MEP Coordination & Interfaces

The fundamental difficulty is organisational before it is technical. In a conventional project one services contractor installs more or less everything in sequence on one site, and coordination happens continuously as the work proceeds. In a modular project the services are split across a factory-site boundary: the manufacturer installs and tests the services *inside* the modules, often months before site work, while the main contractor and services engineer handle the *building-wide* services — the incoming mains, the primary risers, the central plant, the distribution that serves more than one module. Two organisations, two programmes, two workplaces, building two halves of one system that must meet as if they were designed by one hand.

The tool that makes this work is an interface schedule (sometimes an interface register or responsibility matrix): a disciplined document that lists every point where the factory scope meets the site scope and records, for each, exactly what is provided on each side, who makes the connection, and who tests it. Every pipe tail, every cable, every duct end, every riser junction is an entry. Nothing is allowed to fall into the gap between 'the factory assumed the site would do it' and 'the site assumed the factory had done it' — the gap where modular services most often fail. The interface schedule is not bureaucracy; it is the design of the boundary made explicit.

The designer's contribution is to make the boundary as simple, clear and few as it can be. A services strategy that hands the factory a clean, standardised scope (all in-module first- and second-fix) and the site a clean scope (mains, risers, plant) with a small number of well-defined connection points is far easier to coordinate than one where responsibility zig-zags back and forth. This is the same DfMA logic applied to organisation: reduce the number of interfaces, standardise the ones that remain, and define them precisely. And it is why the services engineer and the manufacturer must be coordinated from the earliest design stages — the interface cannot be retrofitted once both halves are built. The binding content of each side's design (the sizing, the loads, the compliance) is the engineer's and the manufacturer's; the clarity and discipline of the boundary between them is a coordination task the whole team owns.

The interface becomes physical at the connection points — the specific places where the factory's services are joined to each other and to the site's services. These cluster at the module joints: where modules stack, the risers must be coupled; where modules sit side by side, horizontal runs must be joined across the corridor or party line; where the modular structure meets the in-situ podium or core, the in-module services must connect to the building mains and the primary risers. Each of these is a designed connection, not an incidental meeting of pipes.

Three properties make a connection point work. First, access: the joint must sit in a reachable zone — a ceiling void, a dedicated service riser cupboard, a removable access panel — because a connection you cannot reach is one you cannot make, test, or maintain over the building's life. A joint buried inside a sealed module wall is a latent defect. Second, tolerance: the connection must accommodate the real, slightly-imperfect position of two independently-built elements; modules are not placed to the millimetre, so the connection detail needs flexible couplings, telescopic sections or deliberate slack to close a gap that is never exactly the drawing dimension. Third, speed and reliability: many systems favour plug-and-play connectors — proprietary push-fit or pressed couplings, pre-terminated electrical connectors, flanged duct joints — so that connections are quick, repeatable and hard to get wrong on site, which matters when the same joint is made hundreds of times up a building.

The designer's job is to locate these connection points deliberately, give them access and tolerance, and keep them few. Concentrating services around aligned risers means the connections stack neatly in one accessible shaft rather than scattering through the building. Keeping long runs within single modules reduces the horizontal crossings. Grouping the mains connection into a defined plant interface keeps the factory-to-building junction in one coordinated place. The connection details themselves — the specific couplings, the fire-stopping where a service crosses a compartment line, the acoustic treatment of penetrations, the structural fixings — are binding engineering that belongs to the services engineer and the manufacturer's tested system. What the architect and designer own is the *geometry and logic* of the interface: where it is, that it is reachable, that it tolerates reality, and that there is no more of it than the building needs.

On a site-built job, a services clash — a duct that runs into a beam, two pipes that want the same space — is often discovered and resolved on site by trades with room to adjust. In a modular building that safety net is gone: a clash discovered in a finished, delivered module is a grinder, a scrapped box, and a hole in the programme. So modular MEP leans hard on digital coordination to catch clashes *before* anything is built — in the model, where a fix is a mouse-click, not a demolition.

The method is a federated BIM model: the architectural, structural and services models, plus the manufacturer's own detailed module model, combined into one coordinated whole. Against that federated model the team runs clash detection — automated checks that flag where two elements occupy the same space (a hard clash) or where a service lacks the clearance it needs for installation, access or insulation (a soft clash). Every clash is logged, assigned and resolved in the model before the information is released to the factory. Because the module will be built exactly as modelled, at speed and many times, the model has to be right: the federated, clash-checked model becomes the single source of truth that both the factory and the site build to, and the thing that guarantees the factory's half and the site's half will actually meet.

This is where the early-commitment discipline of lesson 8.1 pays off. You cannot clash-detect a design that is still fluid; BIM coordination forces the services, structure and architecture to be developed together and resolved before manufacture, which is exactly what the factory needs. The model also carries the interface information — the connection points, their tolerances, their access — so the boundary is coordinated digitally, not discovered physically. Modular construction and BIM are natural partners for precisely this reason: the off-site process demands the up-front, coordinated, clash-free information that BIM is built to produce, and Module 10 returns to BIM as a practice tool. The binding design within the model — the sizing, the loads, the compliance, the fire and acoustic engineering of every penetration — remains the services engineer's and the specialists'; BIM is the medium in which the team coordinates that design and proves, before cutting metal, that all the pieces fit.

The last piece of the interface is commissioning — proving the finished services actually work, safely and to the design intent — and in a modular building it is split across the boundary just as the installation is. The factory carries out factory testing of the services within each module before it ships: pressure-testing pipework, continuity- and insulation-testing circuits, checking drainage falls, often witnessed and documented so each module leaves with a test record. This is a real quality gain — a defect is caught and fixed at the bench — but it proves only the services *inside* the module, under factory conditions, in isolation.

The services only become a working building system once the modules are connected on site, so a second layer of commissioning happens after assembly: every site-made joint is tested (pressure and flow for water and drainage, continuity and insulation for electrical, leak and balance for ductwork), the connections to the mains and risers are proven, and the integrated system — pumps, controls, ventilation, fire systems — is commissioned and balanced as a whole. The commissioning plan must therefore span the boundary explicitly: it says what is tested in the factory and recorded, what is re-checked after transport (because a joint that passed at the bench can be disturbed by the journey and the lift), what is tested at each site connection, and how the whole system is finally proven. Who witnesses and signs off each stage is part of the interface schedule.

Two honest cautions belong here. First, a bench test is not a site guarantee: transport, craneage and handling can disturb a joint that was sound in the factory, so re-testing after installation is not optional box-ticking but essential. Second, split responsibility is a risk: when the factory tests its half and the site tests its half, it must be unambiguous who owns the *joints between* and the *integrated whole*, or a problem at the boundary becomes a problem nobody owns. The designer's role is to ensure the commissioning strategy is planned across the boundary from the outset and that access is designed in so every joint can be reached and tested — while the binding commissioning itself, the test regimes, the performance criteria and the sign-off, belongs to the services engineer, the commissioning specialist and the manufacturer's tested system. Defer the binding MEP design and its proving to those qualified hands; design the boundary so their work can be done.