Air Compressor Surging and Pulsation: Diagnosis for Centrifugal and Screw Units

How much does it shift? There is no universal answer. Elliott and MAN Turbo machines in certain refinery wet gas applications have had field-calibrated surge lines shifted four to six percent to the right compared to the factory test report. The more the gas molecular weight deviates from the test gas and the more the discharge system volume and resistance characteristics deviate from the test stand conditions, the larger the shift.

This means the surge control line set in the anti-surge controller (most installations use CCC’s S1 or S2 series) has less margin than the engineer thinks. Conducting a field surge test during commissioning is the only rigorous way to resolve this. The method is to activate the protection system and then gradually reduce flow or increase discharge pressure, pushing the operating point step by step toward the boundary, using high-speed data acquisition to capture the dynamic pressure signal of the first surge event, and then calibrate. Not many installations have done it, because risk management approval is frequently not obtainable, especially on older in-service machines.

Understanding surge as the compressor itself convulsing will mislead the diagnostic direction. The discharge piping plus buffer tank forms a Helmholtz resonance cavity, and the compressor is the energy input end of this resonance system. When the energy the compressor inputs can no longer sustain the downstream backpressure, the gas decelerates, reverses, discharge pressure drops, forward flow re-establishes, discharge pressure recovers, and the cycle repeats. The frequency is determined by the acoustic parameters of the discharge system and has nothing to do with compressor speed.

The same machine, connected to a compact system on the discharge side, can produce surge at three to four hertz, with each reversal making an audible bang. Connected to a long pipe network plus a large receiver, the frequency drops to fractions of a hertz, showing up on the DCS discharge pressure trend as a slow oscillation, and the operator staring at the screen may not realize something is wrong.

Low-frequency surge has a longer duration per reversal event. Longer duration means the thrust bearing endures reversed axial loading for a longer time. Under forward loading the oil film is established. When reversed loading is suddenly applied, the oil film must re-establish on the opposing thrust face, which takes time. During this transition, if the load already exceeds the transient load-carrying capacity of the oil film, the babbitt metal contacts the thrust collar directly. During an overhaul, if the babbitt surface shows discoloration turning purple with signs of peeling, this is characteristic of prolonged overheating, pointing to low-frequency surge. If the surface shows scoring and localized spalling, that looks more like oil film breakdown caused by high-frequency impact.

When subsynchronous vibration and noise changes appear in the low-flow region of a centrifugal compressor, the first thing to do is determine whether it is rotating stall or surge. Install a dynamic pressure transducer on the discharge piping, vary the speed, and watch the pressure oscillation frequency response. If the frequency changes proportionally with speed, it is rotating stall. If it does not track with speed, it is surge.

Among centrifugal compressors covered by API 617, designs with vaneless diffusers are more prone to rotating stall at low flow than those with vaned diffusers. In a vaneless diffuser, when the gas flow angle deviates from the design value, there are no vanes to correct the flow direction, and separation occurs more easily. Some machines just run with mild stall in a certain range below design flow, with vibration levels slightly elevated but within the API 670 alarm limits. If this condition is misidentified as incipient surge, the control logic may intervene unnecessarily, introducing new process disturbances.

The operating point ended up on the left side of the surge line. That is the only direct cause. Sudden reduction in downstream demand is the most common trigger scenario. An upstream process unit trips, a downstream user switches over, a batch ends, and flow drops. The anti-surge controller should open the recycle valve to maintain minimum compressor throughflow.

Gradual increase in downstream resistance causing discharge pressure to slowly climb. The time scale for this type of event is days or even weeks. Discharge-side heat exchanger fouling, separator liquid level running high, a manual valve inadvertently left at half-open position. The diagnostic clue is to plot the historical trend of discharge pressure against flow and see if there is a slow upward drift.

Compressor performance degradation: impeller fouling, labyrinth seal clearance increase, diffuser damage. These cause reduced head-generating capability at the same flow, equivalent to the surge line shifting rightward on the map. Impeller fouling needs separate discussion. The deposit layer is unevenly distributed. Sometimes the surge line does not shift uniformly but bulges outward in a specific speed range. A machine with comfortable surge margin at 85% speed might have no margin left at 78% speed. Gas composition changes also alter the gas molecular weight, which affects the relationship between polytropic head and volume flow, causing the operating point to drift.

Anti-surge valve sticking can be found through valve stroke testing. There is a problem that cannot be found without dedicated testing: the valve position has arrived, but the flow has not. The controller receives valve position feedback, calculates from the valve Cv curve that “17% open should provide this much recycle flow,” and considers the margin sufficient. The problem is that the Cv curve has changed and nobody knows. After two or three years in this service, Fisher or Masoneilan valve trims develop seal face wear and deposits in the small-opening range that cause the Cv to deviate significantly from the factory calibration curve. The 10% to 20% opening range is precisely the most frequently used working range in anti-surge scenarios.

The verification method: at a stable operating condition, manually force the valve to a known opening, observe how much compressor flow increases, and compare with the theoretical value. If the deviation exceeds 30%, the valve’s performance in surge protection can no longer be trusted. Resistance changes in the recycle piping are another blind spot. The recycle line has a silencer (whose internal packing compacts over time), a recycle cooler (fouling, increasing pressure drop), and a check valve (whose disc may stick at a half-open position). These pressure drop changes are gradual and have no dedicated monitoring points.

When surge occurs, gas begins to reverse from the discharge side back toward the compressor. The discharge check valve should close to isolate the compressor from the downstream high-pressure pipe network. A conventional swing check valve relies on the dynamic pressure of the reverse flow to push the disc. When the reverse flow first begins, the velocity is low, the disc wobbles at a partially open position. By the time the reverse flow velocity is high enough, the disc slams shut, and the resulting water hammer pulse is superimposed on the pressure fluctuation from the surge itself.

Spring-loaded non-slam check valves from Goodwin or Neles have a spring preload on the disc so that when the forward flow velocity drops to near zero, the disc has already begun its closing motion without needing reverse flow to push it. The closing process is smooth, and the water hammer pulse is much smaller. This matter is rarely included in considerations during the compressor selection phase. The discharge check valve specification is typically determined by piping design, without the piping engineer having discussed with the rotating equipment engineer what dynamic characteristics the check valve at this location needs.

Screw compressors do not surge; this premise must be established first. API 619 is the standard for screw compressors, and the word surge does not appear in it. The working principle of positive displacement machinery means that as long as the rotors are turning, gas is forced forward.

Screw compressor pulsation originates from the discrete nature of the compression process. Each revolution of the male rotor produces several discharge events equal to the number of male rotor lobes, and each discharge event is a pressure pulse. A machine with a 4-lobe male rotor at 3000 rpm has a fundamental frequency of 4 times 50 equals 200 Hz, plus harmonics. Atlas Copco’s ZR series, Kobelco’s KNW series, and GHH Rand machines all operate on this basic principle; the differences lie in rotor profiles, Vi design, and noise reduction measures.

Every screw compressor has a built-in volume ratio Vi determined by the rotor profile and discharge port geometry. Vi multiplied by the pressure at suction closure, then converted through an isentropic process, yields the pressure that the gas in the tooth space should have at the moment the discharge port opens. If this pressure equals the pressure in the discharge piping, the gas flows out smoothly.

Under-compression: the system discharge pressure is higher than the pressure inside the tooth space. The instant the discharge port opens, high-pressure gas from the piping rushes into the tooth space, producing a steep pressure-rise pulse superimposed on the normal lobe-passing pulsation. Efficiency loss comes from two directions: recompressing the backflow gas consumes power, and the turbulence induced by the backflow increases internal friction losses. Discharge temperature runs high.

Over-compression: the system discharge pressure is lower than the pressure inside the tooth space. The gas has been compressed beyond what is needed, and when the discharge port opens it expands outward. Extra compression work was performed for nothing.

This energy waste will not trigger any alarm. Specific power is simply not a tracked parameter in the monitoring systems of the vast majority of plants. Operators watch discharge pressure, discharge temperature, and motor current, all within normal ranges. Six screw air compressors each rated at 160 kW, assuming an average efficiency loss of six percentage points, 7200 operating hours per year, local industrial electricity price 0.085 euros per kWh: one machine wastes an extra 160 times 0.06 times 7200 times 0.085, about 5875 euros per year. Six machines comes to roughly thirty-five thousand. This number does not appear in the maintenance budget because nothing is broken, nor in fault records because there are no alarms. It just flows into the total figure on the electricity bill.

Fixed-Vi machines have no way to self-adapt when system pressure fluctuates. Variable-Vi slide valve machines can theoretically track pressure changes. The complication: many machine models share a single slide valve for both capacity regulation and Vi adjustment. When the controller faces two conflicting objectives, discharge pressure control takes priority over Vi optimization, and Vi gets sacrificed. Some manufacturers use a dual slide valve design that decouples these two control degrees of freedom. It costs more.

Discharge piping has its own acoustic natural frequencies, determined by the pipe geometry and the speed of sound in the gas inside. If one of the pipe’s natural frequencies coincides with the lobe passing frequency or one of its harmonics, that frequency component of the pulsation is amplified by resonance, and radiated noise increases dramatically.

Speed of sound changes with temperature. As ambient temperature goes from 5°C in winter to 35°C in summer, the speed of sound in air goes from 334 m/s to 352 m/s, a change of over 5%. Pipe natural frequencies shift accordingly. A pipe natural frequency that in winter misses the lobe passing frequency by 8 Hz might land right on it in summer. This is why some screw compressor stations have noise complaints with a clear seasonal pattern. Maintenance personnel change gaskets, tighten flanges, check bearings, find nothing, and in two days the noise goes away on its own. The temperature dropped and the resonance condition was no longer satisfied. The countermeasure: install a side branch resonator on the discharge piping, a length of closed-end branch pipe that provides quarter-wavelength resonant absorption at the target frequency.

API 619 requires screw compressor performance testing per ASME PTC 9 or ISO 1217. Performing periodic performance tests and comparing current specific power to the commissioning baseline is the most direct means of tracking rotor condition degradation. Discharge temperature as a standalone indicator is unreliable, simultaneously influenced by suction temperature, suction pressure, discharge pressure, and cooling.

The parameter with diagnostic value is discharge temperature deviation: at the current actual operating conditions, calculate the theoretical discharge temperature using the isentropic compression formula, then add the known isentropic efficiency offset for that machine, yielding what the discharge temperature should be with the machine in new condition. Measured value minus this theoretical value is the deviation. If the deviation is increasing month over month, and cooling system issues have been ruled out, it points to increased internal leakage.

The impact signature when a screw compressor ingests liquid looks completely different from normal pulsation on a time-domain waveform. Normal pulsation is a periodic signal with good repeatability. Liquid slugging appears as randomly occurring isolated spikes, manifesting on the frequency spectrum as broadband energy elevation rather than discrete harmonics.

Warning signals are faint. Discharge temperature during a liquid slug event exhibits a rapid fluctuation pattern of first dropping then rising, lasting a short time, requiring at least one-second sampling intervals to see. Most plants’ temperature acquisition cycles are ten or thirty seconds, too slow to capture it. The reality is that most screw compressor liquid slugging events are only discovered after visible damage has occurred. The focus of prevention is on source control: whether the suction piping liquid separator design is adequate, whether drain traps are functioning, and whether operating conditions might produce condensation inside the suction piping.

The configuration of two or more centrifugal compressors operating in parallel sharing a common discharge header is very common. Machine A approaches the surge boundary, and the controller opens Machine A’s recycle valve. Machine A’s effective discharge flow drops. If total downstream demand has not changed, the entire shortfall is pushed onto Machine B. If Machine B is already running at high load, it cannot absorb the entire shortfall. Header pressure rises, Machine B’s operating point climbs toward the surge line. Machine B’s controller also begins opening its recycle valve.

Now both machines have their recycle valves open, neither is contributing effective flow to the header, downstream pressure drops, both recycle valves close back, both machines reload, pressure rises again, both approach the boundary again. When the amplitude is large enough, one or both machines surge and trip. The trip record reads “Machine A surge trip at 14:37, Machine B surge trip at 14:38.” Neither machine shows a single-machine problem. The problem is in the coupling. The key is to treat parallel operation as a system when designing anti-surge protection, rather than protecting N independent machines separately.

Surge is an aerodynamic phenomenon. Its primary signal is oscillation of pressure and flow, not vibration. Vibration is a secondary signal resulting from pressure oscillations acting on the rotor and casing, having passed through a conversion from aerodynamic to structural mechanics, with information loss. A dynamic pressure transducer installed near the discharge flange, sampling at 2 kHz or above, yields a pressure waveform that can directly answer: is there surge or incipient surge, what is the frequency, what is the amplitude, is it surge or rotating stall.

API 670 does not list dynamic pressure measurement as a standard configuration. Most installations have a full set of vibration monitoring with no permanently installed dynamic pressure transducers on the discharge piping. The value of vibration data in surge matters lies after the fact rather than during the event. After a surge trip, check whether the axial displacement trend shows a step change, check whether the vibration spectrum shows new components, and decide whether a case-opening inspection is needed. Conventional process pressure transmitters typically sample one per second or lower and have no resolving power for the dynamic characteristics of surge.

There is one thing that involves no hardware modification whatsoever, purely DCS screen development. Project each centrifugal compressor’s current operating point in real time onto the compressor map for the operator to see, mark its distance from the surge line, color-coded.

What operators normally see is the individual numbers for discharge pressure, suction pressure, flow, and temperature, each with its own alarm limit. When all numbers are within limits, the screen is all green. The operating point may be only 3% margin from the surge line, and no single parameter will reach its alarm because the alarms are set on absolute values, not on relative position on the compressor map. Real-time operating point visualization closes this gap. When operating conditions push the point toward the surge line, the color changes from green to yellow to red, and the operator can intervene manually before the anti-surge controller kicks in. Implementing this requires digitized OEM performance curve data, a calculation module in the DCS to perform suction condition correction, and screen development. The workload is not large.