The Bearing Failure That Was Visible Six Months Before the Trip

A screw compressor tripped on high discharge temperature at 3:47 AM on a Tuesday. The maintenance log recorded the event as a bearing failure - unexpected, sudden, catastrophic. But the vibration data from the previous six months tells a different story. The first detectable bearing defect frequency appeared 187 days before the trip, at an amplitude of 0.08 inches per second, buried in the velocity spectrum around 1,200 Hertz. By month three, that amplitude had doubled. By month five, sideband frequencies emerged around the running speed, indicating modulation from a progressing inner race defect. The trip was not unexpected. It was the final data point in a pattern that had been writing itself for half a year, and no one was reading it.

A Healthy Screw Compressor Vibration Signature Across Operating Speed

A screw compressor in good mechanical condition produces a vibration signature dominated by the fundamental rotor speeds and their harmonics. For a typical twin-screw machine operating at 3,000 RPM, the primary vibration energy concentrates at 50 Hertz - the male rotor speed - and at multiples corresponding to the lobe pass frequency, which depends on the number of lobes on each rotor. A 4-lobe male rotor and 6-lobe female rotor, for example, generates a lobe pass frequency of 200 Hertz. The overall vibration level on a healthy machine rarely exceeds 0.15 inches per second in the velocity spectrum, and the temperature at the discharge port stabilizes within 10 degrees Celsius of the baseline established during commissioning.

What matters for monitoring is not the absolute value but the stability of that signature. A healthy compressor produces a vibration pattern that repeats across every operating cycle, regardless of load variation. The amplitude at each frequency component stays within a narrow band, typically plus or minus 0.02 inches per second, as long as the rotors remain balanced, the bearings remain intact, and the oil condition remains within specification. Any deviation from this stable pattern is the earliest indicator of mechanical change - long before temperature or pressure readings move.

Bearing Defect Frequencies Appear in Vibration Spectra at Predictable Locations

Rolling element bearings in screw compressors generate defect frequencies that are mathematically related to the bearing geometry and shaft speed. An inner race defect produces vibration at a frequency equal to half the number of rolling elements times the shaft speed, plus the shaft speed itself. For a typical cylindrical roller bearing on the male rotor, this frequency falls between 250 and 350 Hertz depending on the specific bearing dimensions. An outer race defect appears at a lower frequency, typically 150 to 220 Hertz, because the rolling elements pass the outer race defect at a rate determined by the cage speed rather than the shaft speed.

The amplitude at which these frequencies become operationally significant depends on the machine and the measurement location. In practice, a defect frequency that rises above 0.2 inches per second in the velocity spectrum is cause for investigation. Above 0.4 inches per second, the defect is progressing. Above 0.6 inches per second, failure within weeks is probable. These thresholds are not theoretical - they come from hundreds of compressor teardowns where the vibration data was correlated with the actual bearing condition. A plant that collects vibration data monthly at best may catch the amplitude at 0.6 inches per second and schedule a shutdown. A plant that collects data continuously sees the pattern at 0.08 inches per second and has six months to plan the replacement.

Data Change Rate Distinguishes Early Bearing Wear from Incipient Failure

The progression from first detectable anomaly to failure follows a characteristic curve that is nonlinear. In the first three months, the defect frequency amplitude may increase by 0.02 to 0.04 inches per second per month - a slow, almost imperceptible rise that looks like normal variation on weekly readings. In the fourth month, the rate of change accelerates. The amplitude may increase by 0.1 inches per second in a single month. By the fifth month, the increase becomes exponential, with the amplitude doubling every two to three weeks. This acceleration is the critical diagnostic signal, not the absolute amplitude at any single point.

A plant that relies on monthly vibration rounds will see three data points in the slow region and one data point in the acceleration region. The operator may interpret the fourth reading as a sudden jump and conclude the failure was abrupt. In reality, the jump was the visible portion of a curve that had been developing for months. Continuous monitoring captures the entire curve, including the inflection point where the rate of change shifts from linear to exponential. That inflection point, typically occurring between 0.2 and 0.3 inches per second, is the optimal window for planning a bearing replacement - early enough to avoid failure, late enough to confirm the diagnosis.

Temperature Rise Confirms Bearing Deterioration but Arrives Too Late as a Primary Signal

Discharge temperature in a screw compressor is influenced by multiple factors: suction conditions, oil flow rate, oil cooler performance, ambient temperature, and internal clearances. A bearing defect adds frictional heat, but that heat is mixed with the heat of compression and diluted by the oil flow. In practice, the discharge temperature does not show a measurable increase until the bearing defect amplitude exceeds 0.5 inches per second, at which point the mechanical damage is already severe. The temperature rise that appears in the final two to three weeks before failure is typically 5 to 8 degrees Celsius above the normal operating range - a clear signal, but one that leaves almost no time for planned intervention.

Temperature data must be interpreted alongside vibration data, not in isolation. A 3-degree rise in discharge temperature with no corresponding change in suction conditions, oil temperature, or ambient temperature is suspicious. But without vibration data, the operator cannot determine whether the cause is a bearing defect, an oil degradation issue, or a cooling system problem. The combination of a rising defect frequency amplitude and a slow temperature drift is diagnostic. The vibration data tells the operator what is failing. The temperature data tells the operator how much time remains. By themselves, neither signal is sufficient for reliable prediction.

Discharge Pressure Stability Degrades as Bearing Clearance Increases

As bearing wear progresses, the rotor clearances increase. The male and female rotors shift from their designed positions, altering the internal leakage paths within the compression chamber. The result is a gradual degradation of discharge pressure stability. A healthy compressor maintains discharge pressure within plus or minus 0.2 bar of the setpoint under steady load. In the weeks before a bearing failure, that stability window widens to plus or minus 0.5 bar, and pressure fluctuations become visible on the trend chart as a widening band rather than a single line.

This pressure instability is often misattributed to control valve problems or suction pressure variations. Operators adjust the pressure control setpoint, check the suction strainer, and recalibrate the pressure transmitter - all without finding the root cause. The pressure data alone cannot distinguish a bearing-related clearance change from a control system fault. But when the pressure instability is correlated with a rising bearing defect frequency in the vibration spectrum, the diagnosis becomes unambiguous. The pressure data provides the operational symptom. The vibration data provides the mechanical cause. Together, they tell a complete story that neither signal tells alone.

The Operational Difference Between Detection and Prediction

A plant that detects bearing defects at 0.6 inches per second - the amplitude at which most monthly vibration rounds catch the problem - has approximately two to three weeks before failure. In that window, the maintenance team must order bearings, schedule a shutdown, coordinate production downtime, and execute the replacement. If the bearing is not in stock, the lead time alone may exceed the remaining operating life. The plant faces a choice between running to failure and accepting an emergency shutdown, or shutting down early and accepting unplanned production loss. Neither option is acceptable from an operational perspective.

A plant that detects the pattern at 0.08 inches per second has six months. The bearing can be ordered from the original manufacturer at standard lead time. The replacement can be scheduled during a planned maintenance window. The production team can adjust the production schedule to accommodate the downtime. The cost of the replacement is a fraction of what an emergency repair would cost, and the production loss is zero because the downtime was already planned. The difference is not in the quality of the maintenance team or the skill of the operators. The difference is in the data resolution - continuous monitoring versus monthly snapshots - and the ability to see the curve before it bends.