In a GMP plant a work order is the record that equipment went back into service in its qualified state. Before it closes it needs as-found and as-left values, parts by number, technician sign-off with accountable review, and links to any deviation or change control, all kept under an Annex 11 audit trail.
Long-lead AHUs are often ordered before a single end-of-design DQ report is signed, so a casing leakage class of L2 against a URS that asked for L1 surfaces at IQ as a deviation. Splitting DQ along the procurement packages, with each section signed before its order is released, finds it while the order can still change.
During start and stop, whichever fan moves first decides which way air flows through the door gaps. Positive suites start supply first and stop extract first; containment suites do the opposite. Interlock the order, define the response to a single fan trip and test it at OQ with a real power cut.
IQ should start from a verified installation. Before a turnover package moves from construction to qualification it needs a marked boundary, every IQ-blocking punch item closed, commissioning and balancing results, walked-down as-builts, calibration certificates and the equipment file. If one is missing, the package goes back.
Two accountable parties on one RACI row means nobody decides. The rows most likely to collect two As cross a handover: design changes after tender, vendor drawing approval, long-lead releases and readiness for IQ. Read every row as "X decides" and fix the ones that need two names before kickoff.
Operating theatres and isolation rooms rarely get a shutdown window, so maintenance fits between surgery lists. An isolation room may hold only 2.5 Pa negative, which one open ceiling panel next door can erase. Plan the dampers that hold the neighbours, the dirty-filter route and the tests before handback.
EU GMP Annex 15 allows some FAT results to be used at IQ and OQ when transport and installation do not affect them. For an AHU that means listing what the trip can change, from fan alignment to software version, retesting those on site and referencing the rest.
A loaded filter sheds part of its dust cake when the clamps release. Withdraw it through the upstream access section or sideways through a casing door, never through the section behind it. Then check every clamp by hand and log the new filter's differential pressure as the baseline.
ISO 16890 rates what a filter captures; Eurovent energy classes, A+ to E, compare what it costs to run at reference conditions. Two ePM1 70% filters 50 Pa apart in average pressure drop differ by about 2,000 kWh a year at 10,000 m³/h. Ask for the class and the pressure drop at your airflow.
A plant room at 32 °C and 70% RH has a dew point near 26 °C. With 12 °C air inside, a TB3 casing frame sits at 21 to 24 °C and drips, while TB2 spans 24 to 27 °C and only its upper end clears. Ask for the tested kb and check it against the room's dew point.
MBH means thousands of Btu/h: 24 MBH is 24,000 Btu/h, about 7.0 kW. Entered as kW it oversizes the room about 3.4 times. Before converting, check whether the datasheet figure is electrical input or heat to the room, and keep the latent part for the moisture balance.
Shrinking a 10,000 m³/h duct from 700 × 500 mm to 600 × 400 mm raises velocity from about 8 to 12 m/s and velocity pressure from 38 to 80 Pa, so every fitting costs about twice. Set separate limits for mains and branches in the design basis.