Swamp Launch Rocket Design and CFD

I built the Swamp Launch IREC competition vehicle in SolidWorks, the geometry that everything else keys off of. The same CAD model became the basis for the CFD study and the wind tunnel test planning, so the simulation and the physical test were checking the same shape.

I achieved drag estimates within 15% of RASAero predictions by developing mesh-independent ANSYS Fluent subsonic CFD models on UF's HiPerGator supercomputer, building Cd-versus-Mach tables from Mach 0.1 to 1.0 and optimizing nosecone, fin, and boattail geometry.

I refine the mesh and boundary-layer setup around the rocket before each run and documented the workflow for each regime: physics setup, wall Y-plus targets, drag calculations, and which equations apply, so it is repeatable by the rest of the team.

I authored Mach Ranges, the team's standardized CFD documentation, which takes a rocket through a complete ANSYS Fluent simulation at every speed from Mach 0.1 to 1.0 in steps of 0.1. Each Mach number gets the same 42 numbered steps in four stages: understanding the aerodynamics, geometry setup, mesh setup, and physics setup. Anyone on the team can pick up a case and run it the same way.

The document tracks how the setup has to change as the flow does. Up to Mach 0.3 the flow is treated as incompressible with constant density and viscosity. From Mach 0.4 the energy equation turns on, with ideal-gas density and Sutherland's law for viscosity, and the inlet and outlet become a pressure far field. At Mach 0.9 and 1.0 the flow goes transonic, so the regions of interest grow from the wake to the nose cone and fins, where normal shocks and a bow shock start to form. The boundary layer is sized for a wall y+ of 30 to 300, so the first-layer height and layer count are recalculated at every speed.

A companion Cd-vs-Mach process turns the playbook into a production run. It groups the ten cases by geometry (full rocket for 0.1 to 0.3, a symmetric half-model without rail buttons for 0.4 to 0.7, full rocket again for 0.8 to 1.0), sets a naming convention so several people can split the runs, and ends in the drag-coefficient curve the team uses. The excerpts below are public versions: competition-rocket operating conditions and dimensions are redacted, and most pages are left out.

To check the CFD against reality, I designed a dual-load-cell wind-tunnel test assembly with precision-machined couplings and Arduino-based real-time data acquisition, then mounted the rocket in a subsonic wind tunnel so measured drag could be compared directly against the ANSYS CFD data.

I authored the team's Wind Tunnel Testing Process, which records how the test was designed, run and analyzed, down to its assumptions and limitations. The model is quarter scale, the largest that keeps the rocket under one tenth of the test section, and is matched to flight by Reynolds number. It was resin printed in sections and sanded to 330 grit so the surface roughness scales with the model. The sting is kept under 30% of the rocket diameter and checked for sag, and a CFD run of the tunnel domain puts the rig's own error at about 6%. The document closes with an honest debrief: vibration in the cantilevered mount limited the useful test speed, and it lists the changes for the next rig.

As Responsible Engineer for flight-dynamics simulation and aft-plane pressure sensing, I validate the models against flight data and refine the boattail instrumentation to calculate pressure drag, supporting the team's 10,000-ft IREC apogee target.

That validation follows the team's Flight Data Analysis Process, which I helped with. After burnout, the flight computer's inertial (east-north-up) accelerations, roll, pitch, yaw, pressure and temperature are integrated in MATLAB to give velocity, position and Mach number. The deceleration along the velocity vector gives drag force and Cd. A 3-2-1 Euler rotation into the body frame gives angle of attack and sideslip. The excerpt shows the method and a drag result from a 2026 test launch; the code, vehicle mass and control-system data are left out.