A Quarterly Generator 'Test' That Proves Nothing

Read the findings together - individually each is ambiguous, together they describe a degrading set that has never been proven.

• 250 kW on a 1,440 kW prime rating is about 17% load. Running a diesel at that level causes incomplete combustion: cylinder glazing, unburnt fuel in the exhaust and, critically, fuel washing past the piston rings into the sump. An exhaust temperature of 190 °C under light load, read against the manufacturer's expected exhaust temperature for that load, points the same way.
• The 4% fuel dilution in the lube oil confirms it. Fuel is entering the oil. That accelerates wear on bearings and rings and, in the extreme, creates a crankcase vapour ignition risk. This is a consequence of the test regime, not an unlucky failure.
• Two crank attempts plus a charger in alarm on a silenced BMS point means the starting battery has been sitting on open-circuit voltage with no recharge. The engine started only because residual charge sufficed. The next event may be a no-start, and it will occur during a utility outage, not during a scheduled test.
• Nine-month-old diesel with no test is at the edge of practical storage life. Water, sediment, microbial growth and oxidation products all manifest as filter plugging under load - again, only when the generator is actually needed.
• What the test proves: the engine can turn and run. What it does not prove: the ATS transfer and retransfer sequence, the UPS on battery, the chilled water restart and step-load recovery, the generator's ability to accept the building load in steps, or fuel-system delivery at load. In other words, almost every element the facility depends on.

Conclusion: the regime is not a test, it is a start verification, and it is actively damaging the asset.

Replace 'run for 30 minutes' with measurable acceptance criteria in the contract.

• Load-bank testing: annually, and after any major engine or fuel-system repair, at a minimum of 30% of prime rating - preferably 50–80% - for long enough to reach stable temperatures (typically 2–4 hours). Record load, exhaust temperature and oil pressure at each step and compare against the manufacturer's figures for that load. The purpose is to burn off deposits and to prove the set can carry real load, so the criterion must be stated in kW and duration, not in engine hours.
• Full transfer test: at least quarterly, with the ATS actually transferring the building load, the UPS observed on battery, and the complete cooling restart sequence included. Measure and record the transfer time, the time to accept load, and any load steps. This is the only test that exercises the failure the facility is buying the generator for.
• Fuel management: lab test to the applicable diesel specification (water, sediment, microbial, flash point, cetane, FAME content) on delivery and annually thereafter; periodic polishing and water removal from tank sumps; a dated stock rotation policy so fuel is consumed and replaced within its storage life rather than sitting for years.
• Starting battery: route charger-fail alarms to a monitoring path that cannot be silenced indefinitely, add battery monitoring (float voltage, impedance) with an annual capacity test, and hold a maintained spare battery on site. The criterion is not 'it started' but 'the battery passed a capacity test'.
• Alarm governance: quarantine nuisance alarms by fixing the cause; an alarm-silencing culture converts a monitoring system into decoration.
• Reporting: monthly trend of exhaust temperature, oil analysis, fuel test results and load achieved, reviewed against previous months rather than filed.

Treat the generator as being in a restricted-confidence state and say so in writing, then correct in order of immediate risk:

1. Battery and charger, within days. This is the cheapest, fastest fix and the highest-probability cause of a no-start. Replace the charger, test or replace the battery, and prove two consecutive clean cranks.
2. Fuel, before any load test. Test the bulk tank and day tank, polish or replace the fuel, and change filters. Running the set hard on contaminated fuel plugs filters mid-test and creates a false failure - or a real one during an outage.
3. Then a supervised load-bank and transfer test at 50% or more of rating, with the building load transferred, the UPS observed on battery, the cooling restart included, and the step-load profile recorded against the manufacturer's curve. Only this test re-establishes confidence.
4. Interim mitigation while 1–3 are open: assess whether a mobile generator can be connected, review the published risk with the business, and document it in the emergency operating procedures - including who authorises a controlled shutdown if the utility fails.
5. Then repair the engine consequences of prolonged light-load running: borescope inspection, oil and filter change, injector check, and a repeat oil analysis after a proper loaded run.

The sequencing matters: testing before the fuel and battery are sound risks adding a second failure on top of the first, and it destroys the credibility of the very test you need the business to trust.