Wind
Blade, tower and nacelle inspection where access is expensive and weather windows are short, with agents planning campaigns around both.
Access is the cost, not the repair
Getting a technician onto a blade is a rope team, a weather window and a stopped turbine. The inspection itself takes minutes. Everything expensive is the getting there, which is why blades get looked at annually whether or not anything has changed, and why a crack that appears in month two waits for the annual campaign.
Four failure modes, and what each one costs you late
Wind is the sector where the inspection decision and the access decision are the same decision, so the cost column is usually about timing rather than about the defect.
Leading-edge erosion
- How it shows
- Roughening and material loss along the outer third of the blade
- What it costs late
- Yield loss that grows quietly, often several percent before anybody schedules a repair
- What catches it
- RGB at blade resolution, compared against the same blade in earlier passes
Lightning strike damage
- How it shows
- Burn marks at the receptor, sometimes with delamination behind it
- What it costs late
- A strike that damaged the down-conductor can end in loss of the blade
- What catches it
- Capture planned from the recorded strike event rather than waiting for the annual campaign
Cracks and delamination
- How it shows
- Surface cracking along bond lines, trailing edge splits
- What it costs late
- Propagates under load, and the gap between a repair and a replacement is months
- What catches it
- Classified from close-range capture and tracked pass to pass for growth
Tower and foundation corrosion
- How it shows
- Coating breakdown, rust staining, grout deterioration at the base
- What it costs late
- Structural over a long horizon and expensive to access once advanced
- What catches it
- The same flight covers tower and transition piece, with findings attached to the turbine
Severity here is how we rank the class by default, not a judgement about your site. Every threshold is set per deployment, and a mode you consider routine can be dropped down the ranking on day one.
Six things the layer runs on a wind fleet
Everything here is shaped by one constraint: access is expensive, so the job is to know exactly what is worth stopping a turbine for.
Plan capture around access and weather
Turbines grouped into one window rather than visited one at a time, scheduled against the forecast and the maintenance calendar you already run.
Capture blades at defect resolution
Each surface at the resolution its defect classes actually need, so erosion, cracking and lightning damage are classified rather than noted.
Measure growth, not just presence
The same blade compared against earlier passes. A category that has not moved in two years is a different decision from one that has.
Trigger capture from what happened
A lightning strike recorded in SCADA plans an inspection of that turbine, instead of the damage waiting for the annual campaign.
Group work into windows
Findings below the threshold for their own campaign are held and bundled, so one access covers three turbines rather than three accesses covering three.
Keep the blade history
Repairs, categories and passes per blade, so warranty and end-of-life decisions are argued from a record rather than from memory.
Erosion that did not need its own campaign
Not everything found has to be dispatched. The useful decision here was to group the repair into an access window that was already happening, which is a decision the layer can only make because it knows what else is outstanding.
Leading-edge erosion, blade B
WTG-14 · blade B
- category
- 3 of 5
- extent
- 2.1 m
- growth since
- 2025-04
- standard
- IEC 61400
Grouped into the next access window
Below the threshold for a campaign of its own, so it was grouped with two other turbines and proposed for an access window already scheduled.
Identifiers and values are illustrative.
How a wind deployment starts, and then repeats
The sequence is built around access windows, because that is the expensive resource in this sector.
Connect the fleet
Read-only into turbine SCADA, including alarm and strike events, plus whatever blade history you already hold. Outbound only, agreed with your operations team.
Baseline every blade
One full pass across the fleet, categorised per blade. Without it there is no growth rate, and growth rate is the number this sector actually decides on.
Run annually, and on events
Blades annually, tower and foundation on condition, plus capture triggered by recorded strikes. Findings are grouped into the next access window rather than raised one at a time.
Widen the limits when the record earns it
Grouping and scheduling are usually the first limits a wind team opens, because the decision is bounded and the cost of being wrong is a rescheduled window rather than a turbine.
What lands on your side after a campaign
Built for the two conversations this sector actually has: what to repair now, and what the warranty position is.
Blade condition record
Every surface captured, categorised and positioned along the span, per blade and per pass.
Web viewer · PDF
Growth rate per finding
Extent and category compared against earlier passes, which is the number a repair decision is actually made on.
CSV · API
Grouped work plan
Findings bundled into access windows, with the turbines and the scope for each one.
CSV · O-OPS
Event-triggered inspections
Capture planned from recorded strikes and alarms rather than from the annual calendar.
O-AGENT · SCADA
Tower and foundation condition
Coating, corrosion, grout and fixings from the same flight, attached to the turbine.
Web viewer · CSV
IEC 61400 evidence pack
The campaign written up for your warranty claim, insurer or asset manager.
PDF · signed
Everything is kept per blade and per pass, which is what makes the second campaign more useful than the first. Growth rates only exist if the earlier passes were retained.
How often, and what for
Anchored to IEC 61400 and to the reality that access is the expensive part. Condition-driven beats calendar-driven here more than anywhere.
- BladesTypical cadenceAnnually, plus after any recorded strikeWhat the pass looks forErosion, cracks, delamination, lightning damage, coating lossReferenceIEC 61400
- Tower and nacelleTypical cadenceAnnually, externalWhat the pass looks forCoating breakdown, corrosion, oil traces, fixingsReferenceIEC 61400
- Foundation and transitionTypical cadenceAnnually, or on conditionWhat the pass looks forGrout condition, corrosion, scour where visibleReferenceIEC 61400
- Array and accessTypical cadenceAnnuallyWhat the pass looks forAccess tracks, cable route condition, substation and yardReferenceISO 55001
These intervals are the starting points the standards describe, not a measured result of ours. Insurers, warranty terms and your own history will move them, and the point of planning from condition is that the interval stops being the thing that decides.
All four modules, on a wind fleet
The agent's most valuable act on a wind fleet is often not dispatching. It is grouping what was found into an access window that already exists.
Detect
Flies blades, tower and foundation at the resolution each defect class actually needs.
Leaves behind
Blade, tower and foundation passes, categorised and positioned along the span.
Analyse
Streams turbine SCADA, including strike events, so a capture can be planned by the thing that happened.
Leaves behind
Turbine SCADA including strike and alarm events, and the trend between campaigns.
Operations
Holds blade history, previous repairs and the warranty position per turbine.
Leaves behind
Blade history, previous repairs and the warranty position per turbine.
Decide & act
Plans capture around weather and access, and groups work into windows instead of raising it one turbine at a time.
Leaves behind
The access plan, the grouping decision, and a decision record for every action.
- 01O-EYEDetectInspection & digital twin
- 02O-ARCAnalyseSensors, IoT & telemetry
- 03O-OPSOperationsThe system of record
- 04actsO-AGENTDecide & actThe agentic layer
Every action writes back to the record, so the model sharpens with every job.
Questions from this sector
Offshore as well as onshore?
Both are within what drone capture does; what changes offshore is logistics and approvals rather than the analysis. We would scope an offshore campaign explicitly rather than treat it as an onshore job flown over water.
Do turbines have to be stopped?
For close-range blade capture, yes, and the stop is short. Planning around it is part of the value: grouping turbines into one window costs less production than stopping them one at a time across a season.
Can you measure erosion severity, not just see it?
Category and extent are reported per finding. Because passes are versioned, the more useful number is growth since the last one. A category that has not moved in two years is a different decision from one that has.
What about internal blade inspection?
Not from a drone. Internal inspection stays a rope or crawler job, and those findings belong in the same record. We are not going to claim aerial capture sees inside a blade.