What a Four-Point Load Test Proves Before a Generator Ships

A generator that starts, idles and holds 400 V on a meter has passed the easiest test in the book. The failures that cost money on site — a governor that cannot accept a step load, a radiator that was sized for a cooler day, a breaker set below the alternator rating — stay hidden until the set is carrying a real building.

Why a start test is not a test

At no load, the engine is doing very little work. Fuel delivery is at its minimum, the turbo is barely spinning, coolant flow is at its lowest, and the alternator is producing only excitation current. Almost every component that fails in service is comfortably inside its limits in this state. Running a set unloaded for ten minutes tells you the wiring is correct. It does not tell you the machine will hold a hospital at 02:00.

The four points, and what each one reveals

Cunmi runs every set through four resistive load steps before it is packed. Each step is chosen because it exposes a different class of fault.

Twenty minutes at each step is enough for temperatures and pressures to stabilise. A set that only sees a two-minute bump at full load never reaches thermal equilibrium, so the most expensive faults — the ones that appear after forty minutes of running — are never triggered.

The overload block: one minute above rating

Prime-rated sets are expected to carry a 10 % overload for one hour in twelve. We verify that capability rather than assume it: a short block at 110 % confirms the engine has fuel headroom and that the protection curve does not trip inside the permitted window. A set that trips at 105 % will fail the first time a chiller bank restarts.

What usually fails on the first run

• Governor droop outside the declared class. A set sold as G2 that behaves like G1 will dip further than the load can tolerate when a large motor starts.
• Cooling margin consumed by the enclosure. The radiator was selected on the open set, then packed into a canopy with 40 °C ambient. The 100 % step finds this in an hour; a site finds it in August.
• Protection set too tight. Breaker and relay settings copied from a larger set trip the machine below its own rating.
• AVR drift under load. Voltage that sits at 400 V unloaded but sags to 385 V at full load is an excitation problem, not a load problem.

What you should receive with the set

A load test is only useful if it produces a record. For every Cunmi set we issue a test sheet carrying the serial numbers of engine and alternator, the four load steps with voltage, frequency, current and power factor at each, coolant and oil temperature at stabilisation, fuel consumption at 100 %, protection settings as left, and the signature of the technician who ran it. That sheet is what an installer needs when the commissioning engineer asks whether the machine has been proven.

FAQ

Is a load test the same as a factory acceptance test?

No. A factory acceptance test is a broader document that can include dimensional checks, wiring inspection, protection calibration and witness testing by the client. The load test is one chapter of it. For export orders we can arrange witnessed testing with remote video if the buyer cannot travel.

Can a load test be done on site instead?

Yes, with a rental load bank, and it is worth doing at commissioning on critical installations. The difference is timing: a fault found in our bay is a scheduling problem; the same fault found on site is a delay, a crane, and a conversation with the client.

Do you test containerised and silent sets as well?

Every configuration is tested in its final enclosure, because the enclosure is part of the thermal and acoustic system. Testing an open set and then fitting a canopy proves nothing about the package the customer receives.

Ask for a specimen test report