Air Compressor Commissioning Procedure from First Startup to Full-Load Verification

Commissioning gets two days on the Gantt chart. Sometimes one.

Most of what goes wrong with compressors in their first three years traces back to these two days, and most of what goes wrong during these two days traces back to piping strain, which is why this article spends a disproportionate amount of space on piping strain and relatively little on things like safety shutdowns, which are important and which are also adequately covered in every OEM manual ever printed.

Piping Strain

A Kaeser CSD compressor, 160 kW, installed at a food processing plant. Persistent 8% specific power penalty above the CAGI data sheet value. Three service visits. Rotors within tolerance. Oil injection rate correct. Separator differential normal. On the fourth visit, someone unbolted the discharge flange and the pipe moved 3 mm. Correcting the pipe support eliminated the penalty entirely. The British Compressed Air Society published this in a technical bulletin. The fix took an afternoon. The three prior service visits, plus the excess energy consumed over the months between installation and diagnosis, cost considerably more.

Here is what happens mechanically. The discharge pipe comes off a header that was hung before the compressor arrived. The holes almost line up. A fitter uses a chain hoist to close the gap. Bolts in. Hoist off. The casing now carries a sustained bending load through the discharge flange. On screw compressors where the discharge-end bearing housing is part of the main casting (Kaeser CSD/CSDX, Atlas Copco GA, CompAir L series all do this), the bore distorts asymmetrically. Rotor tip clearance on machines in this power range sits between 0.04 and 0.15 mm. A few hundredths of bore distortion tightens one quadrant, opens the opposite one. Tight side: friction, heat. Open side: internal gas leakage from discharge back toward suction.

At the first rebuild, the technician sees uneven wear. Writes "contamination." Oversize rotors go in. Same pipe goes back on with the same strain.

The Kaeser CSD installation manual specifies 0.3 mm maximum pipe-to-flange offset. Checking takes twenty minutes. Unbolt the flange, watch if the pipe moves.

This one item probably matters more than the rest of this article combined.

Foundation Bolts

Torque in a star pattern, then bar the shaft over by hand, two full turns. If binding appears that was not there before torquing, the foundation is not flat. Shim the low foot.

On reciprocating compressors, un-shimmed frame warp loads one main bearing more than its neighbors. That bearing position becomes the site of every future failure. The relationship between frame distortion magnitude and the resulting bearing life reduction is poorly characterized in published literature. The mechanism is understood. Quantifying it for a specific installation is not something anyone has a good method for.

Pre-Lube

On an Ariel frame with ten-plus lube points distributed along a multi-throw crankcase, pre-lube takes 45 seconds. On a packaged screw compressor, a few seconds. The Stribeck curve describes the transition from boundary to hydrodynamic lubrication as shaft speed increases. Before the hydrodynamic wedge develops, the journal rides on whatever boundary film the pre-lube pump established. Without pre-lube, the journal contacts babbitt, hard particles embed in soft metal, and fatigue life shortens by an amount that is certain in direction and unknown in magnitude.

Phase Rotation and Restart Timing

Phase rotation: meter, not bump start. Reverse rotation on a screw element lets pressurized separator oil spin the rotors backward faster than the motor would have driven them forward. Helical rotor scoring within two seconds. Scrapped rotors and often a scrapped housing bore.

Restart inhibit timing after a trip: 30 seconds on small machines, two minutes above 75 kW. The motor manufacturer's thermal capacity curve should set the exact number. When that data is unavailable, two minutes.

First Start

Sixty seconds, unloaded.

The thing to pay attention to is sound. Oil pressure and discharge temperature are on gauges and they are important and they get logged. Sound does not get logged. A compressor with good alignment makes a dull steady hum. Coupling offset produces a rhythmic pulse at shaft speed. A bearing defect produces a tick that may require a screwdriver blade pressed against the housing to hear. These are different problems requiring different corrective actions, and they are distinguishable within seconds of reaching speed, and they are available before any vibration analyzer has been set up and configured and has collected enough data for an FFT.

Oil pressure stabilizes in about ten seconds on a positive-displacement pump. Cold oil: slow steady climb. Air in suction: erratic. Mis-set relief valve: stable but low.

Discharge temperature on oil-flooded screw compressors climbs even unloaded, from friction and gas re-expansion between the rotors. If the rise rate exceeds about 2 deg. C per minute, the thermostatic valve is stuck in bypass. The wax motor inside the valve can lose its charge during months of storage. Passing 100 deg C starts a varnish cycle: oil oxidizes, deposits form on hot surfaces, deposits restrict flow, temperature rises, oxidation accelerates.

Touch the bearing housings after shutdown. Check the sight glass for foam. Foam before any load means the anti-foam chemistry in the oil is wrong for this circuit. Shell Corena S4 R and Mobil Rarus SHC 1025 are both good PAO compressor oils. Different anti-foam packages. A machine specified for one, filled with the other, may foam even though every data sheet property matches.

25%, 50%, 75%, full load. Fifteen to thirty minutes each.

Temperature should track pressure ratio. Current should follow the power curve. When a ratio breaks at a given load step, the break isolates the subsystem.

At 50% load, spend extra time. Slide valve hunting on screw compressors happens here and only here. At full load the valve is at end of travel. At no load, same. At mid-stroke, static friction in the actuator exceeds the controller's minimum output increment, and the valve sticks, breaks free, overshoots, sticks again. Fixing it is PID tuning: widen the proportional band, lengthen the integral time, iterate. How much to adjust depends entirely on the installation. An hour of tuning work. This hour almost never appears in commissioning schedules, not because anyone decided it was unimportant.

Reciprocating compressors with step unloading at intermediate loads have a torsional dynamics concern. Certain cylinder combinations excite torsional natural frequencies. API 618 Section 3.9.2 requires torsional analysis for petroleum-service compressors. The Fifth Edition added forced-response analysis because undamped natural frequency studies were systematically underestimating torsional stresses. Dresser-Rand documented coupling failures on multiple installations in service bulletin DSB-108, traced to torsional resonance at unloading combinations that the original undamped analysis had missed.

Non-petroleum installations may have no torsional analysis at all.

On those machines, monitor coupling surface temperature with an infrared thermometer at each unloading step during commissioning. Listen near the coupling guard. Torsional resonance severe enough to damage a coupling produces audible cyclic noise before failure. Lateral vibration sensors will not detect it.

Cooling

Approach temperature is the measurement that matters. Gap between cooled fluid leaving the aftercooler and cooling medium entering the cooler. Stays roughly constant across ambient conditions.

The arithmetic: approach temperature plus site historical maximum ambient. Compare to the high-temperature trip setpoint. Margin under 10 deg C means summer trips.

The CAGI performance verification handbook identifies insufficient cooling water flow as the leading cause of elevated discharge temperatures on water-cooled installations. Finding it requires an actual heat balance: water flow rate times temperature rise times specific heat, compared to motor input power corrected for efficiency. When the heat balance shows a deficit above about 10%, the water flow is short. The shortage is in the plant header. Undersized pipe, a valve partly closed since construction, a cooling tower not at design capacity.

Based on informal polling among Compressed Air Challenge practitioners, the heat balance calculation gets performed during fewer than one in twenty commissionings. Not for technical reasons. The checklist does not call for it and the crew does not carry flow measurement equipment. Whether this should change is a question that answers itself once anyone calculates the cumulative energy cost of running a compressor with a 10% cooling water deficit for fifteen years.

Safety Shutdowns

Oil pressure trip: differential, not absolute. On a screw compressor at 8 bar separator pressure with 3.5 bar on the differential gauge, absolute supply is 11.5 bar. Bearing protection depends on differential because gas pressure surrounding the bearing is what creates the bearing load. The Ingersoll Rand UP series maintenance manual specifies the trip setpoint explicitly as a differential value. When a third-party controller replaces the OEM unit, this distinction gets lost.

Temperature and pressure shutdown tests: simulate each trip condition, measure response time. For temperature, the lubricant supplier's data sheet specifies maximum continuous and maximum excursion temperatures. For pressure, verify blowdown valve opens after the trip. If blowdown sticks closed and auto-restart is enabled, the motor starts against full separator pressure. Locked-rotor current, six to eight times rated. NEMA MG-1 Section 12.52 gives thermal withstand time for squirrel-cage motors in this condition at 10 to 25 seconds depending on frame size.

E-stops: every button, motor killed, blowdown open, manual reset required.

Full Load, Four Hours

Specific power. kW/m³/min corrected per ISO 1217 Annex C. The CAGI data sheet for the machine lists the tested value. Field measurement should land within 5% after ambient corrections.

A machine at 6.8 instead of 6.0 wastes 13% of input energy. On a 150 kW unit at 8,000 hours per year: 156,000 kWh. A receiver pump-up test per ISO 1217 Annex D.2 measures flow well enough if the receiver volume is known and the pressure transmitter is decent. Thirty minutes of work. The receiver pump-up test is probably the most under-used commissioning technique in the compressed air industry relative to the value of the information it provides.

Vibration baselines: three axes, every bearing, permanently marked measurement points. ISO 10816-3 Table C.1 gives the Zone A boundary for newly commissioned machines at 2.3 mm/s RMS in the relevant speed range. The measured value becomes the trending reference.

Retorque bolts after the run.

The commissioning report needs numbers, instrument serial numbers, calibration dates, operating conditions. "OK" next to a checklist item has no diagnostic value at any future date.

— end —