Where and When Roof Penetrations Should Be Formed

The record's first move is to separate the planes. Each interrupted plane needs its own return, and the returns are directional in different ways: a bond to the drainage plane has to be lapped so water passes over it, a bond to the air barrier only has to be unbroken, and a bond to the vapour control layer has to be unbroken on a particular side of the insulation. Making one good return and calling the perforation closed leaves the other planes open, and the symptom then appears wherever the open plane mattered rather than at the pipe.

It also separates the sleeve from the seal. The sleeve gives the building a clean, stable edge that a bond can hold to; the seal closes the remaining annulus around an object that will not stay still. Sealant applied straight into a rough perforation is being asked to be both at once, and it lets go at whichever edge moves first.

The warm deck record states the position rule in one sentence: grouping penetrations, and keeping them away from upstands and low points, is easier to detail than sealing each one where it happens to land. That is a design instruction about the roof layout rather than about any component.

The penetration record explains why the low points matter. Where a perforation sits on a sloping surface decides how much water reaches it, and placed on the upslope side of an obstruction, or close to an outlet, the same detail has to handle concentrated flow instead of the sheet flow it was conceived for. An outlet is itself a penetration, and it is the one place on a roof where flow is deliberately concentrated.

Where the plane crossed is horizontal, the record names a component whose purpose is purely positional: an upstand or raised termination that lifts the seal clear of the surface it sits in, so the joint is not the low point that water, silt and leaf debris collect against between visits.

The plinth record extends the same idea to equipment and warns where it goes wrong. A plinth puts a termination out in the field of the roof rather than at its edge, with water running past on every side, and when that termination ends up below the water it was meant to stand above, because a build-up was raised afterwards or an outlet silted, the roof's weather line stops underwater at a position concealed beneath the plant. Water then appears in the room below, commonly nowhere near the plinth itself.

The inverted roof record states the sequence rule plainly: grouping penetrations, and forming them before the roof is covered rather than after, is materially easier than working back down through a completed build-up. The penetration record agrees from the other direction, noting that perforations are frequently made after the envelope is closed, by a trade working from one side with no view of what is behind, and that marking, sleeving or forming the openings while both sides are still open changes what is physically possible later.

The warm deck record turns that into a question for professionals which is worth asking on any project: what provision is being made for future penetrations so that they do not have to be improvised into a finished roof. Asking it early costs a conversation. Not asking it produces a hole cut through a finished assembly by whoever has arrived with a duct.

On an inverted roof a penetration made after completion crosses ballast, filter, insulation and membrane, and the record notes each has to be reinstated in an order nobody can check afterwards. It also notes that the penetration ends at a membrane nobody will see again, which removes the possibility of verifying the work that has just been done.

On a composite panel roof the consequence is structural as well. Anything crossing the panel breaches the outer face, the core and the liner in one move, and cutting the panel interrupts the member that was spanning between purlins. The record asks how roof-mounted plant will be supported without cutting panels in unplanned positions, which is a different question from how a pipe is sealed.

On a built-up twin-skin roof the record is specific about which return gets missed. Large single-storey roofs carry a great deal of plant, and every support and duct passes through both sheets and the insulation, so the liner has to be resealed around each one rather than simply cut, or the assembly's air control is undone item by item.

It also describes the symptom that follows, which does not look like a penetration failure at all. A break in the liner lets warm moist air from the space below into the cold upper part of the build-up, where it condenses on the underside of the outer sheet and runs to somewhere unrelated before it appears. Diagnosing that as a leak in the outer sheet is the mistake the assembly invites most often.

The record's reason for treating a penetration as an assembly rather than a repair is that the penetrating element will not hold still. Pipes lengthen and shorten with what they carry, ducts pass on vibration from plant, cable bundles are pulled and rerouted, and structural members deflect under load. A seal formed against a moving object either moves with it or is progressively defeated by it.

That same rigidity is what makes replacement destructive. Cutting a bonded service free takes the returned planes with it, which is why the record treats how a penetration is expected to be maintained as an input at the start rather than a discovery later, and asks whether the seal and its returns can be opened and remade without cutting into the surrounding plane.

The record names the organisational failure without ambiguity. A penetration belongs to a services package and the plane it crosses belongs to an envelope package, and that split is why the collar is frequently in nobody's scope. It asks which package is responsible for the collar itself, as opposed to the pipe on one side and the membrane on the other.

It also asks whether the perforations are being set out in advance, or created by whichever trade arrives with a service to run. Those two questions together decide whether a roof ends up with a planned set of crossings or an accumulated one, and the difference is not visible from the finished surface.