Understanding Baseline in Vibration Analysis
Baseline — also called baseline data or the reference signature — is the first set of vibration measurements recorded when a machine is new, freshly commissioned, or otherwise in a known-good condition. It is the yardstick against which every later reading is judged, and it is what lets a condition-monitoring programme tell the difference between “running normally” and “starting to fail.” A good baseline captures overall levels, frequency spectra, time waveforms and phase at every measurement point and direction — in short, the fingerprint of a healthy machine.
Accurate baseline data is the foundation of effective predictive maintenance. Without it, trending has no reference point, and you are left guessing whether today’s reading is normal for that machine or the early sign of trouble. The closely related concept of baseline data covers the same idea from the data-management side.
1. Why Baseline Data Matters
A baseline earns its keep in four distinct ways:
- It enables change detection. Current readings are compared to the baseline; deviations flag developing problems, small departures are caught early before they turn severe, and the gap quantifies how far the machine has drifted (for example, a percentage rise from baseline).
- It establishes normal operating characteristics. It documents what “good” looks like for this specific machine, accounting for designs that are inherently rougher than others, setting realistic expectations, and drawing a clear line between normal and abnormal.
- It anchors alarm limits. Alarm levels are often set as multiples of baseline (2×, 3×, 4×), which makes them machine-specific rather than generic, more sensitive to that unit’s own changes, and less prone to false alarms.
- It makes trending meaningful. Plotting current data against baseline over time shows the rate of change, predicts when intervention will be needed, and validates whether a repair actually worked.
2. When to Establish a Baseline
Ideal times
- New-equipment commissioning: after installation, alignment and initial run-in — the best moment of all.
- After a major overhaul: following a rebuild, rewind or bearing replacement.
- After balansiranje: once vibration has been brought down to an acceptable level.
- After a known-good condition is verified: when the machine has been confirmed to operate correctly.
Acceptable times
- Programme start-up: when condition monitoring begins, use the current state provided the machine is functional.
- After minor maintenance: routine work that does not touch major components.
- Fleet baseline: an average across several identical units in good condition.
Poor times (avoid if possible)
- When the machine already has a known problem.
- During abnormal operating conditions.
- When the trend is already climbing.
- Immediately after start-up, before thermal stabilisation.
3. What to Include in a Baseline
Vibration parameters
- Overall levels: RMS velocity, peak, or acceleration at each point.
- Frequency spectra: the FFT showing all frequency components.
- Time waveforms: the raw vibration signal versus time.
- Phase: phase angles at the dominant frequencies — particularly the running-speed (1×) component.
- Multiple directions: horizontal, vertical and axial at each bearing.
Operating conditions
- Speed: the actual RPM during measurement.
- Load: operating load or output.
- Temperature: bearing and process temperatures.
- Pressure/flow: process parameters for pumps, fans and compressors.
- Environmental: ambient temperature and humidity where relevant.
Equipment information
- Equipment ID, location and description.
- Date of the baseline measurement.
- Measurement locations and sensor types.
- Instrument settings (frequency range, resolution, averaging).
- Any special notes or observations.
The reason to record speed and load so carefully is that vibration depends on both. A baseline taken at 80% load is not comparable with a reading at full load, so the conditions must be ones you can reproduce.
4. Baseline Data Quality
Measurement conditions
- Thermal equilibrium: the machine at full operating temperature.
- Steady state: stable conditions, not a transient.
- Representative: the normal operating point, not start-up or shutdown.
- Repeatable: conditions that can be duplicated in future.
Data quality
- Multiple measurements: take three to five, then average or confirm they agree.
- Adequate resolution: enough spectral lines to resolve the important components.
- Full frequency range: capture everything relevant, from low frequency up past 10 kHz where bearing defects live.
- Low noise: a clean signal-to-noise ratio, which in practice means a well-mounted accelerometer.
5. Using the Baseline for Comparison
Numerical comparison. Calculate the percent change as [(Current − Baseline) / Baseline] × 100. Typical alarm criteria sit at +50%, +100% and +200%, with different thresholds for different parameters. This simple ratio is the backbone of most trend analysis.
Spectral comparison. Overlay the current spectrum on the baseline spectrum and look for new peaks (new faults), amplitude growth in existing peaks, and any shifted components. This is where the diagnostic value of a stored spectrum — rather than a single overall number — really shows.
Waveform comparison. Compare the shapes of the time waveforms to detect changes in periodicity, the onset of impacting, or clipping. It is more subjective, but it reveals changes in character that an overall number hides.
6. Updating and Maintaining the Baseline
When to update
- After major repairs: a new baseline following overhaul, rebalance or alignment.
- Equipment modifications: any change to the machine’s configuration.
- Permanent operating-condition changes: a lasting change in speed, load or process.
- Improved condition: after a successful vibration reduction.
When NOT to update
- After vibration has risen — you would erase the very trending history that warns of failure.
- During abnormal conditions.
- After minor maintenance that does not affect vibration character.
- Simply because time has passed; a baseline is meant to be a stable reference.
Version control
- Archive old baselines rather than overwriting them.
- Document the reason for every baseline change.
- Date and identify each version.
- Keep the full historical record.
7. Fleet and Generic Baselines
For sites running several identical machines, a fleet baseline — averaged from a number of units in good condition — represents a typical healthy signature and is useful for new units or after a repair, though individual baselines should still be built up over time. Where no machine-specific data exists at all, generic industry baselines drawn from standards such as ISO 20816-1 (the modern successor to ISO 10816) or from experience give typical levels by machine type. They are less specific but better than nothing — and they connect naturally to formal vibration severity zones.
8. Common Mistakes and Best Practice
The recurring errors are easy to name: running monitoring with no baseline at all; capturing a poor-quality baseline during abnormal conditions or with sloppy technique; relying on a single measurement without checking repeatability; inadequate documentation of conditions and settings; setting a baseline while a fault is already present; and updating too often, which throws away trending history.
Best practice is the mirror image. When establishing a baseline, take comprehensive measurements at all points and directions, repeat them to confirm repeatability, document the conditions fully, store spectra and waveforms (not just overall levels), and photograph the measurement locations so they are reoccupied identically next time. When managing baselines, keep a centralised database, enforce version control and change notes, review and validate periodically, archive historical versions, and train staff on why the baseline matters.
In the field, capturing that first reference is a natural part of commissioning. After a rotor is balanced and aligned, engineers use a portable two-channel instrument such as the Balanset-1A to record the overall level, 1× amplitude and phase, spectrum and waveform at each bearing — the clean, post-correction snapshot that becomes the machine’s baseline and the anchor for every future comparison. Once the reference exists, an Overall Vibration Level Calculator helps convert later spectra into a single comparable figure for trending.
Baseline data is, in the end, the cornerstone of vibration monitoring. Capturing high-quality measurements while the machine is healthy, documenting them thoroughly, and protecting their integrity while updating only when genuinely warranted is what makes meaningful trending and early fault detection possible — and that is what keeps machines running and maintenance well timed.
