Quick answer: Offshore wind drone inspection now runs two ways. Remote docks, like the vehicle-mounted dock KDDI Smart Drone tested in Akita in March 2026, check a turbine's blades in 30 to 60 minutes without a boat. Close-range piloted flights still produce the millimetre-level imagery needed to grade defects and plan repairs.
Japan's 2040 offshore wind target implies about 3,600 turbines, and a truck-mounted drone dock in Akita checks one in 30 to 60 minutes: what it still misses
A drone dock bolted to a vehicle can now inspect offshore wind blades from shore in 30 to 60 minutes per turbine, with no boat and nobody standing at the launch point. KDDI Smart Drone proved it with an Akita offshore operator in March 2026. My read: remote docks are now the fastest restart check in offshore wind, and repair decisions will still come from close-range flights plus a reporting pipeline that can keep up.
I fly blade inspections offshore and onshore in Taiwan, and I build AI tools that turn inspection imagery into reports. So I read the Akita result with two hats on. As a pilot, I'm impressed. As the person who has to decide what a smudge on pixel 4,000 actually means, I have questions.
In Japanese, this whole field gets searched as 洋上風力 点検 ドローン: offshore wind, inspection, drone. The search volume is growing for a simple reason. Japan has a lot of turbines to build and very few people who can climb them.
What did the Akita vehicle-mounted dock trial actually prove?
It proved that an automated dock on a truck can launch, fly a preset route to offshore turbines, and bring back blade exterior imagery, all triggered remotely from an office. The trial ran from March 23 to 26, 2026, and the results were published in June.
The public details worth knowing:
- The dock sits on a vehicle and handles takeoff, landing, and battery charging on its own. The release photos show a DJI Dock 3.
- Operators started flights from the office. No pilot drove to the launch site.
- Blades were inspected while slowly rotating or briefly stopped.
- Each turbine took 30 minutes to 1 hour, from the vehicle arriving to the inspection finishing.
- Automatic slowdown and return near restricted airspace worked as designed.
The business case is downtime. Some exterior checks used to require stopping generation until a crew could get out there. After a lightning strike or a storm, that wait can stretch across days of bad sea state. Japan's offshore wind planning assumes a capacity factor of roughly 30%, according to NTT's R&D background on the topic, so every lost hour hurts.
Why is offshore wind blade inspection so hard in the first place?
Because the turbine is kilometres out to sea and the weather decides your schedule. A conventional offshore inspection needs a crew transfer vessel, a sea state the boat can handle, and wind the drone can fly in.
Enterprise inspection drones in the DJI Matrice 300 and 350 class are rated for maximum winds of about 12 to 15 m/s. Offshore, that ceiling arrives fast. Stack a wave limit for the vessel on top of a wind limit for the aircraft, and the usable window shrinks to a few days in some months.
The blades themselves keep growing. Current offshore rotors use blades longer than 100 metres, which means more surface, more images, and more places for damage to hide.
Japan adds its own problem: winter lightning on the Sea of Japan coast, where Akita sits. Lightning damage is the classic reason a turbine stops and waits for someone to go look.
Remote dock vs piloted close-range flight: which one do you need?
You need both, for different questions. The dock answers "is it safe to restart?" quickly. The piloted close-range flight answers "how bad is it, and what do we repair?"
| Method | Where the operator is | Best use | What you get | Main limit |
|---|---|---|---|---|
| Rope access | On the blade | Repair, tap testing, hands-on checks | Touch and direct measurement | Slow, turbine stopped, labour shortage |
| Piloted close-range drone | On a vessel or transition piece | Annual condition survey, defect grading | High-resolution images, around 1 mm per pixel is a common target | Needs vessel and weather window |
| Vehicle-mounted remote dock | Onshore office | Post-storm and post-lightning triage, restart checks | Fast exterior overview on a fixed route | Range from shore, preset routes, stand-off distance |
| Two-drone radio sensing (NTT research) | Remote | Detecting structural change between flights | Radio signal changes, not photos | Still research stage |
| Contact LPS drone | Nearby | Lightning protection continuity | Electrical measurement | Specialist equipment and training |
The NTT approach is worth a sentence. It uses 2 drones as a weak-signal transmitter and receiver, with the blade sitting in the radio path between them, and looks for changes in the received signal over time. It detects that something changed. It won't tell you what the crack looks like.
Contact-based lightning protection checks by drone have also been demonstrated in Japan. That matters because receptor and down-conductor continuity is invisible to any camera.
What do remote inspections still miss?
They miss what needs angle, distance, or contact. A dock flying a preset route from shore is excellent at spotting obvious damage. Grading needs more.
The gaps I'd watch:
- Early leading-edge erosion. The first stages show up as subtle texture changes. You need consistent light, close stand-off, and the same angle every year to track progression.
- Trailing-edge splits and bond line cracks. These often only show from specific angles. A fixed route may never point the camera where the crack is.
- Surface coverage on rotating blades. A slowly turning rotor gives you a different blade position on every pass. That's fine for triage and painful for year-over-year comparison.
- Lightning protection faults. A receptor can look perfect and still be disconnected inside.
- Internal damage. Delamination and spar cap issues don't appear on the outer shell until they're serious.
None of this makes the Akita result less useful. It defines the job. Docks increase inspection frequency. Close-range flights and hands-on work keep the depth.
Where does the real time go in a drone blade inspection?
In the report. The flight around one turbine takes well under an hour. Turning that flight into a signed defect report can take far longer.
A full close-range survey of 3 blades, 4 surfaces each, can produce hundreds to more than a thousand images depending on overlap and stand-off. Every image has to be checked. Every defect needs a location on the blade (root, mid-span, tip; leading edge, trailing edge; pressure side, suction side), a size, a severity category, and a comparison against last year.
That's the part I automate. In my own pipeline the AI does the first pass:
- sorts images by blade, surface, and span position
- flags candidate defects and drafts a severity suggestion
- matches each finding against the previous inspection of the same turbine
- writes the first draft of the report
A human inspector still confirms every finding and signs the grade. The AI removes the hours of scrolling. It doesn't get the final word.
Here's my point of view on docks: put 10 of them on trucks along a coastline and you multiply the images coming in. If the reporting side stays manual, you've just moved the bottleneck from the boat to the office. The operators who win this will be the ones who fix both ends.
Who should own the inspection data?
The asset owner should. Raw images, flight positions, defect labels, and severity history belong with the turbine, in storage the operator controls.
This sounds obvious until you switch inspection vendors and discover that 3 years of labelled defect history lives inside someone else's platform in a format nobody else can read. Year-over-year comparison is where most of the value sits. Losing it resets you to zero.
My rule for any program I build: open file formats, a fixed defect taxonomy, and exports the client can take anywhere.
What does this mean for Japan's offshore wind build-out?
It means inspection capacity has to scale faster than the workforce. Japan's Offshore Wind Industry Vision targets about 10 GW by 2030 and 30 to 45 GW by 2040. NTT estimates that implies roughly 3,600 offshore turbines along Japan's coasts by 2040, based on 12.6 MW class machines from the first auction round.
Japanese industry is already moving. KDDI's own newsroom notes that much blade inspection still relies on rope access at height, and that the sector needs to cut manual labour. NEDO completed a research program in November 2023 aimed at fully automating drone blade inspection. Kansai Electric announced plans for a 5G-connected drone inspection service. Underwater drones are going after foundations and cables.
My goal is to work in Japan's offshore wind sector, bringing what I've learned on Taiwan's offshore and onshore sites. Taiwan built its offshore fleet a few years ahead of Japan, and the operational lessons transfer: weather windows, vessel logistics, and the reporting load that follows every campaign.
How should an operator structure a blade inspection program in 2026?
Use 3 tiers and keep the data in one place.
| Tier | Trigger | Method | Output |
|---|---|---|---|
| 1. Triage | Lightning alarm, storm, SCADA fault | Remote dock or quick piloted flight | Restart or hold decision |
| 2. Condition survey | Annual or campaign-based | Piloted close-range drone, all 4 surfaces | Graded defect report with history |
| 3. Intervention | Severe or progressing defect | Rope access, internal inspection, LPS test | Repair scope and sign-off |
Then add the boring parts that make the whole thing work:
- one defect taxonomy across every vendor and every year
- consistent flight patterns so images line up season to season
- AI pre-sorting with a human sign-off on every grade
- raw data stored under the operator's account
The Akita trial shows the first tier can run from an office. Tiers 2 and 3 still need people close to the blade. If you run offshore or onshore assets in Taiwan or Japan and want to talk about the reporting side, that's the part I work on every day.