Fully Adhered EPDM in a Late-Season Cold Snap
A 40,000 sq ft office roof in the upper Midwest receives fully adhered 60-mil EPDM over 2-in. polyiso. The specification calls for a solvent-based bonding adhesive at a coverage rate of 1 gal per 90 sq ft, with the membrane rolled into the adhesive after the adhesive has flashed off. The crew starts at 7 a.m. on an overcast October day: 41 F at the deck, rising to 55 F by mid-afternoon, dropping to 33 F that night. To keep pace, the crew rolls adhesive at roughly 1 gal per 110 sq ft and lays membrane only 2 to 3 minutes behind the adhesive. Three days later, about 60 blisters and wrinkles, 6 to 24 in. across, appear across the first 6,000 sq ft, clustered mostly over insulation board joints on the north half of the roof. A blister is cut open: the adhesive on the substrate is still glossy and stringy, and a heavy wet transfer of adhesive remains on the back of the membrane. A moisture meter reads 12% in the polyiso facer near one joint.
1. What failure mechanism best fits this evidence, and what alternative mechanism would you rule out, and how?
Read the worked answer
The physical evidence points to trapped solvent, not a wet substrate.
The tell is the condition of the adhesive. Glossy, stringy adhesive on the substrate plus a heavy wet transfer on the membrane back means the membrane was laid into adhesive that had not flashed off. Solvent was sealed in the adhesive layer, then warmed by the sun and by the 55 F afternoon, and the vapor pressure lifted the membrane into blisters. Three conditions in the setup all push the same direction: coverage 22% below spec (thicker adhesive films hold solvent longer), membrane laid within 2 to 3 minutes, and a 33 F night that drastically slows solvent release and cure.
Why the substrate moisture theory is weaker: if moisture were driving the blisters, you would expect the adhesive to be dry and stuck to the membrane rather than still wet, water droplets or beading on the membrane underside, and blisters concentrated over wet insulation fields rather than uniquely over board joints where adhesive puddles. A 12% facer reading is elevated and worth a survey, but one reading near a joint does not explain wet adhesive everywhere.
Before writing the report, cut a second blister over a visibly dry area away from joints. If the adhesive under it is also wet, the mechanism is confirmed and the problem is system-wide, not local.
2. Define the corrective action for the affected 6,000 sq ft, and state the threshold at which you would stop repairing and remove.
Read the worked answer
Repair is only defensible where the adhesive bond is verified, not where it merely looks flat.
Step 1, evaluate: cut and open representative blisters across the whole affected area, including at least one in each 'zone' between board joints. Run peel or pull tests on the flat membrane to see whether the adhesive is stuck, transfer-bonded, or still wet. If the flat areas are wet beneath the sheet, there is no repair for that panel.
Step 2, repair the localized blisters: cut in a plus or X pattern (or remove a patch), clean and dry the substrate and membrane back, apply the specified adhesive at the specified rate, allow full flash-off, roll in, and install a properly sized and rounded patch with splice adhesive or tape per the manufacturer. Do not cut a slit and glue the flap; do not weight blisters down and walk away.
Step 3, threshold: if the wet-adhesive condition covers a large continuous share of the area, or if pull tests show a consistent absence of bond, remove and replace the membrane and adhesive in that area. Patching 60 blisters over a wet adhesive bed produces a roof that will blister again next summer, and the cost of repair plus future callbacks exceeds replacement now. Get the manufacturer's field representative to sign off on whichever path is chosen, and photograph everything.
3. What must change before the crew continues on the remaining 34,000 sq ft?
Read the worked answer
Treat the first 6,000 sq ft as a failed test patch and change the process, not just the schedule.
Temperature: stop solvent-based bonding adhesive work when the ambient or surface temperature is at or below the manufacturer's minimum, or is falling toward it. The 41 F start against a 33 F night is outside normal solvent adhesive practice. Options are a manufacturer-approved cold-weather or low-temperature adhesive, or waiting for a suitable window. Get the limit in writing from the manufacturer for the specific product.
Coverage: enforce 1 gal per 90 sq ft with measured pours or a marked container, and count squares per pail. The 110 sq ft/gal rate is a specification violation, not a rounding error.
Flash-off: establish a go/no-go check the crew can actually perform, such as a dry-touch test and a visual check that adhesive no longer transfers to a clean glove. Set a rule that membrane is laid only over adhesive that has flashed, and limit the size of the adhesive run so it does not get ahead of the membrane.
Rolling: require roller use per manufacturer after the membrane is in place, and require that wrinkles and bridging be worked out before adhesive sets.
Controls: a documented surface temperature log, a manufacturer's representative on site for the restart, and a new test patch verified before full production resumes. Confirm with the manufacturer whether a primer is required at these temperatures.