Expertise

Equipment meant to work in the sea is usually tested by the people who built it, in a tank or off a quay. What is missing is someone who takes it out and uses it the way an operator actually does — in current, in a seaway, with cold hands and two metres of visibility — and writes down what went wrong.

I have a boat in Drøbak and the fjord outside the door. A prototype can be in the water the day after it arrives in the post, for the cost of diesel, instead of waiting for a vessel day or a slot at a test facility. What comes back is video, stills, measurements and a written assessment with findings ranked by importance — including the things that did not work.

Tests can be repeated across a season. That is the only way to learn anything real about biofouling, wear, and what happens to a part that is left in place.

Funnels, protective caps, cable and hose routing, camera and light mounts, tool holders, dummy stabs for dry training, full-scale training models. None of it is load-bearing, pressure-bearing or subject to certification — but it stops an operation when it is missing, and every workshop currently solves it with tape and a piece of aluminium.

The volumes are too small for the established suppliers, and their pricing shows it. What I sell is not the plastic, it is calendar time: from a phone call or a bad photograph to a physical part in your hand in the next period at home, with one or two iterations the same week.

Where the geometry is governed by ISO 13628-8 / API 17H the dimensions are public. Training models, blanking plugs and dummy versions can be built against them without access to the manufacturer's drawings — and without touching anything that carries load or pressure.

The Blueye X7 has seven guest ports, the X3 Ultra has three, and the BlueROV2 has open electronics and an established third-party market. The sensors exist and the platforms exist; what is missing every time is the adapter between them — the bracket, the skid mount, the tool holder, the cable route.

The combinatorics make it unattractive for everyone else: many platforms times many sensors gives hundreds of variants, each with a market of perhaps twenty units. Injection moulding is out of the question at that volume and CNC is too expensive. This is precisely the niche 3D printing exists for, and these parts are free-flooding and carry no pressure — which is where FDM is genuinely strong.

One concrete starting point: a light arm that moves the source 25–30 cm away from the lens. Almost all hobby and prosumer gear mounts the light next to the camera, which physically guarantees backscatter in particle-rich water. Parts are sold with photographs from real use in the fjord, not renders.

The requirements for escape openings and rotting cord are specified in millimetres, and enforcement is unforgiving — fines arrive without warning, including for pots bought in good faith in a Norwegian shop. From 1 February 2026 an escape hole is also mandatory in every pot used to catch fish, for commercial and recreational fishers alike. The Directorate of Fisheries has published fitting guides — that is, instructions for doing it yourself, not a product you can buy.

The kit is pure mechanics: escape rings in the correct dimensions, a holder for rotting cord that actually parts when it should, and untreated cotton cord within the requirement. No electronics, no CE marking, no certification. Printed in PP or PETG and fitted to the pot types actually sold in Norway.

The Institute of Marine Research estimates that recreational fishers lost around 8,330 lobster pots in 2024, and only 25 per cent of recovered ghost gear still had its rotting cord intact. The mechanism works less well than the regulation assumes. The margin per unit is small and the numbers are large — and there is nobody standing between the research report and the shop shelf.

In the summer of 2026 a two-year project began to clear the Oslofjord of ghost gear, funded with NOK 20 million from the Norwegian Environment Agency and using an AUV to map lost gear in the new no-fishing zones. Detection is being solved. Recovery is not: it is still manual and weather-dependent, and grapnels tear apart both the gear and the seabed.

What I can contribute is tooling that recovers without destroying, marking solutions that make lost gear easier to find again and return, and field trials of both from my own boat in my own area. Local knowledge of where people actually set their pots appears in no report.

From 1 July 2026 the threshold in the Norwegian procurement rules is NOK 500,000 excluding VAT. Research institutes, municipalities and county authorities can buy work of this size directly, without competition and without pre-qualification.

FDM parts are porous. Even at 100 per cent infill they take up water, because the layers are not chemically bonded, and pressure increases ingress. Published implosion data exists for SLA and SLS, but not for FDM — and not for the materials people actually run at home and in small businesses.

The programme is dull and cheap: identical coupons in PP, PETG, ASA, ASA-CF, PA-CF and TPU, weighed dry, pressure-cycled, weighed again. Creep under sustained load in cold water. Biofouling over a season. A pressure vessel made from steel pipe, or a water filter housing with a hand pump, costs under ten thousand kroner, and 8 bar corresponds to 80 metres — enough to separate usable from useless.

Everything is published with raw data, including the runs that failed. This sells nothing directly. The point is that the question is asked every week in the communities around Blue Robotics and Blueye without anyone having a numerical answer, and that whoever has the answer stops having to prove their competence again in every first meeting.

A structure can be formally correct against the standard and still be impossible to operate in current with two metres of visibility. The grab point is where the drawing says, but the manipulator cannot reach it; the camera line is blocked by an obstruction nobody considered; the torque tool will not fit at the angle it has to sit in.

The review is done from the using end and is systematic: access, grab points, camera lines, clearance, whether the interface is where ISO 13628-8 / API 17H says it should be, and whether a pilot who cannot see the whole structure at once can actually hit it. Delivered in writing, with findings ranked and sketches or photographs.

The same applies to handling on deck. What happens during mobilisation, in a seaway, with a cold glove, with the wrong tool, when someone is in a hurry. That failure mode rarely appears in the calculation.

The bottleneck in subsea autonomy is not algorithms but labelled data and failure-mode knowledge. No dataset tells a model what an incipient defect looks like in turbid water at ninety metres, how a grab failure develops in the second before it becomes an incident, or which scenarios an autonomous inspection will miss. It exists only in the heads of people who have seen it.

On the training side the need has just been formalised: IMCA added the Remote Pilot grade II level in April 2025 and updated the requirements for approved introductory courses in August 2025 with reference to simulator requirements. The simulator platforms are mature. Content is the bottleneck, because producing it requires operational experience the platform vendors do not have in-house.

The deliverable is written scenarios with triggering cause, what the pilot sees, the right and the wrong response, and how it ends — or labelled video and paid participation in a research project. This needs preparation: ten good scenarios is ten working days.