Inspecting Carbon Structures

Carbon fails differently from metal — linearly, suddenly, and often invisibly at high-load points — so inspection is your primary defence. Aluminium and steel yield and deform before they break, redistributing stress and giving you visible, audible warning. A carbon/epoxy laminate is essentially linear-elastic all the way to rupture: it absorbs abuse silently, accumulating sub-surface delamination and fibre damage that it then releases with little or no notice. The whole discipline of carbon inspection is built around one reality: the damage that matters most is frequently the damage you can barely see, and the number that matters most is a strength loss you cannot see at all. Get good at looking and tapping, understand what each method can and cannot resolve, work the load paths hardest, and escalate anything ambiguous before it becomes a failure at sea. This is a core item on the annual maintenance schedule and works hand in hand with your standing rigging inspection.

Why carbon hides its damage — the failure sequence

To inspect carbon well you have to understand the order in which it breaks, because you are hunting the early stages before they cascade. A racing laminate is a stack of unidirectional or woven plies — thousands of high-modulus carbon filaments (tensile strain-to-failure only about 1.5–2%) locked in a brittle epoxy matrix — laid at chosen angles and, in panels, bonded to a foam or honeycomb core to form a stiff sandwich. Load does not break all of this at once. It fails in a predictable progression:

1. Matrix cracking. The off-axis plies carry load transverse to their fibres, and the matrix's transverse strain-to-failure is a fraction of a percent — far below the fibres'. So the resin micro-cracks first, in the 90° and ±45° plies, long before anything structural gives way. This is the milky, crazed, light-scattering change you are trained to spot.
2. Delamination. Those matrix cracks turn the corner and grow along the ply interfaces as interlaminar shear cracks, separating adjacent plies. Interlaminar shear and through-thickness tension are the laminate's weakest directions — there are no fibres bridging them, only resin — which is exactly why this is where hidden damage lives.
3. Fibre breakage. Only at the end do the load-bearing fibres rupture, at which point the part fails abruptly. Because the stress–strain curve stays essentially straight to that point, there is no plastic plateau to warn you.

The insidious case is impact. A dock, a submerged object, a dropped winch handle or a hard grounding delivers energy that drives matrix cracking and delamination internally while the resilient outer skin springs back looking almost untouched. Engineers call this barely visible impact damage (BVID) and define it against a viewing standard: damage not readily detectable under normal lighting from roughly 1.5 m (about 5 ft), typically leaving a surface dent of only 0.25–0.5 mm. The energy band that produces it — broadly 8 to 64 J in laboratory studies, i.e. a moderate knock — is precisely the band where delamination dominates the damage mechanism. A dent the size of a fingernail can sit above a delaminated footprint several times its diameter. This is why casual visual inspection is not enough, and it feeds directly from how the load paths and structural engineering of the boat concentrate stress at predictable spots.

The invisible number: compression after impact

Here is the fact that should change how seriously you treat a "harmless" knock. Delamination removes interlaminar shear capacity, which splits the laminate into thinner sub-laminates. Under compression — the loading a leeward chainplate laminate, a mast partner or a keel floor sees constantly — each thin sub-laminate buckles locally at a far lower load than the intact section would. That local instability is the trigger for final failure. Published compression-after-impact (CAI) testing routinely shows residual compressive strength falling by roughly 20–30% from a single low-velocity impact that leaves only BVID on the surface. In other words, a fitting area can look 95% intact and be 70–75% as strong in compression. That is the gap tap-testing and NDT exist to close, and it is why compressive load paths — not just tensile ones — deserve the closest scrutiny.

A related slow killer is fatigue. Even at stress levels below the static elastic limit, cyclic loading (every wave, every tack, every rig cycle) accumulates matrix micro-cracking and interface debonding that show up first as a measurable loss of laminate stiffness before any macro-crack appears. Race-optimised carbon appendages and grids are designed to a finite number of load cycles for exactly this reason; matrix crazing spreading around a hard-working fitting can be the visible leading edge of fatigue, not just a cosmetic blemish.

Reading the surface: what good and bad look like

Start every inspection in strong, raking light, ideally with a torch held at a shallow angle so shadows reveal texture and any deviation from a fair curve. Work symmetrically: compare the port chainplate laminate against the starboard, one ring frame against its twin. Your eye is a comparator, not an absolute gauge — it catches the odd one out. A healthy structure is uniform, glassy and consistent. Warning signs, and the mechanism each one betrays:

• Cracks, especially any radiating from or running into a fastener, fitting edge or hard corner — stress concentrators where matrix cracking initiates.
• Whitened, milky or crazed resin — dense matrix micro-cracking scattering light. Around a fitting this is the first visible stage of the failure sequence, often signalling overload or fatigue.
• Print-through — the weave telegraphing through the finish, indicating resin-starved or over-worked laminate where fibres have shifted. Near a high-load fitting, investigate it rather than dismissing it as cosmetic.
• Star fractures or spider cracks radiating from a point — the classic signature of a localised impact, and a prompt to tap the surrounding area for the wider delaminated footprint.
• Lifted, cracked or "tenting" paint at a bonded joint or fitting — frequently sitting directly over a substrate crack, a disbond, or a fitting working loose beneath.
• Distortion, a flat spot, or a proud/sunken area on what should be a fair curve — the skin has debonded or the core beneath has crushed, and the panel has lost its sandwich stiffness locally.
• Rust weeping, staining or a gap at a metal-to-composite joint, most critically the keel-stub interface — evidence of movement, water ingress and load working the joint.

Photograph anything ambiguous, note the date and a measured size, and re-check it next inspection. A stable mark is far less alarming than one that grows — and only a dated, sized record lets you tell the difference.

Tap-testing: the physics, the technique, the honest limits

Tap-testing is the workhorse for finding disbonds and near-surface delamination with no equipment, and it has real physics behind it. Model the laminate above a defect as a mass on a spring: a well-bonded region is stiff, so the coin's contact with the surface is brief and the impact force peaks high and sharp; a disbond or delamination softens that local spring, so the contact time lengthens and the peak force drops. The audible result is that solid laminate rings with a sharp, clear, higher note while a defect returns a dull, dead, lower one — the shift you hear is the force–time pulse moving to lower frequency as its duration grows. Instrumented tap testers exploit exactly this, measuring contact time directly to remove some of the operator's guesswork, and they sit in an interesting acoustic niche: much shorter wavelength than vibration analysis, much longer than ultrasound.

Technique: tap lightly and repeatedly in a close, regular grid, keeping tap force and rhythm consistent so your ear judges the contrast between adjacent taps, which it does far better than absolute pitch. Overlap your grid around any dead note to size its edges.

Know the limits, because they are hard boundaries, not caveats:

• It is operator-dependent and subjective, and it characterises stiffness only, not the full internal picture.
• It is reliable for shallow defects and thin skins. On a skin-to-core disbond it typically resolves features down to roughly 10 mm in diameter — and no better.
• As skin thickness increases and curvature increases, sensitivity falls; deep defects under thick laminate and disbonds beneath heavily curved sections can be missed entirely.

So a dead note means something is wrong here — not how big, how deep, or how serious. Treat it as a flag to escalate, never as a diagnosis.

The high-load areas that decide the boat

Concentrate your effort where the structure earns its keep. On a Grand Prix one-design these are the chainplates and their surrounding laminate, the mast step and partners, the keel floors and grid, ring frames and bulkhead-to-hull bonds, the rudder bearings and their twin-rudder housings, and the retractable bowsprit heel fitting.

In a modern carbon boat, chainplates are commonly laminated in — carbon tows fanned out to disperse the concentrated shroud load across the hull skin, a longitudinal stringer and the adjoining bulkhead over many plies at mixed orientations, so no single point becomes a hard spot. Inspect that fan for cracking, matrix whitening, print-through, and any staining or relative movement of the surrounding laminate when the rig is loaded — a fitting that visibly "clicks" or works under load is urgent regardless of how clean the laminate looks around it. At bulkheads, floors and the keel grid, look for tabbing lifting, cracks at the join line, panel edges peeling from the hull, and — at the keel-stub joint specifically — weeping, rust staining or any gap that opens under load, all classic precursors to a keel-attachment problem on a weight-optimised racer.

Two situations override the whole checklist. After a hard grounding, collision or rig event, inspect the primary structure carefully even if you can see nothing — that is precisely when BVID and CAI strength loss hide. And treat any water in a cored panel as structural, not cosmetic: once moisture enters the core it degrades the skin-to-core bond, and every freeze–thaw or thermal cycle expands the disbond a little further, so a soft spot found wet this season is often larger next season.

The Melges 40 hull structure concentrates these paths as any performance boat does, but the exact laminate schedules, insert locations, keel-bolt torques and design load cycles are boat-specific — verify them against the class rules and the builder's own structural documentation rather than assuming, and never work to a remembered number.

When to escalate — and what a specialist actually adds

Escalate to a composite engineer whenever you find a change in a primary load path, any crack that reaches a fastener, any damage larger than a coin, any unexplained tap-test dead zone, any wet or soft cored panel, or any suspicion following an impact or grounding. The cost of a scan is trivial against a rig or keel failure offshore.

A specialist brings instrumented non-destructive testing that resolves what your eye and ear cannot, and the two mainstream methods are genuinely complementary:

• Phased-array ultrasonic testing (a C-scan). A multi-element probe — commonly around 5 MHz with 64 elements — sweeps and focuses the beam electronically to map delamination through the thickness and size it accurately in plane. Depth (axial) resolution improves with frequency, but higher frequencies attenuate faster in the resin-rich, fibre-scattering laminate, so the surveyor trades penetration against resolution and needs proper coupling and setup. This is the method that turns "a dull spot here" into a sized, located defect map.
• Pulsed (flash) thermography. A brief high-energy flash heats the surface and an infrared camera watches how the heat diffuses; a shallow disbond, delamination or water-soaked core cell blocks or alters the flow and shows as a thermal anomaly. It is fast and non-contact and excellent for shallow defects and moisture ingress, but it obeys a hard physical rule of thumb: reliable depth estimation needs the defect's diameter to be more than roughly six times its depth, so small or deep defects fall outside its confident range.

An experienced surveyor chooses — or fuses — these to build a real internal map before anyone touches a grinder. That map is what converts a suspicion into an engineered decision to monitor, repair or replace.

Getting repairs right

Structural carbon repair is specialist work, full stop. Restoring the original strength means matching the fibre type and modulus, the ply orientation and full stacking sequence, the resin system and the fibre volume fraction, cutting the correct scarf (a shallow taper — often in the order of 1:20 or gentler — so load feeds gradually across the joint rather than through a step), then consolidating under vacuum and curing correctly, frequently with a post-cure to reach the design glass-transition temperature. Miss any of those — too steep a scarf, a resin-starved ply, an under-cured matrix that softens when the deck heats in the sun — and the repair can look flawless while carrying a fraction of the design load. On a chainplate, keel floor or rudder bearing, a ground-out patch faired smooth and painted can be weaker than the damage it replaced.

That is why the goal of routine inspection is not to fix, but to find, flag and get properly assessed. Look and tap in good light, compare port with starboard so your eye works as a comparator, work the compressive and tensile load paths hardest, treat any water intrusion as structural, and escalate anything that changes. Done consistently, this is how a carbon boat stays fast and trustworthy season after season — see the wider carbon construction picture for why the material rewards this discipline.

Inspection intervals, keel-bolt torque values, design load cycles and repair procedures must always follow the builder's documentation and a qualified composite engineer. Any figures specific to a given boat are to be verified against its class rules and structural drawings.