Ethan Sheehan

Ethan Sheehan — Aerospace engineering at the University of Bristol. Propulsion Lead, HyPower Bristol.

Aerospace engineering at the University of Bristol Bristol, UK

Engine 1, Test 3 at steady state on the Airborne Engineering thrust stand

Scroll
Everything, at speed 16 clips — 7 programmes

I build and fire liquid rocket engines.

MEng Aerospace Engineering student at the University of Bristol and Propulsion Lead at HyPower Bristol, the university's rocketry team.

I led BristolSEDS Experimental Propulsion from Engine Lead to Propulsion Lead across four engine programmes, including the 5.94 kN regeneratively-cooled engine that set the UK student thrust record in January 2025.

5.94 kN
UK student thrust record, Jan 2025
>80,000 N·s
total impulse fired
6
hot fires, Engine 1
£120k+
funding & sponsorship secured
BSEP Engine 1 — Maleficent

Ignition, slow motion. The plume establishing on Test 3.

01 — Propulsion: BSEP

Four engines, a torch igniter, and the fires that proved them.

FIVE
Test 1 hot fire from the AEL cell camera: ignition, a 3 s burn and shutdown
01 — Engine Lead — 2024-25

BSEP Engine 1Maleficent

Regeneratively-cooled Inconel 718 bipropellant engine. 5.94 kN in January 2025: the UK student thrust record.

Peak thrust
5.94 kN
Chamber pressure
23.55 bar
Hot fires
6
Full case study
Cell camera view of the burn: a yellow ignition plume turns purple, then a fireball, then the engine shuts down
02 — Propulsion Lead — 2025

BSEP Engine 2Snow White

Lightweight CP1 chamber with a coaxial-swirl injector: the first use of the Al-Fe-Zr alloy in a liquid rocket engine.

Burn
5.0 s full duration
Average thrust
1.59 kN
Chamber mass
~3x lighter
Full case study
X-ray CT section sweep following one spiral cooling channel around the chamber wall, the wall appearing as a bright branched section with channel voids on a white background
03 — Propulsion Lead — 2025-26

BSEP Engine 3Rapunzel

Spiral-channel regen chamber for EuRoC 2026: designed, printed in CP1 aluminium and CT-scanned end to end, waiting on its first hot fire.

Design thrust
5 kN
Channels
56 spiral
Peak wall temp
-31.6 K at 27°
Full case study
CAD render of Engine 4: the isogrid-latticed chamber and nozzle, three-quarter view from the injector end
04 — Propulsion Lead — 2025-26

BSEP Engine 4Elsa

Bristol's first LOx/IPA engine, for Race 2 Space 2026.

Design thrust
5 kN
Burn
8 s
Oxidiser
LOx
Full case study
Close GoPro clip of the final burn: the igniter runs for several seconds, then flame breaks out of the body wall and a shower of sparks erupts.
05 — Igniter designer, HyPower Bristol — 2024-25

BSEP Torch IgniterTuppence

Hydrogen/oxygen torch igniter: eight firings, retired after melting through 5 mm of stainless. Succeeded by Pascal.

Firings
8
Inlet pressure
3→12 bar
Wall
5 mm melted through
Full case study
02 — Flight campaigns: HyPower

Taking the engines to the stand and to Portugal.

TWO
Wide 4K view of the test bay from outside: the sparkler ignites in a shower of sparks, a short orange plume streams out of the bay, then white vapour as the oxidiser runs out
06 — Propulsion Lead — September 2025

Integrated Test Campaign 2025

Three burns in September 2025: a 7 s full-duration burn measured at 3.99 kN peak, and a 5.58 kN peak on the last.

Burns
3
Full-duration burn
7 s
Peak thrust
5.58 kN
Full case study
Handheld pan along the flight propulsion system on the bench: the printed chamber and bell nozzle, the injector and valve block with its feed lines and pressure transducers, then the white tank sections and their bulkhead plumbing
07 — Propulsion Lead — 2025

EuRoC 2025Waxwing

HyPower's 2025 vehicle at the European Rocketry Challenge: 25 teams, first of the UK entrants, grounded 800 m from the pad.

Teams
25
UK entrants
1st
Vehicle
83 kg, 4 m
Full case study
03 — Control systems

Controllers that had to run on real hardware.

TWO

A gain-scheduled cascade PID on an STM32, simulated and then verified bit-exact into the flight C before it was allowed near a bench. A phased LQR that lands a thrust-vectored rocket after its chute has already opened, scored over three hundred Monte Carlo runs rather than one good one.

The thread is the same as the propulsion work: build the loop, then build the rig that can prove it holds. Two more control units, an LQR stability augmentation for a Cirrus SR22 and a hand-written GA and PSO tuning spacecraft attitude controllers, sit under Coursework.

Bench telemetry for the 7 bar test: supply pressure decaying from 15.9 to 10.4 bar, regulated pressure rising to about 6.5 bar against the 7 bar setpoint, and the servo transient
08 — Simulation and flight-code pipeline — 2026

BSEP E-Reg

A software-defined pressure regulator for HyPower's pressure-fed rockets: a servo ball valve and a gain-scheduled cascade PID on an STM32, simulated, verified bit-exact into flight C, then bench-tested.

Best hold
3.96 / 4.0 bar
Replay
5001/5001 bit-exact
Tank held
55 bar
Full case study
Descent and landing of Monte-Carlo run 59: altitude above ground against crossrange, the vehicle drawn at its own logged attitude every few seconds with the plume shown where thrust is commanded, flaring to touchdown 660 m downrange
09 — Year-3 individual research project, supervised by Prof. Tom Richardson — 2025-26

RP3: TVC Landing Simulation

Landing a thrust-vector-controlled sounding rocket after main-chute deployment: a 3-DOF Simulink simulation, a phased LQR controller and a 300-run Monte-Carlo campaign.

Monte-Carlo success
209/300
Best landing pitch
−0.63°
Mean propellant
12.6 kg
Full case study
04 — Machine learning

Reinforcement learning, computer vision, and the tests that scored them.

FOUR

Reinforcement learning, computer vision, and the instrumentation that says whether any of it actually worked. A CNN that looked right until a test-set leak was found and killed. A gauntlet of 140,000 games per arm before an agent was allowed to ship. A policy held out against twelve unseen seeds. Live crack-tip inference at about 25 fps on a single Jetson.

What interests me is the measuring apparatus around a model: what it is scored against, how it behaves on data it has never seen, and how you find out that a good-looking result is wrong.

Third-person Unreal chase view of the minidrone rounding the corner of the red L-track above a grey arena floor with yellow and blue decoy squares
10 — RL path-planning policy, the team's primary submission — 2026

Optimal Line-Following DroneBOLD

A reinforcement-learning path planner for the MathWorks Minidrone Competition Europe 2026: a Parrot Mambo follows a red line and lands on the marker, with the policy shipped as baked weights and zero toolbox dependency.

Tracking RMS
0.0438 m
Unseen seeds
12/12
Landing
3-20 mm
Full case study
Delaminating composite specimen edge with the detected crack tip marked, on a deep purple glow
11 — Research intern — 2026

ILK DresdenTorchCrack

Live AI crack-tip detection for composite fracture tests, running on a Jetson AGX Orin. Summer 2026.

Live rate
~25 fps
Frame
5 MP
Compute
1 Jetson
Full case study
Pan across the tensile rig during a live test, from the Shimadzu frame to the acquisition monitor
12 — Research intern — 2025

Bristol Composites Institute

Curvature-to-stress prediction for 3D-printed flax/PLA hygromorph strips: 72 specimens, 10 batches, a test-set leak found and killed, and a paper on why the shape transfers and the scale does not.

Master-curve R²
0.92
Specimens
72 / 10 batches
Self-calibration R²
0.74
Full case study
Grouped bar chart of four reinforcement-learning arms: each temperature-1 sampled-policy gain of several percentage points stands beside an argmax gain of about zero, with the no-reward control lowest
13 — Environment and evidence, two-person team — 2026

Pokémon TCG Agent

Kaggle Pokémon TCG AI Battle, summer 2026. The agent was three behaviour-cloned networks and a shallow search; the product was the measurement instrument that reversed six of our own decisions before they shipped.

Games per arm
140,000
Sampled gain
+7.48 pp
Argmax gain
+0.20 pp
Full case study
05 — Aerodynamics and ground systems

CFD validation, and the ground software a team inherits.

TWO
Split-field plot of the ARIA nosecone solution at Mach 1.8: numerical schlieren above the axis, Mach number below, with the exact Taylor-Maccoll conical shock at 34.43 degrees drawn as a red dashed line lying on the computed shock
14 — Sole author: CFD, V&V campaign, paper — 2026

Nosecone CFDARIA V&V

Density-based OpenFOAM simulation of the ARIA sounding-rocket nosecone at Mach 1.8, validated against NACA wind-tunnel data to within the tunnel's own uncertainty.

Mean |ΔCp| vs wind tunnel
0.0040
Apex p/p∞ vs exact theory
+0.2 %
LTS speed-up
12.7×
Full case study
The AVDASI 2 docs landing page in July 2025: Flight Lab red header, a paragraph explaining the auto-synced code repo, and Avionics, Aerodynamics and Structures sections in the left nav
15 — Intern — Summer 2025

Bristol Flight Lab

AVDASI2 avionics teaching resources: the Cube guide, an example ground station and the docs site the Flight Lab still runs.

Org commits
41 / 123
Docs site built in
28 commits
Images surviving
29 / 29
Full case study
06 — Coursework

The units behind the rest of it.

SEVEN
Root locus of the F-16 longitudinal poles as the centre of gravity moves from 25 to 35 percent of mean aerodynamic chord
16 — Individual coursework — 2025

Flight Dynamics & Advanced Control

A Cirrus SR22 hand-flown in X-Plane and reduced to five dynamic modes against the military handling standard, then an LQR stability augmentation system that flies an F-16 with its centre of gravity pushed past the neutral point.

Dynamic modes flight-tested
5
Cruise range, sim vs POH
976 vs 941 NM
Unstable CG stabilised
35% MAC
Full case study
Root locus of a two-degree-of-freedom aerofoil coloured by airspeed; the second mode crosses into the right half-plane at 100 m/s, marked as the flutter speed
17 — Individual coursework — 2025

Dynamics and Control of Linear Systems

Flutter speed of a two-degree-of-freedom aerofoil read off its root locus, then spacecraft attitude controllers tuned by a hand-written genetic algorithm and particle swarm in MATLAB and Simulink.

Flutter speed
100 m/s
Controller tunes compared
10
Cost vs baseline tune
320x lower
Full case study
Hand-drawn artefact study on grid paper of a SpaceX Merlin 1-D: side, front and bottom views annotated for the turbopump, preburner, combustion chamber and regeneratively cooled nozzle, in a University of Bristol title block
18 — Group challenge and individual portfolio — 2023

Engineering by Design

EWB Challenge for a Cambodian village: a 50-drone delivery fleet chosen by weighted pairwise comparison, and a week-7 pencil study of a Merlin 1-D.

Options weighed
5
Winning score
3.18
Fleet
50 drones
Full case study
Two panels: measured fundamental frequency against overhang length on the cantilever curve at 175 GPa with the 190 GPa reference curve, and the nine Young's modulus values against the 175 GPa mean and 190 GPa stainless lines
19 — Oct 2023 - Apr 2024

Engineering by Investigation

Year 1 lab unit. Young's modulus of a steel ruler to within 8%, from a microphone, a breadboard preamp and a Raspberry Pi Pico logging at 5 kHz.

Young's modulus
175 GPa
Sampling
5 kHz
Error vs 304 stainless
-8 %
Full case study
Space Shuttle Atlantis lifting off Pad 39A at Kennedy Space Center on a column of exhaust, 12 November 1995
20 — Individual coursework — Oct-Nov 2024

Space SystemsSTS-74

STS-74 mission study: requirements, GMAT orbit, Hohmann transfer, configuration trade and power sizing for the Mir Docking Module.

Orbit period
91.5 min
Array
13 m² for 235 W
Hohmann Δv
90 m/s
Full case study
The finished BlueBird UAV on a bench trolley in the lab: full blue film-covered wing across two benches, silver instrument pod under the carbon fuselage spar, students in maroon lab coats working around it
21 — Avionics test lead, group design project — 2024-25

AVDASI 2: BlueBird AvionicsBlueBird

Avionics test lead on a six-person team: Hall-effect flap sensing, servo calibration, aircraft integration and three days in the wind tunnel.

Flap angle accuracy
1 deg bench, 1-3 deg tunnel
Tunnel campaign
3 days at 20 m/s
Wing vs prediction
20-40% down
Full case study
Integrated Test Campaign 2025

Hotfire 2, the full-duration burn. Shock diamonds hold the length of the plume.

Index

All 22 projects, filtered by topic.

22 shown
About

Ethan Sheehan

Bristol
Ethan Sheehan

MEng Aerospace Engineering student at the University of Bristol and Propulsion Lead at HyPower Bristol, the university's rocketry team.

I led BristolSEDS Experimental Propulsion from Engine Lead to Propulsion Lead across four engine programmes, including the 5.94 kN regeneratively-cooled engine that set the UK student thrust record in January 2025.

Before this year: a summer research internship at ILK Dresden on live crack-tip tracking, and a summer at the Bristol Flight Lab building the avionics teaching resources the unit still runs.

DegreeMEng Aerospace Engineering, First class
InstitutionUniversity of Bristol
Dates2023-2027
PrizeRoderick Collar Prize
BeforeA levels: Maths, Further Maths, Chemistry, Physics. British International School of New York, 2021-2023
NowMEng year 4 at Bristol, Propulsion Lead at HyPower