Composite Repair Basics

A structural composite repair has to rebuild the laminate itself — the fibres, their orientation, the fibre-volume fraction and the cure state — not just the surface you can see. In metal, a weld or a doubler shares load through an isotropic material with a single yield stress. In a carbon or glass laminate the strength lives in thousands of directional fibres locked in a cured, cross-linked matrix, so a genuine repair must recreate that architecture, its resin content and its degree of cure, and bond it back into the structure through a joint whose strength is set by adhesive shear, not by the fibre. That is why a patch that looks perfect can still be dangerously weak. This article builds on carbon construction and the carbon inspection guide.

Why a composite fails differently from metal

A laminate's stiffness and strength come from fibres oriented deliberately — 0 degrees for fore-and-aft (axial) load, plus or minus 45 degrees for shear and torsion, 90 degrees for hoop and transverse load. The matrix holds the fibres in place, transfers load between them by shear, stabilises them against micro-buckling in compression, and sets the transverse and interlaminar properties — but it is an order of magnitude weaker than the fibre and it, not the fibre, governs most failure modes a repair addresses. Damage rarely severs fibres cleanly. Far more often it separates plies (delamination) or the skin from its core (disbond), leaving the fibres intact but no longer working as a stack. Restoring strength therefore means three coupled things: replacing the right fibres in the right directions to re-establish continuous load paths, restoring the fibre-volume fraction so the panel regains its designed stiffness, and curing the matrix to the cross-link density the original resin reached. Get the orientation wrong, leave the laminate resin-rich, or under-cure the matrix, and the repair becomes the weakest point.

This matters acutely on a modern race boat because load intensity combines with a matrix-dominated failure mode: delamination. A Grand Prix one-design carries a great deal of sail on a light, stiff platform and channels rig and keel loads through concentrated structure, so panels are thin, highly worked, and unforgiving of a soft or porous patch. The publicly published facts for the Melges 40 describe an epoxy-infused carbon hull, deck and internals over a foam core, designed by Botin Partners, built by Premier Composite Technologies to ISO structural regulations (commonly cited as Category B), with an all-carbon rig and an electrically actuated canting keel that swings to around 45 degrees. A repair here is not cosmetic tidying; it is re-engineering a stressed member. (Any specific ply schedules, laminate thicknesses, fibre types, resin system or bond specifications for a boat must be verified against the class rules and that boat's own build documentation — never assumed. The figures above are public class-level facts, not a repair specification.)

Assess before you touch anything

Good repairs start with honest assessment, because the visible mark rarely bounds the damage. The fibre can be intact while the interlaminar interfaces around it are wrecked, and it is those interfaces that fail. Work through it in order:

• Map the extent. Clean the area and look for cracks, crazing, star fractures, resin-rich whitening (matrix micro-cracking scatters light) and any softness or oil-canning underfoot on a panel. A gouge that scuffs the top ply is a different problem from an impact that has driven a delamination front through the outer skin.
• Tap test the surround. A delamination or disbond is a spring-mass discontinuity: the debonded skin has lower local stiffness, so it rings duller and lower than sound laminate. Tap methodically with a coin or coin-tap tester and outline the dull zone with tape; it almost always extends well past the visible mark. Tap testing resolves near-surface, larger delaminations but is insensitive to deep, tight or small ones.
• Beware barely visible impact damage (BVID). A low-velocity impact drives a conical delamination stack through the thickness — small at the impact face, largest at the back face — while leaving a surface dent of only about 0.5 to 1.0 mm. That barely visible mark can coincide with compression-after-impact residual strength reduced by up to 50 per cent, because delaminated sub-laminates buckle at a fraction of the intact stack's load. If the geometry suggests a hard hit, treat it as significant even when it looks trivial.
• Locate it against the load path. The same-sized ding is trivial in a lightly loaded fairing and serious next to a chainplate, the keel structure, a ring frame or the mast step, where a delamination sits in a peak-stress field and propagates under fatigue. Position drives everything that follows.

Where you cannot see the internal condition, ultrasonic (pulse-echo A-scan or C-scan) or active thermography by a specialist maps hidden delamination that tapping misses — the ply-depth of the deepest debond dictates how far back the scarf must be taken, so this is worth doing before committing to any repair near primary structure.

Cosmetic versus structural — the line that matters

The single most expensive mistake in composites is treating a structural repair as cosmetic. A cosmetic repair restores appearance and seals the surface: filling a scratch, fairing a scuff, renewing gelcoat or clear coat. It carries no load and requires no engineering. A structural repair must restore strength, stiffness and the laminate's cure state in a part that resists real forces — matching fibre type, areal weight, ply count and orientation, bonding into a properly prepared scarf, and curing to the correct Tg.

The two are identical once faired and painted, and that is exactly the danger: filler over broken structural plies looks finished and is still failed, but now the failure is invisible. The rule is simple: if the damage reaches the load-bearing plies, sits within roughly one fastener-diameter of a high-load fitting, or rings dull on a tap test, treat it as structural until a competent person proves otherwise.

The engineering of a scarf repair

A sound structural repair removes the damage and rebuilds continuous load paths rather than bridging over them; the geometry that makes this work is the scarf, and its mechanics dictate every number that follows.

• Scarf, don't overlay — and understand why. Damaged material is ground back into a shallow taper so new plies bond over a long, gradually stepped surface. Across an ideal scarf the load transfers as near-constant shear in the bondline, whereas an external overlap piles load into sharp shear and peel spikes at both ends. The average adhesive shear stress scales with the taper: for a scarf of half-angle θ carrying far-field stress σ, τ is of order σ·tan(θ), so halving the angle roughly halves the bondline shear. Because common structural film adhesives have a shear allowable of order 30 MPa, recovering a highly strained carbon laminate (around 4000 microstrain) needs a taper angle near 3 degrees — which is why structural scarf ratios run about 20:1 to 60:1 (and up to 100:1 on the most highly loaded aerospace parts), not the 12:1 sometimes quoted for lightly loaded work. Well-executed scarf repairs typically recover on the order of 80 per cent of pristine strength while restoring essentially full stiffness. A patch slapped over a hole creates a stiff step, a peel initiation point and a fatigue crack starter — a classic failure mode.
• Stepped-lap is the same idea, discretised. Cutting the taper as a staircase of ply steps is easier to execute by hand and can rival a smooth scarf, but each step corner concentrates stress, so step length and count must keep those peaks below allowables.
• Match the laminate exactly. New plies replicate the original fibre type, areal weight and orientation, ply for ply, so the repair recovers the same directional stiffness and the loads redistribute as designed. Substituting a heavier cloth, a different modulus fibre or skipping a plus/minus 45 pair changes the panel's stiffness locally, and stiffness mismatch pulls load into or out of the repair unpredictably.
• Rebuild the core. In sandwich structure the foam or honeycomb core carries transverse shear and holds the skins apart at the design core thickness; that spacing is what gives the panel its bending stiffness (which scales with the square of the separation). Crushed, sheared or wet core is cut out and replaced with matched-density core, potted at the edges, and both skins re-bonded — otherwise the panel behaves as two thin, unsupported laminates with a fraction of the original stiffness.
• Restore fibre-volume fraction, then cure. Consolidation and cure, covered next, together decide whether the rebuilt stack reaches its properties.

What good looks like: a fair repair with no proud step, a sharp ring on the tap test across the whole area including the join, full-thickness plies bonded over a long scarf, no resin pooling. What bad looks like: a proud external patch, dry or resin-starved cloth (fibre not fully wetted, so no shear transfer), a resin-rich glossy patch (excess matrix, low fibre fraction, low stiffness), trapped air, a dull tap-test ring at the join, or an untapered edge.

Consolidation, voids and the cure question

How the repair is laid up, consolidated and cured determines whether it truly restores strength, and this is where amateur and engineered repairs diverge most.

Consolidation and voids. Vacuum bagging draws the stack down under atmospheric pressure, compacting the plies, extracting entrapped air and volatiles, and bleeding excess resin to lift the fibre-volume fraction. The physics sets a hard ceiling: even a perfect vacuum yields only about one atmosphere, 14.7 psi, and a realistic shop differential is nearer 6 to 12.5 psi (roughly 12 to 25 inches of mercury). That is why a production part cured in an autoclave at several atmospheres of positive pressure achieves a lower void content than any vacuum-only repair, and why every psi of consolidation matters. Voids are not cosmetic: interlaminar shear strength falls by roughly 7 per cent for each 1 per cent of void content up to about 4 per cent (some carbon/epoxy data show around a 35 per cent ILSS loss by 4 per cent voids), because voids are pre-existing interlaminar cracks that concentrate stress and link up. A perforated release film over the laminate meters resin bleed into the breather, so you can pull higher vacuum for better compaction without starving the laminate of matrix.

Cure and glass-transition temperature. The cure is the part people most underestimate. As an epoxy cross-links, its glass-transition temperature (Tg) climbs with the degree of cure — captured by the DiBenedetto equation, Tg rising from the uncured toward the fully cured value as conversion approaches 100 per cent. The trap is vitrification: if the cure temperature sits below the resin's momentarily developing Tg, the matrix glasses off and the reaction effectively stalls before full conversion, locking in a low Tg and incomplete properties. A room-temperature wet-layup patch on a part whose resin was designed to be cured hot therefore arrests early — it may feel hard yet fall short of design strength, and, because its Tg is low, it can soften on a black deck in the Australian sun where surface temperatures climb well above ambient. Prepreg (cloth pre-impregnated with a metered resin fraction, B-staged to roughly 15 to 30 per cent cross-linked and stored frozen) is consolidated under vacuum and taken through a defined heat cycle, driving conversion up and delivering a reproducible Tg that matches a factory carbon component. Where hot cure is impossible in situ, a controlled post-cure — held meaningfully above the intended service temperature, commonly of order 20 to 40 degrees C above it — advances conversion and lifts Tg after the fact. The carbon construction of the original part sets the standard the repair has to meet.

Surface preparation, moisture and prepreg handling

Two silent killers sit either side of the cure, and both are procedural rather than exotic.

Bond-line preparation. The scarf is a bonded joint, and bonded joints fail at contaminated interfaces long before they fail in the adhesive. The prepared surface must be clean to the fibre, freshly abraded, and free of release agents, mould wax, silicone (from polishes) and skin oils. Grit-blasting or careful abrasion, followed by a solvent wipe with clean cloths changed often, raises surface energy so the adhesive wets and forms chemical bonds; a peel ply torn off immediately before bonding leaves a fresh, high-energy surface and can raise surface energy by more than half — but it must be a bond-compatible peel ply (no silicone finish) or it does the opposite. Handle prepared scarf faces with clean gloves only; a single fingerprint re-contaminates the bond.

Moisture — the failure mode that strikes at cure. A marine laminate, and especially a wet core, absorbs water. Take that laminate straight to an elevated-temperature cure and the trapped moisture flashes to steam: internal vapour pressure blows skins off cores (blown face sheets), drives fresh disbonds away from the repair, and nucleates voids in the fresh adhesive bond line. The mandatory step is a slow, controlled dry-out under gentle heat and vacuum before any hot cure — low temperature over long time, ramped gently, so water leaves as vapour without over-pressurising the interfaces. Rushing the ramp cooks the moisture into damage, and a repair that skips drying can be weaker after cure than before it began.

Prepreg out-life and cold chain. If prepreg is used, it is a perishable, time-and-temperature-limited material. It is stored at around minus 18 degrees C, where shelf life runs roughly 6 to 24 months, and it has an out-life — the cumulative, irreversible clock, typically of order 250 to 750 hours at 21 degrees C or below, that it may spend out of the freezer before it must be cured. Resin advancement (further B-staging at room temperature) quietly consumes tack, drape and ultimately the achievable cure long before anything looks wrong. Rolls must be thawed sealed to dew point before opening, or condensation contaminates the very bond you are trying to make. Expired or heat-abused prepreg looks identical to good material and produces a substandard repair.

Temporary versus permanent — and when to call a professional

There is a legitimate place for a temporary repair: getting a boat off the water safely, or through a regatta, without pretending strength has been restored. A resin-and-cloth splash over a crack, or tape and a backing plate, can seal water out and hold a lightly loaded area briefly — provided everyone treats it as a stopgap, keeps water out of the laminate, and de-rates the loads. A permanent structural repair restores the engineered strength and cure state and carries full design load indefinitely. Never let a temporary fix disguise unrestored structure — a faired-over stopgap reads as finished to the next person who loads the boat.

Call a composite specialist when damage is in or near a high-load area, when the tap test shows delamination beyond the visible mark, when core is involved, when the laminate may be wet, or whenever you cannot see the full ply schedule or achieve the correct cure. The honest test is whether you can rebuild the exact laminate — right fibre, right orientation, right fibre fraction, right scarf, dried, consolidated and cured to the builder's Tg, in a load path you fully understand. If not, the role of your own inspection is to find and flag the damage, not to guess at the fix. Continuing to load a compromised structure, or fairing over a failure, is how a small ding becomes a rig-down or a hull breach at the worst possible moment.

Structural composite repairs must be carried out by a qualified composite engineer to the builder's engineered standard.