Why bond, and when not to

An adhesive joint is a thin layer of polymer holding two parts together over an area. That simple description hides the two reasons bonding is chosen and the two reasons it fails. Bonding spreads load over a whole surface instead of a few points, and it joins materials that cannot be welded together, without heat, holes or marks. But the layer is weak in some directions and strong in others, and it only holds if the liquid adhesive wetted the surfaces properly before it set.

This chapter sets out what bonding offers and where it is weak, introduces Plant K’s two bonding jobs, and shows the first and most common mistake in judging a bonded joint: multiplying the adhesive’s data-sheet strength by the bonded area.

What bonding offers

Mechanical fasteners and welds carry load at discrete points or lines. A rivet or stud passes all of its share of the load through a small area around a hole or a weld nugget, and the surrounding sheet sees a stress concentration there. A bonded joint passes load through its whole area, so the sheet sees a far smoother stress field. In thin sheet this matters more than anything else: a 2 mm skin with a stud welded to it has a small heat-affected spot and a hard point; the same skin with a bonded flange has neither.

The other advantages follow from the same idea.

• No marks on the outer face. Bonding needs no heat and no holes, so a visible surface stays flat and unmarked.
• Unlike materials. Aluminium to polycarbonate, steel to glass, composite to metal: bonding joins combinations that cannot be welded and that corrode or crack around fasteners.
• Sealing. A continuous bondline seals as it joins, keeping out water and dust.
• Stiffness and damping. A bonded stiffener acts with the skin along its whole length, and many adhesives damp vibration, which reduces drumming in thin panels.
• Fatigue. Without holes, the sheet has no drilled stress raisers from which fatigue cracks start.

Where bonding is weak

The weaknesses are just as specific, and every chapter of this book returns to one of them.

• Peel and cleavage. A bondline loaded in shear spreads its load over an area. A bondline pulled apart at one edge, like tape being peeled, concentrates the load on a line a fraction of a millimetre wide. Most adhesives are many times stronger in shear than in peel.
• Surface sensitivity. The adhesive only sticks to what it touches. A film of oil, release agent, loose oxide or condensation a few molecules thick can halve a joint’s strength without any visible sign.
• Temperature limits. Polymers soften as they approach their glass transition temperature and creep under sustained load. A joint that is strong at 20 °C may be weak at 80 °C.
• Durability. Water slowly attacks many adhesive-to-metal interfaces, especially on poorly prepared aluminium, and the loss of strength may take months to appear.
• Cure time and process control. An adhesive needs time, and often heat, to reach strength. Mixing ratio, open time and fixturing all have to be right, every time.
• Inspection. A bonded joint that looks perfect may not be. There is no simple non-destructive test that proves adhesion, so the process has to be proved instead.

Figure 1Bonding against mechanical fastening and welding in thin sheet

Property	Bonded joint	Welded studs or rivets	Spot or seam welding
Load path	spread over the whole bonded area	concentrated at each fixing	concentrated at each nugget or along the seam
Marks on the outer face	none	stud or rivet heads, read-through	heat marks and distortion
Unlike materials	most combinations	many, with corrosion care	only compatible metals
Peel and cleavage	weak unless designed out	strong	moderate
Service temperature	limited by the adhesive	limited by the sheet	limited by the sheet
Inspection	process must be proved	visible, torque or pull test	visible, destructive tests
Time to handling strength	minutes to hours	immediate	immediate

How load passes through a bonded joint

A bonded joint is three layers acting together: two adherends and the adhesive between them. Load enters one adherend, crosses the adhesive and leaves through the other. The adhesive is much less stiff than metal: a structural epoxy has a shear modulus around one hundredth of aluminium’s, and an elastic polyurethane perhaps one thousandth of that again. So the load does not cross the bondline evenly. Near the ends of an overlap, where one adherend stops and the other carries all the load, the adhesive takes far more than the average. Chapter 6 calculates how much more.

The direction of load matters as much as its size. There are four basic loadings on a bondline.

1. Shear, in which the two adherends slide relative to each other parallel to the bondline. This is the loading adhesives carry best.
2. Tension, in which the adherends are pulled apart evenly over the whole area. Adhesives carry this well too, but it is rare in practice because a perfectly even pull is hard to arrange.
3. Cleavage, in which two rigid adherends are pried apart from one edge, so the stress is highest at that edge and falls across the joint.
4. Peel, in which a flexible adherend is pulled away from the other, so the stress concentrates on a narrow line at the point of separation.

Good bonded design arranges the parts so that the adhesive sees shear, and tension where unavoidable, and never relies on it in peel or cleavage. That principle decides the shape of every joint in this book.

Plant K’s two bonding jobs

Plant K’s doors have two bonding jobs with different demands, and they illustrate the two main families of adhesive joints.

The stiffeners are a structural bond. Three hat-section stiffeners of 1.6 mm aluminium run vertically inside each 2.0 mm aluminium door skin. Each stiffener has two flanges 20 mm wide and 1,600 mm long, which are bonded to the skin with a two-part toughened epoxy at a bondline thickness of about 0.2 mm. The joint must carry the shear that makes skin and stiffener bend together, resist the drumming of the door in wind and survive being slammed and dropped. Aluminium to aluminium, the two parts expand together with temperature.

The window is an elastic bond. A 6 mm polycarbonate sheet is bonded over a cut-out on the inside of the door. Polycarbonate expands about three times as much as aluminium with temperature, and a dark kiosk door in the Australian sun gets very hot. The joint has to seal and hold the window while letting it move, which is a job for a thick, flexible adhesive rather than a thin, rigid one, as Chapter 8 shows.

The case for bonding the stiffeners

The plant’s current method welds 12 studs along each stiffener, 36 per door, and fixes the stiffeners with nuts. Each stud leaves a slight mark on the painted face, which the paint shop fills and sands. At about 1.5 minutes a mark, the 36 marks take 54 minutes a door; at a labour rate of A$78 an hour, excluding GST, that is about A$70 a door, or A$2,808 a week at 40 doors. Stud welding and fitting nuts take a further 32 minutes a door.

Bonding removes the marks entirely. Preparing the flanges and the skin takes about 12 minutes, and dispensing, placing and fixturing the stiffeners about 10 minutes, while the cure happens in a fixture that the plant already has space for. The adhesive itself costs little: the bondline volume is 2 flanges × 20 mm × 0.2 mm × 1,600 mm × 3 stiffeners = 38,400 mm³, or 38.4 mL, and with fillets and mixing-nozzle waste the plant expects to use about 50 mL a door. At an illustrative A$0.18 a millilitre that is about A$9.

Figure 2Plant K's stiffener fixing, studs against bonding, per door

36welded studs and nuts on the current door

54 minfilling and sanding stud marks on the painted face

22 minpreparing, dispensing and placing three bonded stiffeners

about 50 mLadhesive including fillets and nozzle waste

The figures make bonding look like an easy decision, and on cost it is. The trial batch showed that it is not yet a reliable one.

The error: strength times area

When the plant first proposed bonding, the design review accepted it on a single calculation. The epoxy’s data sheet gave a lap shear strength of 25 MPa on aluminium. Each stiffener has 2 × 20 × 1,600 = 64,000 mm² of bonded flange. The reviewer multiplied the two: 25 × 64,000 = 1,600,000 N, or 1,600 kN per stiffener, more than a hundred times any load the door could see. The twelve studs it replaced had a total weld base area of only about 236 mm², and the bond area was 272 times as large. The conclusion was that strength could not be a concern.

The trial batch disagreed. In the plant’s drop test, a door dropped onto its edge from a set height, two stiffeners on one door separated from the skin over the last 60 to 80 mm of their length, and the separation started at the very end of the flange. A third door showed a stiffener that had pulled away cleanly from the skin over a short length, leaving the adhesive on the stiffener and bare metal on the skin.

The calculation was wrong in three ways, each the subject of a later chapter.

• The data-sheet strength is an average over a short overlap. The 25 MPa comes from a standard lap shear specimen with a short overlap. It is the failure load divided by the overlap area, and in that specimen the stress peaks at the ends. Over a 1,600 mm length loaded from its ends, almost all of the bond is lightly stressed and the ends carry the load. Multiplying an average strength by a long length overstates the capacity enormously, as Chapter 6 shows with the shear-lag model.
• The loading was not shear. When the door was dropped, the skin flexed and the end of each stiffener flange was pulled away from it. That is peel, and the adhesive’s peel strength is a few newtons per millimetre of width, not tens of megapascals. Chapter 7 explains how to design the ends so peel cannot start.
• The third door’s failure was at the interface. Adhesive left on one side and bare metal on the other is an adhesion failure: the adhesive never properly bonded to the skin. That points to surface preparation, and Chapters 2 to 4 explain how to measure and control it.

Figure 3Questions to ask before trusting a bonded joint's strength

• Avoid: Is the bondline loaded mainly in shear, and are peel and cleavage designed out?Chapters 6 and 7
• Avoid: Is the strength figure for this overlap and these adherends, not a short test coupon?Chapter 6
• Ask: Does the adhesive wet and adhere to the surfaces as they arrive in production?Chapters 2 to 4
• Ask: Is the cure complete at the shop's temperature and time?Chapter 5
• Ask: Can the bondline take the thermal movement between the parts?Chapter 8
• Ask: Has the joint been tested after ageing in heat and moisture?Chapters 9 and 10

When not to bond

Bonding is the wrong choice in some situations, and recognising them early saves trials. It is usually wrong when the joint must be taken apart in service, unless a release mechanism or a removable adhesive is designed in. It is a poor choice where the loading is mainly peel and the geometry cannot be changed, where service temperatures approach the adhesive’s limit, or where the surfaces cannot be prepared consistently, such as heavily oiled stock with no cleaning step. It is risky where failure would be dangerous and there is no way of testing the production process, because a bonded joint cannot be inspected into soundness.

Hybrid joints combine the strengths of both methods. A rivet or spot weld at the end of a bonded flange stops peel from starting, holds the parts while the adhesive cures and gives a back-up load path. In aircraft and vehicle bodies, rivet-bonding and weld-bonding are common for exactly this reason. For Plant K, the cheapest version of this idea, a single rivet at each end of each stiffener flange, turns out to be part of the answer in Chapter 7.

A quick screen for a new bonding job

Before a trial, a short screen sorts out jobs that are unlikely to succeed.

1. Which way does the load go? Sketch the joint and the loads. If the bondline is in shear, continue. If it is in peel or cleavage, change the geometry or add a mechanical fixing at the edges.
2. What are the materials and their surfaces as delivered? Oiled sheet, mill-finish aluminium, moulded plastics with release agents and powder-coated parts all need different preparation. Low-energy plastics such as polyethylene and polypropylene need special treatment or special adhesives.
3. What temperature range and environment? Note the hottest and coldest service temperatures, moisture, chemicals and ultraviolet exposure.
4. How much do the parts move relative to each other? Unlike materials or long joints with temperature changes need a bondline that can take the movement.
5. How will the process be controlled and proved? Decide what test coupons will be made, how often and to what acceptance level.

Plant K’s stiffeners passed the screen with one warning, the ends of the flanges, which the original design had not addressed. The window passed with a different warning, the movement between polycarbonate and aluminium, which the original adhesive choice had not addressed.

How this applies to a small Australian business

Bonding can remove whole operations, such as filling weld marks or sealing seams, and join materials that nothing else will. Before relying on it, check the loading direction, the surfaces as they really arrive, the temperature range and the movement between parts. Do not multiply a data-sheet strength by a bonded area: that number answers a different question. Plan from the start how the process will be tested, because a bonded joint has to be proved by its process rather than by inspecting the finished part.

Exercise: a bonded bracket

A 3 mm aluminium bracket is to be bonded to a 2 mm aluminium panel over a 40 by 60 mm area with an epoxy whose data sheet gives 22 MPa lap shear strength. (a) What capacity does the strength-times-area method give? (b) The bracket carries a 50 kg load hanging 150 mm out from the panel face. Which loading does the bondline see? (c) What would you change?

Worked answer. (a) The bonded area is 40 × 60 = 2,400 mm², and 22 × 2,400 = 52,800 N, about 53 kN. (b) The load of about 490 N acts 150 mm from the panel, producing a moment of about 73.6 N·m that tries to pry the top edge of the bracket away from the panel. The bondline sees cleavage, with the highest stress along the top edge, as well as some shear. The 53 kN figure is irrelevant to this loading. (c) Arrange for the load to be carried in shear, for example by bonding the bracket to a vertical face so the weight acts along the bondline, extend the bonded base upward to reduce the cleavage stress, or add a mechanical fixing near the top edge where cleavage would start.