Level 5 integrated systems test with mixed results

Blockers - do not accept handover

1. Battery runtime (finding 1). This is an unambiguous failure. A discharge test at 80% load should deliver substantially longer than the full-load design figure, not 28% less. Possible causes: undersized string, aged or mismatched cells, incorrect end-of-discharge voltage, or a test at elevated ambient temperature. Either way, the designed ride-through does not exist, and every downstream assumption about generator start and transfer depends on it.
2. CRAH auto-restart failure (finding 6). A unit that requires a local manual reset after a power restoration is a single point of failure in the operational sense: after any site-wide power event, the hall may run on reduced cooling until someone walks the floor. This directly contradicts the contract requirement of no single points of failure, and it also indicates an unverified control sequence that may affect other units intermittently. Root cause must be found in the control logic and the restart sequence, not just reset locally and re-tested once.
3. Transfer time (finding 3). Fourteen seconds exceeds the 10-second start-and-transfer criterion of NFPA 110 Level 1, which the contract explicitly references. The UPS masks the effect on the IT load, so this is not automatically an IT outage risk, but the acceptance criterion is contractual and the site cannot be signed off against a standard it fails without a documented, formally agreed deviation.

Conditional - accept with a dated closure plan

• Generator block load (finding 2). The measured dips (9% voltage, 3% frequency) and 4-second recovery are within commonly specified limits, so this passes typical criteria - but the test was a 65% step. If the specification requires a 100% single-step block load, this is a re-test item, not an acceptance.
• Sensor calibration (finding 5). Not a safety blocker, but PUE, capacity reporting and alarm thresholds cannot be accepted on uncalibrated data. Fix and re-verify before the M&V baseline is set.
• Chilled-water cold start (finding 4). Informational unless a recovery-time objective exists. It should be cross-checked against battery autonomy: if a full cooling restart takes 38 minutes, the site needs either longer ride-through or a documented staged-restart plan.

Battery string (finding 1)

• Investigate before retesting: verify the installed string against the design calculation (load profile, end voltage, temperature correction, ageing margin), check cell-to-cell connection torque, and review the test log for the actual end voltage used.
• Retest in accordance with the applicable IEEE maintenance standard for the cell type, at full design load, with calibrated load bank instrumentation and per-cell voltage and temperature recorded at intervals.
• Acceptance evidence: a signed test report showing the design runtime at full load with the specified end voltage, per-cell readings within tolerance, and a manufacturer's statement that the string meets the calculated design life at the installed ambient temperature. If the string is undersized rather than faulty, add capacity - a maintenance activity cannot fix a design error.

CRAH restart (finding 6)

• Review the control sequence, the restart logic, and the alarm history for the failed unit; check for a latent trip condition (low differential pressure, high condensate, control board fault) that presents only on restoration.
• Retest all eight units, not just the failed one, with a simulated power restoration while on load, and repeat three times to demonstrate repeatability.
• Acceptance evidence: all units restart automatically and reach steady-state operation within the documented time, with the sequence-of-events log and the control sequence narrative signed by the controls contractor and the commissioning agent.

Transfer time (finding 3)

• Measure the sequence by stage (utility loss detection, generator start, generator stable, transfer, UPS on generator) to determine whether the delay is in the generator, the ATS setting, or the sensing logic - a deliberately delayed transfer to allow generator stabilisation may be correct engineering but is still non-compliant with the stated standard.
• Either retest to achieve the criterion, or produce a formal technical deviation approved by the customer, showing that UPS autonomy comfortably covers 14 seconds plus margin and that the deviation is documented in the O&M regime.

Overall. No handover certificate should be issued until blockers 1 and 6 are closed and finding 3 is either retested or formally deviated, with the customer's signature on the record.

Design and as-built record

• As-built drawings for power, cooling, containment, fire and network routing, signed as accurate by the contractor, plus a marked-up comparison against the design intent showing every deviation.
• Single-line diagram, protective device coordination study and arc-flash study, with the settings actually applied in the field verified against the study.

Test evidence

• All Level 1 to Level 5 scripts with witnessed results, witness names and signatures, and a clause-by-clause statement of pass, fail or deviation against the specification.
• Factory acceptance test certificates for UPS, generators, switchgear, CRAHs and chillers; load-bank calibration certificates; battery test report; generator load and step-load test records.
• Sensor and instrument calibration certificates with traceability, and a point-to-point verification of every monitoring point against the DCIM or EPMS.

Operational readiness

• O&M manuals, spares list with on-site stock confirmed, and a maintenance schedule with the first year's activities dated.
• Training records for the operations team, including a witnessed demonstration that they can execute the restart and failover procedures unaided.
• A written sequence-of-operations document for loss of utility, loss of a UPS, loss of a chiller and loss of a generator, each validated during the integrated test.

Commercial and closure

• A punch list with each item assigned an owner, a closure date and a severity classification, and a defined retention or holdback tied to closure of the blockers.
• Warranty start dates and defect liability periods per system, and the escalation contacts for each vendor.

Why this matters. The commissioning report describes one day in the life of the facility. The evidence package is what lets the operator prove, two years later during an incident or an audit, that the facility was handed over in the state it was designed to be - and lets the operator defend an availability commitment to a tenant.