HULK DEFENSEUncrewed underwater vessels

How it is engineered

No vessel has been built yet, so the engineering record is the evidence. This page is that record — including the parts of it that are not flattering.

Method

One model. Everything else reads it.

master.json is the single source of truth. The CAD, the DXF, the Fusion build script, the bill of materials, the costing workbook, the engineering deck, the drawings and this website all read it and type nothing of their own.

Arrangement, structure, weight and buoyancy are solved together rather than in separate spreadsheets. Move a compartment and the ballast, trim and stability consequences follow automatically — which is how a change stays a change instead of becoming a rework.

Every build re-checks that nothing fouls anything else and that the vessel closes on weight and trim. If it does not, the build stops.

Machinery space — generator set, shafting and auxiliaries
Machinery space — generator set, shafting and auxiliaries

Self-checking

The CAD refuses to build if the design is wrong.

The Fusion build script aborts unless every V10 invariant holds: covers flush with the deck, trunk top meeting the hatch seat, compartment top meeting the trunk base, cover stiffeners fitting inside the trunk depth, and no cover overhanging the flat deck.

The hull is lofted through 17 rounded-rectangle sections whose offsets reproduce the model's own section table to 5 × 10⁻⁶. The lofted envelope volume comes out at 182.69 m³ against the model's stated 182.7 m³.

  • Covers flush with the deck
  • Trunk top meets the hatch seat
  • Compartment top meets the trunk base
  • Cover stiffeners fit inside the trunk depth
  • No cover overhangs the flat deck
  • Every arrangement box inside the hull envelope
  • No two boxes intersecting without a declared exemption

What the model emits

One file in. Every artefact out.

Change a plate thickness in the model and the CAD, the drawings, the bill of materials, the deck, the report and this website all move together on the next build. There is no second copy of any number to fall out of step.

That is why the engineering pack is internally consistent: not because it was proofread, but because nothing downstream is allowed to hold a figure of its own.

Generated from master.json

  • CAD — STEP AP242 assembly, external and internals
  • A parametric Fusion 360 build script with named components
  • Dimensioned general arrangement, plan and midship section
  • 172-line bill of materials from 226 researched components
  • 136-page engineering report, 90 tables, 6 appendices
  • 44-slide engineering deck
  • Every figure and drawing on this website
Alongside in the shed with a hatch pair open and two modules seated
Alongside in the shed with a hatch pair open and two modules seated

Sourcing

Specified against real components.

The bill of materials is built from 226 researched components across 172 costed lines in 14 systems, with 61 supplier references behind them — not from parametric rules of thumb. Every line names a part or a defensible analogue, so a reviewer can follow any item back to what it actually is.

At V10 the thirty highest-value lines that carried no published price were re-researched against live vendor pages and the sourcing re-recorded. Quantities are driven by the model rather than entered by hand, which means the material take-off moves with the geometry instead of being re-counted each revision.

Where no published price exists for a class of hardware, the line says so and names the vendors that were checked. That is more useful to a reviewer than a confident number nobody can trace.

Structural argument

Four load cases, and the one that governs is the fault.

Pressure balance means the compartment boundary's design load is a choice of architecture rather than a fact of the sea, so naming the cases explicitly matters more here than on a conventional vessel.

Two safety factors are applied. On the design case the allowable stress is yield divided by 2.00. On the emergency case it is yield divided by 1.25 — deliberately close to yield, because that case is survival of a fault that should not occur, not continued operation.

Pressure balance turns a 2.011 bar structural problem into a 1.00 bar one
Pressure balance turns a 2.011 bar structural problem into a 1.00 bar one
CaseDifferentialConditionGoverns
Normal±0.15 barThe compartment tracks sea pressure to within the control band throughout the dive.Governs nothing. An order of magnitude inside the design case.
Fault+0.50 barCharge control has failed high and the relief valve has lifted. The compartment is over-pressurised against the sea.Governs dog sizing and dog count. Uplift on one cover is 162.5 kN.
Design1.00 barThe structural design differential — twice the relief setting and 6.7 times the control band.Governs plate thickness and stiffener scantlings on the compartment shell.
Emergency2.011 barTotal loss of pressurisation at operating depth: the compartment at one atmosphere with the full head outside it. Survival only.Governs hatch cover thickness and stiffening. The cover is driven onto its seat at 66.6 t, and the dogs are unloaded.
Sea pressure seats the cover — 66.6 t on the seat at operating depth, dogs unloaded
Sea pressure seats the cover — 66.6 t on the seat at operating depth, dogs unloaded

Why a seated plug closure

In the emergency case sea pressure drives each cover onto its seat at 66.6 t, carried in bearing on the 100 mm seal lands all round at a mean bearing stress of 0.8 MPa. The dogs are unloaded; they only ever see load in the opposite direction.

The fault that most threatens the vessel loads the hatch in the direction the hatch is strongest, and unloads the hardware that is weakest. A cover that opened outward, or that relied on its dogging to resist the head, would have the opposite property. That is the whole safety argument, and it is one sentence long.

Assertions

Eleven of the fifteen things the model refuses to emit without.

Beyond the 38 containment and clash checks, the model guards the design directly. Each of these is evaluated on every run, and a failure stops the build rather than producing a warning somebody has to notice.

  • Compartment volume matches Configuration A to within 0.02 m³
  • A payload module passes the hatch opening and stands inside the compartment
  • No hatch cover overhangs the flat deck
  • Cover stiffeners fit inside the trunk depth
  • Dog load stays inside manual-class hardware
  • The masthead lands between 3 and 5 m above the water
  • The mast trunk fouls neither the genset pod, the control planes nor the compartments
  • The fixed-ballast channel is shallow enough to clear the trim tanks
  • Main ballast and compensating volume both fit inside the installed tankage
  • Trim closes to within 20 mm
  • BG stays above 0.25 m

Systems

Named hardware, not a block diagram.

Navigation, communications and the electrical interface are specified against researched parts with published lead times. What follows is the architecture; the part numbers, the suppliers and the lead times are in the pack.

Navigation and autonomy

Inertial navigation
Dual-antenna INS/AHRS with thermal calibration, rugged grade
Bottom-referenced velocity
Doppler velocity log, 300 m rated — the item that makes long-duration submerged dead reckoning viable
Obstacle avoidance
Multibeam forward-look imaging sonar
Altimetry
Single-beam altimeter and echosounder
Surface fixing and heading
Dual RTK GNSS receivers on survey-grade multiband antennas
Autopilot
Two independent flight controllers running ArduSub
Mission computer
GPU-class module running ROS 2 above the autopilots
Depth
Redundant extended-submersion pressure transmitters
Compartment monitoring
Leak and bilge sensing, distributed per compartment
Transverse section at station 6.30 m, looking forward
Transverse section at station 6.30 m, looking forward

Communications

Satellite broadband
Phased-array terminal on the masthead — the high-rate path for command and video offload
Cellular
Maritime LTE router with an external masthead antenna — the inshore path
Satellite short-burst data
Independent low-rate path of last resort, on its own antenna
Long-range LPWAN
Direct link to a mothership or shore station for telemetry and acknowledged waypoint command
Vessel identification
Class B AIS transponder with an integrated zero-loss splitter
Voice and DSC
VHF transceiver on the mast
Masthead architecture
Separate antennas with physical separation across the mast head. No single multiplexer exists that spans 150 MHz to Ka band, and specifying one would be an error
Surge protection
Per-coax, banded to the frequency of the run it protects
The three-stage mast at full extension against a harbour skyline
The three-stage mast at full extension against a harbour skyline

Electrical and interface

Compartment penetrations
SubConn / MCBH bulkhead connector sets — 44 bulkhead ends and 44 cable ends across the vessel
Payload environment
Dry, one-atmosphere junction boxes inside the compartment. Ordinary topside equipment travels as it is
Why not a wet stab plate
A pressure-balanced compartment fills with air, not water. Wet-mate blind connectors solve a problem this architecture does not have, and add a failure mode it does not need
Battery installation
Separate one-atmosphere pressure housings, forward and outboard, on anti-shock mounts
Fuel
Pressure-compensated — the tank sees no differential at depth
One compartment as a unit: the sealed box, two trunks, two covers
One compartment as a unit: the sealed box, two trunks, two covers

One thing the brief asked for that the hardware cannot do

The long-range link is specified for 10 km over water, which needs the slowest and most robust configuration the technology has. That yields roughly 50 bits per second of usable sustained throughput with 1.5 to 3 seconds of latency — one short telemetry frame every few seconds. It carries position, depth, heading, state of charge, tank levels, fault codes and an acknowledged waypoint command comfortably. It cannot carry a control loop, and describing it as a remote-control channel would be a misstatement of what the hardware does. Control-grade links run over the satellite and cellular paths.

Drawings

Every drawing is a view of the model, not a picture of it.

These are emitted by the same file that emits the CAD and the bill of materials. A dimension on a drawing cannot disagree with the same dimension in the structural calculation, because there is only one of it.

Structural arrangement — compartments, trunks, coamings and covers on the ring frames
Structural arrangement — compartments, trunks, coamings and covers on the ring frames
Eight ring frames, section-matched, 12 mm web 220 mm deep
Eight ring frames, section-matched, 12 mm web 220 mm deep
One compartment unit — the sealed box, two trunks rising to the seat, two covers flush with the deck
One compartment unit — the sealed box, two trunks rising to the seat, two covers flush with the deck
Tankage and ballast — eleven tanks with the fixed-ballast channel in the keel
Tankage and ballast — eleven tanks with the fixed-ballast channel in the keel

See the full drawing set

The same rigour goes into your configuration.

Tell us the payload and the mission and we will size it against the model rather than against a brochure.

Talk to engineering Request the record