Machinery testing and commissioning — proving it before it matters

A machine arrives on site having already passed tests at the works. Those tests proved specific things and not others, and most commissioning problems live in the gap between what was tested and what was assumed.

Works testing: three different tests

They are routinely confused, and they answer different questions.

Mechanical run test

The machine runs at speed, usually on air or water rather than the process fluid, for a defined period. What it proves is the machine’s own behaviour: rotor balance, vibration amplitude, bearing metal temperatures, seal integrity, lube oil system function.

Performance test

The machine is run against measured conditions and its output recorded — head and flow for a pump, pressure ratio and throughput for a compressor — usually across several points, not one.

String test

The complete train — driver, gearbox, driven machine — run together.

Individual machines can each pass and still misbehave when assembled, because torsional response, lateral behaviour through the coupling and control interaction only exist in the train. A torsional natural frequency sitting near running speed is invisible in any component test.

It is expensive, needs a test facility able to absorb the full power, and is therefore specified for critical unspared trains rather than routinely.

Getting the machine ready on site

Flushing, which is not a formality

The lube system is full of construction debris

When lube oil pipework is fabricated on site it collects welding slag, grinding dust, wire brush bristles, sand and mill scale. All of it is sitting in the pipe when the system is filled.

The first thing a running lube pump does is deliver that straight to the bearing surfaces, at pressure, into a clearance measured in hundredths of a millimetre.

Flushing circulates oil at high velocity — often heated and deliberately turbulent — through temporary loops that bypass the bearings, with progressively finer filters, until the oil reaches a specified cleanliness class and the check patches come back clean.

Bearings wiped in the first weeks of operation are very often the evidence that flushing was cut short because the schedule was tight. See bearings, seals and lubrication.

Alignment, after the pipework

Alignment itself is covered in vibration and alignment. The commissioning point is one of sequence.

A laser shaft alignment system mounted across the coupling of a pump and motor, with brackets clamped to each shaft and a handheld display showing offset and angularity readings. — Align after the pipework is connected. On a machine that runs at 200 °C, a cold alignment is a prediction about thermal growth rather than a measurement.

Machines that run hot grow. A hot alignment allowance offsets the cold setting so the shafts come into line at operating temperature, which means the machine is deliberately misaligned while it is standing still.

Solo run before coupled run

Baselines

The readings taken when a machine is new and known-good are worth recording carefully, because everything afterwards is judged against them.

A large centrifugal pump bolted to a heavy steel test bed in a manufacturer's test hall, connected into temporary looped test piping with flanged spools, gauges on the suction and discharge and instrument cables running to a test console where an engineer is standing. — A works test bed: massive, rigid, and carrying no pipework the machine did not come with. Every condition here is one the site will not reproduce.

Why site problems appear after a clean works test

The works test bed is massive, rigid and free of connected pipework. Site is none of those things.

The foundation and grout may be flexible, or the grout may contain voids
Pipework may be pulling on the casing
Suction conditions are the real ones, so NPSH problems no water test would reveal appear now — see pumps
The process fluid has a different density, viscosity and temperature
The electrical supply may sag on start, so the motor accelerates differently

None of these are failures of the machine, and all of them get reported as one.

What to take away

Mechanical run, performance and string tests answer different questions. Passing one says nothing about the others.
A performance test needs the curve, not the duty point, because the plant will not sit at duty.
Flush the lube system properly. Fabrication debris goes straight to the bearings otherwise.
Align after the pipework, watch the indicators as bolts are tightened, and allow for thermal growth.
Run the driver uncoupled first. Reverse rotation destroys machines in seconds.
Record baselines at commissioning. They can be captured only once, and condition monitoring depends on them.
A clean works test does not predict site behaviour, because the foundation, pipework and fluid are all different.