Using Vibration Analysis for Early Problem Detection

Why Vibration Analysis Works

Rotating machinery vibration is not random. Because the forces that generate vibration are tied to rotational speed and the mechanical structure of the machine, vibration signatures follow predictable frequency patterns. When a fault develops, it typically disturbs one or more of these force relationships in a consistent way, producing a characteristic change in the vibration spectrum.

This means that experienced interpretation of vibration data can identify not just that something is wrong, but what is likely wrong — whether the source is unbalance, misalignment, bearing damage, fluid instability or something else. That specificity is what makes vibration analysis valuable for maintenance planning rather than just fault detection.

The Two Levels of Vibration Data

Overall level monitoring

Overall vibration level is the simplest form of vibration measurement, but the units and meaning depend on the measurement type. Two distinct approaches are used on large rotating machinery, and they are complementary rather than interchangeable:

• Shaft relative vibration — measured by proximity probes (eddy-current transducers) mounted in the bearing housing, reading shaft displacement directly. Expressed in micrometres (µm) peak-to-peak. The governing standard for large rotating machinery is ISO 7919. Shaft measurement is the primary protection signal on most large turbines and generators because it directly captures rotor behaviour, including instabilities that may not be strongly reflected at the casing.
• Casing or bearing housing vibration — measured by seismic transducers (velocity pickups or accelerometers) mounted on the bearing housing or support pedestal. Expressed in mm/s RMS. The governing standard is ISO 10816 / ISO 20816.

A developing shaft instability can produce large shaft displacement while casing vibration remains relatively low — which is precisely why shaft measurement by proximity probe is the primary protection channel on large machines. Where both measurement types are available, both should be monitored and trended together. The "overall level" discussed in the remainder of this section refers to the combined trend picture from whichever measurement type is primary for the machine in question.

Overall level monitoring is effective for trend detection: a machine that is gradually deteriorating will typically show a rising overall vibration level over time. The limitation is that overall level provides no information about the source of the vibration. Two machines at the same overall level can have very different fault conditions.

Frequency spectrum analysis

Frequency spectrum analysis (spectral analysis or FFT analysis) decomposes the vibration signal into its component frequencies. This reveals which frequencies are contributing to the total level and in what proportion. Because different fault types produce different frequency signatures, spectrum analysis makes it possible to identify the likely cause of elevated vibration, not just its presence.

Spectrum analysis requires more sophisticated instrumentation and interpretation skill than overall level monitoring, but it provides substantially more diagnostic value.

Key Frequency Patterns and What They Indicate

1× running speed (synchronous vibration)

A dominant component at 1× running speed (1× RPM) typically indicates unbalance, misalignment or shaft bow. This is the most common vibration component in any rotating machine. A small 1× component is normal and expected. A large or growing 1× component warrants investigation.

• Unbalance: produces 1× in the radial direction (horizontal and vertical). The unbalance force (F = m·e·ω²) is proportional to the square of angular velocity — doubling speed quadruples the excitation force. However, the resulting vibration amplitude response is not simply proportional to speed. It depends on the rotor's position relative to its critical speed and on the system's damping characteristics. Below the critical speed (stiffness-controlled region), amplitude increases with speed, rising more sharply as the critical speed is approached. Above the critical speed — where most large turbines and generators operate at rated conditions — the rotor is in the mass-controlled region and amplitude changes relatively little with small speed variations. Near a critical speed, response is amplified by the system's damping ratio and can be substantially higher than the off-resonance response at the same force level. A rotor showing strongly increasing 1× amplitude with small speed changes should therefore be evaluated in the context of its proximity to a critical speed, not treated as a simple linear scaling of excitation force.
• Shaft bow: produces 1× synchronous vibration that is indistinguishable from dynamic unbalance by amplitude or phase angle at running speed alone. The standard field diagnostic differentiator is the slow roll vector: at very low speed (well below the critical speed — typically below 10% of rated speed), the dynamic unbalance force is negligible. If 1× vibration persists at slow roll speed with consistent amplitude and phase, it indicates mechanical bow (shaft runout) rather than dynamic unbalance. Thermal bow is a distinct condition: the rotor deforms due to differential temperature across its cross-section during standstill, producing a temporary eccentricity that manifests as elevated 1× vibration during the run-up. The thermal bow signature — whether vibration initially rises or falls as the rotor warms up — depends on the type and location of the bow condition and does not follow a single predictable pattern.
• Misalignment: can produce 1× but more characteristically produces 2× or a combination of 1× and 2×, often with a different phase relationship in the axial direction.

2× running speed

A dominant 2× component is a well-established indicator of misalignment. Angular misalignment typically produces dominant 2× vibration in the axial direction, often with a 180° phase difference across the coupling. Parallel (offset) misalignment typically produces elevated 1× and 2× in the radial direction. This distinction is diagnostically useful: high axial 2× with low radial 2× points toward angular misalignment; high radial 2× with elevated 1× points toward parallel offset. A growing 2× component can also indicate mechanical looseness or, in advanced stages, a developing transverse shaft crack. A breathing crack creates cyclic stiffness asymmetry at twice per revolution, which produces 2× vibration; however, this signature only becomes detectable at moderate-to-advanced crack depths — typically when the crack has grown through a significant fraction of the shaft cross-section — and is not a reliable early warning indicator. Early-stage shaft crack detection by vibration analysis alone is unreliable. Reliable crack assessment requires specialist transient analysis, comparing run-up and coast-down vibration behaviour against established baseline data to identify changes in critical speed, 1× and 2× amplitude, and directional response characteristics. Any suspicion of shaft cracking warrants OEM consultation and specialist rotordynamic assessment — it should never be inferred from a 2× level change in isolation. If 2× grows while 1× remains stable, misalignment or looseness is the more probable cause.

Sub-synchronous vibration (below 1×)

Vibration at frequencies below running speed is often the most diagnostically significant. Common sub-synchronous components and their causes:

• 0.43–0.48× (oil whirl): fluid film instability in plain or partial-arc journal bearings. Oil whirl occurs because the rotating oil film in a full-circumferential bearing travels at approximately half the shaft rotational speed, producing a destabilising pressure wave that rotates ahead of the shaft in the same direction. It is associated with lightly loaded bearings, excessive bearing clearances, high oil temperature, and low oil viscosity. Classic oil whirl appears at 0.43–0.48× running speed and can remain at a relatively stable amplitude or grow progressively. An important design qualification applies to the large turbine and generator context: tilting pad journal bearings (TPJB) — which are standard on virtually all modern large steam turbines and generators — are specifically engineered to suppress this mechanism. Because each pad floats independently and adjusts its position in response to shaft loading, the continuous circumferential pressure wave required to sustain oil whirl cannot develop. As a result, classic oil whirl is rarely observed on TPJB-equipped machines under normal operating conditions. If sub-synchronous vibration in the 0.43–0.48× range is observed on a machine equipped with tilting pad bearings, this is an unusual finding that warrants immediate investigation. Possible explanations include severely abnormal operating conditions, bearing damage, incorrect bearing assembly, or a sub-synchronous excitation source other than classical oil whirl — such as aerodynamic excitation in the steam path or a structural resonance.
• Locked at a frequency (oil whip): if the oil whirl frequency (≈0.47× running speed) coincides with a rotor natural frequency, the vibration locks onto that natural frequency even as running speed continues to increase — this is oil whip. Since the whirl frequency increases proportionally with speed until it reaches the first rotor natural frequency, oil whip typically first becomes possible at approximately twice the first critical speed (because 0.47× of 2N_c ≈ N_c). Oil whip is a self-sustaining instability that does not self-limit under the same mechanism that bounds oil whirl — the destabilising force grows with response amplitude rather than diminishing. Amplitudes can increase rapidly to levels that cause severe bearing damage, seal rubs and rotor deflections.
• 0.3–0.5× with load correlation: in some cases, sub-synchronous vibration correlates with load changes rather than speed — this can indicate aerodynamic or steam path excitation forces, particularly at off-design load conditions.

Super-synchronous vibration (above 1×)

Vibration at integer multiples of running speed (3×, 4×, higher harmonics) typically indicates mechanical looseness — pedestal looseness, bearing looseness or structural looseness in the supporting structure. The presence of many harmonics (a "forest of peaks" in the spectrum) is a strong indicator of looseness. A separate and distinct category of high-frequency excitation — blade passing frequency (BPF) — also occurs in turbine and compressor stages. BPF is calculated as the number of rotor blades multiplied by the rotational frequency in Hz. For a typical steam turbine stage with 40–80 blades running at 3000 RPM (50 Hz machines), BPF falls in the range of 2000–4000 Hz; on 60 Hz machines at 3600 RPM, the equivalent range is 2400–4800 Hz — in either case, well above the low-order harmonic region discussed here. Blade passing frequencies therefore occupy a fundamentally different part of the frequency spectrum, require wide-bandwidth instrumentation to measure, and are assessed through separate analysis methods. They should not be confused with the low-order integer harmonics used for looseness and alignment diagnostics.

Bearing defect frequencies

Rolling element bearings (used in some auxiliary systems and smaller machines) produce specific defect frequencies related to the geometry of the bearing — inner race defect frequency (BPFI), outer race defect frequency (BPFO), ball spin frequency (BSF) and cage frequency (FTF). These frequencies are calculable from bearing geometry and running speed. Detection of these frequencies provides early warning of localised bearing damage before it progresses to general failure.

Journal bearings (used in most large turbines and generators) do not produce bearing defect frequencies. Their condition is assessed through overall level, sub-synchronous vibration patterns, temperature monitoring and oil condition.

Trend Interpretation: Rate of Change Matters as Much as Level

Absolute vibration level tells you the current state. Rate of change tells you the urgency. A machine that has operated at 4 mm/s for three years and is now at 4.2 mm/s is not in the same situation as a machine that was at 2 mm/s six months ago and is now at 4 mm/s. The second machine requires investigation even if its current level is within alarm limits.

Two useful rules of thumb for trend interpretation:

• A vibration level that doubles over a short period is more significant than the absolute level suggests. The relevant timeframe varies substantially with fault type: gradual deposit unbalance or slow-developing misalignment may progress over weeks or months; mechanical loosening may progress over days; fluid film instabilities can double within hours or less. The rate of change, and whether that rate is itself accelerating, is more diagnostic than the absolute level.
• Any sudden step change — vibration that moves significantly in minutes or hours rather than days or weeks — always warrants investigation, even if the new level is still within normal operating range. Sudden step changes indicate an abrupt change in machine condition — a component that shifted, loosened or failed — and carry a different and typically more urgent set of possible causes than gradual trend increases.

Vibration During Startup: A Diagnostic Window

Machine startup is a particularly rich diagnostic window because the rotor passes through the entire speed range from zero to operating speed. This reveals information that steady-state operation cannot:

• Critical speed location: the speed at which peak vibration occurs during run-up identifies the rotor's critical speed. If this shifts between outages, it warrants investigation: a genuine shift can indicate a change in shaft stiffness or mass distribution caused by damage or deposit accumulation. Note that apparent critical speed during a run-up is also influenced by oil film bearing stiffness, which varies with oil temperature and run-up rate — comparisons between outages should therefore be made at consistent operating conditions and interpreted in the context of any changes to the startup procedure.
• Thermal bow: thermal bow develops during machine standstill when differential temperature across the rotor cross-section causes the shaft to deflect toward the hotter side, producing a mechanical eccentricity that generates elevated 1× vibration during the run-up. The vibration signature during startup depends on the type and location of the bow condition: some present as high initial vibration that decreases progressively as the rotor reaches thermal equilibrium; others produce vibration that increases initially before stabilising; some affect only specific speed ranges as the critical speed is traversed. Whether thermal bow vibration at any point during a given startup is acceptable must be assessed against the OEM startup procedure, not characterised as generally acceptable or otherwise. Most OEM startup procedures define the maximum permissible vibration at each speed hold point, the expected time window within which thermal bow effects should diminish to acceptable levels, and the action required if vibration does not clear within that window. Proceeding through an OEM-defined speed hold point with vibration above the defined limit — regardless of the attributed cause — carries risk of blade tip rubs, labyrinth seal contact, and bearing edge loading. Thermal bow startup vibration should never be assessed or accepted without direct reference to the applicable OEM startup procedure and its defined limits.
• Phase stability: recording vibration phase angle during startup provides diagnostic information that amplitude alone cannot give. When comparing successive startups, a change in phase angle at a fixed speed point — with amplitude unchanged — can indicate a change in rotor condition that warrants investigation. The 180° phase shift that occurs as the rotor accelerates through a critical speed is a confirming characteristic of a genuine resonance passage; its absence, or an abnormal phase trajectory through the critical speed zone, can indicate a change in rotor damping or stiffness near that speed.

Practical Limits of Vibration Analysis

Vibration analysis is powerful but not unlimited. Several things it cannot reliably do without additional information:

• Vibration cannot determine the severity of bearing surface damage that is not yet generating increased vibration — a bearing with significant metallurgical damage can still produce low vibration if the running clearance has not yet changed significantly.
• Vibration alone cannot determine whether the cause of a detected fault is primary or secondary — a looseness condition identified by vibration may be the primary fault or may be a consequence of another developing problem.
• Vibration analysis requires consistent measurement conditions. Changes in speed, load, oil temperature or adjacent machine behaviour all affect the vibration signature. Comparison between measurements made under different conditions can be misleading.

These limitations mean that vibration analysis is most valuable when used in combination with other condition monitoring signals (bearing temperature trends, oil condition analysis) and with direct inspection findings during outages. A vibration diagnosis that is consistent with oil analysis findings and visual inspection observations is much more reliable than one based on vibration data alone.