tank objective: jet mixing, cooling, interface disturbance, and pressurant consumption must also be compared.

The existing perforated solid is a geometry starting point. A usable diffuser fluid domain needs plenum and external jet volumes, real inlet/outlet boundaries, sufficient resolution through each hole, and near-jet refinement. A porous or distributed-injection replacement would need calibration against a resolved case; visual perforations alone do not justify its coefficients.

Connecting the studies without mixing their assumptions

The intended architecture shares geometry and boundary-condition records across separate models. It does not force the full rocket, microgravity interface, and every diffuser hole into one enormous calculation. Each level can be verified first, then connected through quantities that conserve mass and energy.

External heating maps into a transient wall model. The inner-wall heat flux and wetted-wall state feed the tank calculation. A resolved diffuser study supplies discharge distribution or a tested reduced closure. The tank returns pressure response, thermal state and gas demand to a 0D/1D feed-and-pressurization network, which supplies valve schedules and inlet conditions for the next coupled iteration.

The system model will track gas-bottle depletion, regulator behavior, line heat exchange, valve losses, tank pressure targets, liquid withdrawal and vent events. A bidirectional coupling would also let the changing tank pressure affect diffuser and supply-line flow. Its time stepping, feedback stability, and convergence must be checked; exchanging one set of boundary conditions is only a one-way coupling.

Every exchange will carry units, coordinate frame, time base, interface area and sign convention. Heat-flux mapping should preserve integrated heat rate; gas-flow mapping should preserve species mass; interpolation must not invent negative inventories or lose vented mass. Ground ambient heat transfer and orbital radiation are different environments, so an ascent aero-heating result will not be reused as an orbital boundary condition without a justified mission phase.

The first useful connected demonstration will be a small, traceable reference case: one tank, one pressurant species, one diffuser, one wall boundary and a declared operating sequence. Added fidelity will follow comparison and sensitivity checks rather than being inferred from a more detailed render.

Evidence gates before reporting physics

Residual reduction, a visually stable interface, or a cleanly closed CAD surface is insufficient on its own. Reported drag, heating, pressurant use, pressure-collapse behavior, or a diffuser improvement will require the corresponding numerical and comparison evidence. The current result is a reproducible public-data foundation and a clear view of the engineering architecture.

Public references and project records

• SpaceX Falcon User’s Guide: architecture, dimensions, propulsion, tank arrangement and pressurization.
• NASA: A History of Collapse Factor Modeling and Empirical Data for Cryogenic Propellant Tanks.
• NASA: Zero-Boil-Off Tank experiments and microgravity pressure-control research.
• NASA ZBOT CFD validation record: candidate benchmark for the next stage.

Project and source review: October 2026. The downloads describe this independent concept and its presentation checks. No rocket CFD result or proprietary employer analysis is included.