Adhesive bonding for sheet metal: when glue beats a weld

The stiffener on a powder-coated 1.5 mm CRCA enclosure door is normally welded, and the weld shows through the paint as a shallow dip on the outer face. Filling and sanding that dip is a manual operation on every part, and it still telegraphs under a gloss finish. Bonding the stiffener removes the heat, the dip and the filler, but only if the joint carries the load and survives the coating line.

Adhesive bonding for sheet metal is a load-path decision before it is an adhesive decision. A bonded lap carries shear well and peel badly, and thin sheet is the worst case for peel because the panels rotate as they load. What follows gives the lap-shear and peel numbers for 1.5 mm CRCA, the overlap length beyond which extra adhesive does nothing, the surface preparation that decides whether the joint lasts five years or five months, and the notes that make the joint inspectable on a drawing.

The worked case is a 1.5 mm CRCA door stiffener carrying 1.2 kN in shear over a ten-year life, powder coated after assembly at 190 °C.

What load can adhesive bonding for sheet metal carry?

A 25 mm wide single lap with a 12 mm overlap in 1.5 mm CRCA, bonded with a toughened acrylic at 12 MPa lap shear, carries 3.6 kN in a short-term tensile test. That is roughly one 5 mm spot weld in the same sheet, which lands between 3.5 and 5 kN.

The arithmetic is area times shear strength: 25 mm × 12 mm gives 300 mm², and 300 mm² × 12 N/mm² gives 3,600 N. The figure looks reassuring until the design factor is applied. A structural joint under sustained load is designed at 25 to 35% of its short-term lap shear, because the adhesive creeps under constant stress and because the test specimen is not the production joint. At 30%, the same lap is good for 1.08 kN, which is short of the 1.2 kN requirement. Widen the overlap to 20 mm — 500 mm², 6.0 kN short-term — and the allowable becomes 1.8 kN with margin.

The second number is stiffness. Structural adhesives run 300–1,000 MPa in shear modulus against 79,000 MPa for steel, so the bond line is a soft spring in the load path. Shear deflection is small: at 1.2 kN over 300 mm², with G = 500 MPa and a 0.3 mm bondline, it is 0.0024 mm. The lap also rotates. Load enters through the sheet mid-plane, offset by half the sheet thickness, and that 0.75 mm eccentricity bends the joint and puts peel on both ends. At 1.5 mm sheet the rotation governs, not the shear figure.

The numbers a bonding specification needs

The Ganesh column is where the route is decided. We bond stiffeners, brackets and liners onto formed and coated parts, but the joint geometry has to be flanged or otherwise supported; a flat lap on 1.5 mm sheet with a load pulling the panels apart is a weld, whatever the drawing says.

Why extra overlap stops helping

Shear stress in a lap joint is not evenly distributed. It peaks at the two ends of the overlap and falls to a fraction of the average in the middle, because the two adherends are stiff in tension and the adhesive is compliant in shear. The characteristic length over which the stress decays is:

1/λ = √(E · t · t_a / G_a)

For steel at E = 210,000 MPa, t = 1.5 mm, a 0.3 mm bondline and G_a = 500 MPa, that gives √(210,000 × 1.5 × 0.3 / 500) = √189 ≈ 13.7 mm. Most of the load transfers within 1.5λ, about 20 mm, of each end.

That is the reason a 12 mm overlap is weak and a 25 mm overlap is not twice as strong. Doubling from 12 to 25 mm raises capacity by roughly 40%, because both ends finally have room to develop. Doubling again from 25 to 50 mm raises it by under 10% while doubling adhesive consumption, and it lengthens the bond line that has to be pressed and held flat.

Where the extra length does earn its place is fatigue. A longer overlap lowers the peak stress at the ends, and a joint cycling between 0 and 1.2 kN ten times a day lives longer at 25 mm than at 12 mm even though the static numbers are close.

Flange the joint before lengthening it

Because the failure mode is peel, geometry beats length. Three changes reduce peel more than any adhesive upgrade:

• Flange one adherend. A 10 mm return flange moves the load path closer to the bond line and cuts the rotation.
• Joggle the lap. A step pressed into one panel brings both mid-planes level, so the joint is loaded in pure shear.
• Use a double lap or a channel. Bonding a hat section or a channel over the joint doubles the bonded area and removes the free edge that peels first.

A joggle costs one press or brake operation. A 50 mm overlap costs adhesive on every part and does less.

What the adhesive grade actually changes

Toughened acrylics are the default for production sheet metal because they tolerate minimal surface preparation, cure at room temperature and hold peel better than epoxies. Two-part polyurethanes are the choice for mixed materials and for joints that flex, at the cost of half the shear strength and a longer fixture. Epoxies give the highest shear and the best high-temperature performance, but they are brittle and need a clean, blasted surface and often a warm cure.

Surface preparation decides more than the adhesive grade

A bonded joint fails at the interface long before it fails in the adhesive. Four substrates behave differently.

CRCA. Mill oil sits at 0.5–2 g/m² on incoming sheet, and it is formulated to be hard to remove. Alkaline degreasing at 50–60 °C followed by a water rinse removes most of it; blasting to Sa 2½ removes the rest and leaves a 3–6 µm profile that gives the adhesive something to key into. Abrading by hand with 120 grit works for small parts but leaves a profile that varies with the operator.

Galvanised sheet. Zinc is a poor bonding substrate: it forms a weak, friable oxide and it reacts with some acidic adhesives. A zinc phosphate conversion coating, the same chemistry used before powder, lifts bond strength and consistency. Where the part is also coated, bond to the phosphate, not to the powder.

Aluminium. The natural oxide is 2–4 nm thick and grows to around 10 nm within a day in Mumbai humidity. Bond within four hours of abrading, or use a conversion coating or a primer that the adhesive supplier lists. A joint bonded to a day-old oxide fails at 30–50% of the strength of one bonded to a fresh surface.

Stainless. Abrade and degrease, and do not passivate the bond area. Passivation is there to build the chromium oxide that makes stainless corrosion resistant, and that same oxide is what the adhesive has to grip.

The check that catches most of this is the water-break test: pour demineralised water over the prepared face and watch the film. A continuous film that holds for 30 seconds means the surface is clean and wettable. A film that pulls into beads means oil or contamination is still there, and no adhesive will compensate.

Contamination after preparation

Gloves, not bare hands. Skin oil transfers at 0.1–1 mg/cm² per touch, and a fingerprint in the middle of an overlap is a void that starts a corrosion cell. Parts should be bonded within the shift they are prepared in; a blasted CRCA face flash rusts in 24–48 hours in the monsoon, and rust under an adhesive is a disbond waiting for a load.

The coating line is the constraint, not the adhesive

Most structural adhesives cannot take a powder cure. A standard cure holds the part at 180–200 °C for 10–15 minutes, and a 1.5 mm panel reaches metal temperature after 3–8 minutes in the oven, so the joint sees the full soak. A toughened acrylic with a 40–60 °C glass transition softens, creeps and takes a permanent set.

There are three workable routes:

1. Coat, then bond. Mask the bond area with a plug or a high-temperature tape, coat, cure, remove the mask and bond. The bond face stays bare and the joint is assembled last. This is the route for visible parts where the weld mark is the problem in the first place.
2. Bond, then coat at low temperature. Only works with a 2K epoxy rated above the cure and a coating that cures at 60–80 °C. The thermal expansion mismatch — 50–80 µm/m·K for the adhesive against 12 µm/m·K for steel — leaves the joint in shear before it sees a service load.
3. Weld-bond. Four to six spot welds at 100–150 mm pitch locate the parts and carry peel, while the adhesive carries shear and seals the interface. The welds still mark the face, but far less than a full seam, and the joint is stiffer than either process alone.

Route 1 has a hidden cost: masking. Every plug is a manual operation, and a missed plug means adhesive bonded to powder, which fails at the coating interface at around 2–4 MPa. If the bond area is large, it is often cheaper to coat the part in two passes with the bond face masked by a fixture.

What Ganesh actually runs

Welding is the core joining process here: three CLOOS robotic MIG cells, more than 30 manual MIG, spot and SPM stations, and fixtures designed and built in-house for the assemblies that need them. Bonding enters the route at the assembly stage, after the parts have been cut on the 12 kW fibre laser, formed on the 200 T brake, blasted in the 16 ft × 16 ft chamber and pretreated on the seven-tank line. That sequence matters, because blast and pretreatment are the same operations that prepare a bond face.

We do not hold a rail or aerospace structural-bonding qualification, and a joint governed by a rail bonding standard or an aerospace specification has to be validated by the party that does. What we can do is quote a bonded or weld-bonded route, build the fixture that holds it during cure, and inspect the result.

The welded assemblies capability page covers the welding side of the same decision, and the bending and welding hub indexes the distortion and fixture articles.

What goes on the drawing

1. Joint type and geometry: flat lap, flanged lap, joggle or double lap, with the overlap length dimensioned, not assumed.
2. Adhesive family and grade, named by the customer's approved list where one exists, with the cure schedule and fixture time.
3. Bond face preparation: degrease plus blast to Sa 2½, or abrade, with the profile range.
4. Bondline thickness, typically 0.25 mm nominal with a 0.1–0.5 mm acceptance band, checked on a cured coupon or by ultrasonic gauge.
5. Masking requirement, if the part is coated after bonding, with the bond area shown on the drawing rather than described in a note.
6. Proof load and acceptance: a sample pulled to failure each shift, with the required failure mode — cohesive in the adhesive, not adhesive at the interface.
7. Service temperature and environment, so the adhesive is chosen for the duty rather than by habit.

If the joint is still open, the welded assembly tolerances article shows what a welded frame holds, and fixture-controlled welding covers the fixture logic that a bonded assembly reuses. Send the drawing with the load and the finish to request a quote and the route can be priced as bonded, welded or weld-bonded.

When adhesive is the wrong route

Above 120 °C in service. A genset canopy near an exhaust outlet, or a part that goes through a second cure, needs a weld. Adhesives that hold at 180 °C exist, but they are epoxies with brittle peel behaviour and a warm cure, and the economics rarely favour them on sheet metal.

Loads that pull the joint apart. A bracket hanging a load away from its fixing has a peel component by design, and no adhesive grade fixes that. Change the geometry or weld it.

Joints that must conduct or earth. Adhesive is an insulator. A bonded enclosure needs a separate earth strap with its own fasteners and a masked contact area, and that often removes the cost advantage.

Parts that will be disassembled. Bonded joints are permanent. Anything that is serviced in the field should be riveted, screwed or welded.

Thin sheet with a lot of flex. Below 0.8 mm, panel flutter works the bond line in peel continuously. A weld or a mechanical fastener tolerates that movement; an adhesive eventually does not.

The article that decides between a welded frame and a bonded one is usually the tolerance, not the process. Spot welding for panels and enclosures sets out what a spot pattern holds, and that is the number to compare against the 1.8 kN allowable above.