What an Adhesive Joint Depends On

Why this matters

An adhesive datasheet hands you a large, confident number, and a bonded joint that holds a fraction of that number in service is not a defective product. It is a joint where something other than the adhesive was the weakest link, and the datasheet never claimed to describe that something.

The number on the sheet is a stress, measured on a specific coupon geometry, in one loading mode, at one temperature, at one bondline thickness, taken to failure in a short test on a fully cured sample. Every one of those conditions is a place where your joint differs from the test. If you know which ones you changed, the datasheet becomes useful. If you do not, it is a number that makes a bad joint feel engineered.

Before the container opens

Read the safety data sheet for the specific product and follow it. That is your right and your employer's duty under 29 CFR 1910.1200, and the reason it matters here is that adhesive hazards are mostly inhalation hazards, which a glove does not touch.

• Solvent-carrying and solvent-cement products give off vapor continuously while open and while curing. Work with ventilation that moves air away from your breathing zone, and where the product's safety data sheet calls for respiratory protection, use it under a written program per 29 CFR 1910.134. A dust mask is not a vapor control.
• Two-part urethanes and some primers contain isocyanates, which are respiratory sensitizers: once someone is sensitized, small later exposures provoke a reaction. Ventilation and the respiratory protection the safety data sheet specifies, not gloves alone.
• Cyanoacrylates bond skin in seconds and react exothermically with cotton and wool, enough to burn. Nitrile gloves and eye protection, and never wipe a spill with a cotton rag.
• If you abrade a painted or primed surface as part of the prep, you are creating an airborne hazard from whatever the coating is. Old coatings may contain lead (29 CFR 1910.1025 in general industry, 29 CFR 1926.62 in construction) or hexavalent chromium (29 CFR 1910.1026 and 29 CFR 1926.1126). Both are respiratory controls, not glove-and-goggle controls, and unknown coatings on older equipment are presumed to contain them until assessed.

Four elements in series

A bonded joint is a chain, and its capacity is the weakest link, not the strongest.

The adherend. Whatever the adhesive is stuck to has to carry the load into the bond. A bond to a painted surface is a bond to paint, and the joint's real capacity is the paint's adhesion to the substrate underneath. Same for a bond to a plasticized plastic, a galvanized coating, an oxidized aluminum skin, or laitance on concrete.

The interface. Adhesive to surface. This is what surface preparation decides, and a sibling article owns it.

The bondline. The cured adhesive itself, in bulk. This is the only element the datasheet number describes.

The envelope. Temperature, moisture, chemical exposure, and above all duration. An adhesive can be strong and still not hold a load you leave on it.

Three of those four are outside the datasheet. That is not a criticism of datasheets, it is the reason a bonded joint has to be thought about as a joint rather than as a product selection.

What the published number was measured under

Every structural adhesive strength value carries conditions, and they are usually stated in small print right beside it:

• A geometry, most often a single-lap coupon of stated overlap and adherend thickness. The stress in a lap joint is not uniform: it peaks at the two ends of the overlap and sags in the middle, so the published value is an average over an area that was never uniformly loaded.
• A loading mode. Shear values and peel values are not even in the same units, shear in force per unit area and peel in force per unit width, which is the warning that they are not comparable. A sibling article covers the mode question and the arithmetic of peel.
• A temperature, commonly around room temperature. Most polymers lose a large share of their stiffness and strength over a fairly narrow band as they approach their softening range, and the manufacturer usually publishes a retention curve. Use the curve, not the headline number, for a joint that runs hot.
• A duration. A test coupon is loaded to failure in minutes. A bracket in a mechanical room carries its load every second for years, and polymers creep under sustained load. The continuous-load allowable is a different, much smaller number, and it belongs to the manufacturer.
• A cure state. Full strength requires the full cure schedule, and cure rate is temperature and humidity dependent for most chemistries. Handling strength is not working strength.

The two geometry facts worth carrying

Bondline thickness has a range, in both directions. Too thin and the joint is starved, with dry patches and no ability to accommodate movement; too thick and most structural adhesives lose strength, because the bulk polymer is weaker than a thin, constrained layer. Manufacturers specify a range. Control it with a deliberate spacer rather than by feel: glass beads mixed into the adhesive, a wire shim, or a molded stand-off.

More overlap gives less than proportional strength. Because stress concentrates at the ends of the overlap, extending the overlap adds area in the middle where the stress is already low. Doubling the overlap on a lap joint does not double its capacity, and past a certain overlap-to-adherend-thickness ratio the added length contributes very little. Where the load is close to the joint's limit, adding width helps more than adding length, and a small fillet of adhesive left at the ends of the overlap reduces the stress peak that governs. The specific numbers here belong to the adhesive manufacturer's design data or to a test coupon made from your actual materials.

The joint sheet, filled in

A bracket bonded to the outside of a painted steel plenum, carrying a small sensor assembly. Illustrative values used to carry the method.

That 120 looks like an enormous margin, and it is the number that gets quoted in the shop. Now correct it, one condition at a time.

Temperature. The plenum runs near 140 F in summer. The manufacturer's retention curve for this product shows roughly 35 percent of room-temperature strength at that temperature, giving about 630 psi. Margin against the same 15 psi service stress: about 42 times.

Duration. The load is permanent, and this product's continuous-load allowable is a fraction of the short-term value. Take it as 10 percent of the elevated-temperature figure, about 63 psi. Margin: about 4.2 times.

Note what happened to the comparison. The original 120 times was computed against an uncorrected number, and both later figures are corrected values against the same 15 psi service stress, so the honest statement is that correcting the conditions alone, with no change to the joint, took the margin from an uncorrected 120 times to a corrected 4.2 times. Nothing about the joint changed. Only the honesty of the denominator did.

Adherend. And here is where the joint actually fails, because none of the above touches the weakest link. The plenum is painted. The adhesive is bonded to paint, and the paint's adhesion to the steel underneath is an unknown that no adhesive datasheet addresses. It is testable, with a coating adhesion test on the actual surface, and it is not guessable. On a warm plenum with an aged coating, it is entirely plausible for the coating interface to be the lowest capacity in the chain by a wide margin.

How the failure reads. When it lets go, look at the failure surface. Adhesive on one side and clean paint on the other means the adhesive-to-paint interface failed. Adhesive with paint stuck to it, and bare steel on the plenum, means the paint was the weak link and the adhesive did its job perfectly. Those two look identical from across the room and mean opposite things, and only one of them is a reason to change adhesives.

What that changes about the design. Either take the coating off in the bond footprint so the adhesive lands on prepared metal, which makes the surface prep article the governing document, or stop asking the bond to carry a permanent load and add a mechanical fastener that carries the dead weight while the adhesive seals and locates. The second option is usually right for a bracket on painted equipment, and it is not an admission of failure: a bonded joint that carries no sustained load has no creep problem at all.

What flips the recommendation

The whole analysis above assumes a permanent load in a warm environment on a coated substrate. Change any of those and the answer moves:

• Intermittent or transient load only, with no sustained component, removes the creep correction entirely and puts you back near the short-term allowable at your service temperature.
• A joint that must accommodate movement between two materials with different expansion rates wants a flexible adhesive in a thicker bondline, deliberately, and the low-modulus product's much lower strength number is the correct trade. A rigid structural adhesive across a mismatched pair concentrates thermal strain at the bondline ends and cracks itself.
• A cold or cryogenic service flips the failure mode the other way: many adhesives get stronger and more brittle as they cool, so the governing risk stops being creep and becomes shock and thermal cycling.

How to verify you got this right

• Build a coupon from your actual materials, cured the way the job will be cured, and load it. This is the only way to include the adherend and the interface, which is to say the two elements the datasheet cannot cover.
• Measure bondline thickness on a scrap sample by curing and sectioning it, or by using a deliberate spacer whose size you know. A bondline nobody controlled is a variable nobody can defend.
• Confirm cure schedule against the actual conditions, not the label maximum. Cold and dry slows most chemistries substantially, and a joint loaded before it is cured has been damaged in a way that does not show.
• Write the service temperature and load duration on the job record next to the product name. Six months later, when it lets go, those two lines are what tell the next tech whether the product was wrong or the joint was.

References

• 29 CFR 1910.1200, hazard communication, for the safety data sheet of the specific adhesive
• 29 CFR 1910.134 for respiratory protection under a written program where the safety data sheet requires it
• 29 CFR 1910.1025 and 29 CFR 1926.62 (lead), 29 CFR 1910.1026 and 29 CFR 1926.1126 (hexavalent chromium), where abrading an existing coating is part of the preparation
• Adhesive manufacturer documentation for bondline thickness range, temperature retention curve, continuous-load allowable, and cure schedule
• See related: What an Adhesive Needs in Order to Bond; Why Surface Preparation Decides an Adhesive Joint; Thermal Expansion Mismatch Inside a Joint