Ninety-Minute Water Demand, Fifty Minutes of Tank

The arithmetic first. 2,000 L/min for 90 minutes is 180,000 L, or 180 m3. The reserve is 100 m3. At the rated flow the tank is exhausted after 50 minutes, so the provision covers about 56% of the required duration. That is the headline failure and it does not need a hydraulic model to establish.

Then the qualifications that make it worse.

• Not all 100 m3 is usable. Vortexing and the clearance between the suction and the tank floor mean the effective volume is perhaps 90-95 m3, taking the duration to around 45-47 minutes.
• The 2,000 L/min is a pump curve point, not a system flow. At 7 bar at the pump, friction loss through a ring main and 12 hydrants reduces pressure at the far hydrants. If two hydrants are opened for cotton, which a deep-seated Class A fire will demand, the system flow may not reach 2,000 L/min at all, or the pressure at the branch collapses.
• Cotton is a deep-seated fire. Baled cotton smoulders internally and re-ignites after apparent extinguishment; it needs sustained, high-volume application, often for many hours, not a 90-minute window. The surveyor's 90 minutes is a floor, not a target.
• The pump is unverified. Eighteen months without a churn test, and a log with identical single signatures across 78 weeks, means the pump's condition is unknown. Diesel fire pumps fail on battery, fuel quality, controller faults and seized impellers. A pump that has not been run is not a control; it is an assumption.

Finding. The installation cannot meet the required duration, the flow is unverified at the point of use, and the pump's reliability is undocumented. This is not a paperwork gap to be closed with a fresh log book.

11 minutes 20 seconds for 340 people is too slow for this fire load. A cotton and polyester warehouse can go from smouldering to full involvement quickly, and the building should be clear in a fraction of that time. The number is a symptom; the cause is usually travel distance, exit width, or the absence of anyone directing people.

The seven missing people are the serious finding. A muster discrepancy does not mean seven people were lost; it means the accountability system did not work. They left by the north gate, which tells you:

• the north gate was open and unsupervised during the drill;
• no area warden swept the finishing hall or the north aisle;
• there was no gate control or roll-call reconciliation procedure;
• most importantly, someone would have declared 'all clear' with seven people unaccounted for. In a real event that declaration is what stops the search, and it is how people die in buildings they had already left.

The inaudible alarm is a design failure, not a maintenance failure. In a hall at 92 dB(A), a conventional bell or voice alarm cannot be relied on to exceed ambient. The alarm coverage needs to be measured, not assumed: sound level survey at head height across the hall, tonal alarms selected to sit in a frequency band not masked by loom noise, visual strobes for anyone wearing hearing protection, and a distinct evacuation signal that is not confused with the machinery alarms.

Also note the absence of an incident command structure. The drill gives no evidence of a named incident controller, a deputy, a fire team with a defined role, or a liaison point for the fire service on arrival. The plan exists on paper; the drill shows it was not exercised as a system.

Corrective actions. Unannounced monthly drills with a timed target; warden sweep checklists signed per area; controlled gates with a nominated gate marshal; photo or card-based muster roll at the assembly point; measured alarm coverage; and a post-drill debrief with the workers, not just the managers. Repeat and record until the discrepancy is zero.

The 18 minutes defines the design basis. For the first 18 minutes this site fights the fire alone, or it loses the building. Everything must be designed for that reality rather than for the assumption that the brigade will arrive.

What I would defend.

• Detection before smoke logging. Aspirating smoke detection in the cotton store, because open flaming is a late stage in baled cotton and by then the hall is untenable.
• Automatic suppression matched to the hazard. A sprinkler or deluge system designed for a high-challenge Class A storage fire, with the hydraulic calculation to prove it, not just a ring main and hydrants that depend on people being present and trained.
• Compartmentation. A fire-rated wall between the cotton store and the polyester, and between both and the diesel tank. The diesel tank inside the perimeter fence, near 30 t of baled cotton, is a siting decision that should be reversed with a bunded, separated location.
• Water. The reserve increased to the calculated 90-minute demand plus a margin, or a guaranteed alternative supply such as a fire water main or a documented water relay arrangement with the estate.
• Pump reliability. Weekly churn testing with automatic logging, an annual flow test against the curve, a jockey pump and a standby, and the log counter-signed.
• Pre-incident planning. A joint exercise with the state fire service, a hydrant map at the gate, an entry plan, and a named liaison who meets them on arrival.

What I would refuse. 'We have extinguishers and trained staff.' 'The brigade is only 18 minutes away.' A fresh churn-test log with no physical test. A fire team without breathing apparatus or drills. Reducing the cotton inventory while keeping the same unprotected building and calling it risk reduction - it may help, but it is not a substitute for the missing 80 m3 and the unverified pump. And any argument that the diesel tank is acceptable because nothing has happened yet.