Five Key Methods for Bearing Fault Diagnosis

Overview

This article reviews five common methods for bearing fault analysis and practical diagnostic tips.

1. Abnormal rotational sound analysis

Auditory inspection is used to monitor bearing condition by listening. Common tools include a long-handled screwdriver or a rigid plastic tube about 20 mm in diameter. Electronic stethoscopes improve monitoring reliability. A healthy bearing runs smoothly and lightly without stalling, producing a harmonious, low-noise sound such as a steady "whooshing" or a low-frequency "hum." Abnormal sounds indicate specific bearing faults as follows.

• Hissing: A uniform, continuous hissing arises from rolling elements rotating on the inner and outer raceways and includes irregular metal vibration sounds independent of shaft speed. This typically indicates insufficient grease; replenish lubricant. If equipment has been idle for a long time, especially in low-temperature conditions, a hissing or grating sound can occur due to reduced radial clearance and reduced grease consistency; adjust bearing clearance and replace with grease of slightly higher penetration.
• Periodic "clunk" within continuous whooshing: A regular periodic noise superimposed on the continuous whooshing indicates marks, grooves, or corrosion on the rolling elements or raceways. The noise period is proportional to speed. Replace the bearing.
• Irregular scraping: Irregular, uneven scraping sounds are caused by foreign particles such as iron filings or sand entering the bearing. The sound intensity is relatively low and not related to speed. Clean the bearing and re-lubricate or replace the lubricant.
• Coarse, continuous rubbing: Continuous, irregular rubbing sounds often result from too loose a fit between the inner ring and shaft or the outer ring and housing bore. If sound intensity is high, inspect and repair the fit.

2. Vibration signal analysis

Bearing defects such as spalling, indentations, corrosion, cracks, and wear produce distinctive vibration responses. Using specialized vibration measurement instruments (frequency analyzers, etc.) allows measurement of vibration amplitude and frequency content to infer specific faults. Measured values depend on operating conditions and sensor mounting, so baseline measurements for each machine and sensor position are necessary to establish diagnostic thresholds. Typical bearing diagnostic techniques include vibration signal detection, lubricant analysis, temperature monitoring, and acoustic emission detection. Vibration-based diagnostics are the most widely used and are divided into simple and precision methods.

2.1 Simple diagnosis

Simple vibration diagnostics compare measured vibration values (peak, mean, rms, etc.) with preset thresholds to determine whether a fault exists and whether further precision analysis is required. Judgment criteria can be grouped into three types:

• Absolute criteria: Fixed limit values for measured vibration.
• Relative criteria: Periodic vibration measurements at the same location compared over time; use the vibration when the bearing is known to be healthy as the reference and diagnose based on the ratio of current to baseline values.
• Comparative criteria: Compare vibration values among several identical bearings under the same conditions at the same location.

No single absolute standard applies to all bearings, so a combination of absolute, relative, and comparative criteria is recommended for reliable diagnosis.

Common simple diagnostic methods include:

1. Amplitude-based diagnosis
   Amplitude metrics include peak XP, mean X (for harmonic vibration this is the half-cycle average; for impact-type bearing vibration the absolute-value average), and root mean square Xrms. This is the simplest and most common method: compare measured amplitudes with standard limits. Peak values are useful for faults with transient impacts such as surface pitting. Mean value is similar to peak but more stable and typically used at higher speeds (e.g., above 300 rpm). RMS is a time-averaged metric suitable for faults like wear where amplitude changes slowly over time.
2. Probability density diagnosis
   The amplitude probability density of a healthy rolling bearing typically follows a normal distribution. Faults can cause skew or dispersion in the density curve.
3. Kurtosis diagnosis
   A healthy bearing with normally distributed amplitudes has a kurtosis around 3. As faults develop, kurtosis changes in a manner similar to crest factor. Kurtosis is independent of speed, size, and load and is especially useful for detecting pitting-type defects.
4. Waveform factor diagnosis
   Waveform factor is defined as peak divided by mean (XP/X). It is an effective indicator for simple bearing diagnosis.
5. Crest factor diagnosis
   Crest factor is defined as peak divided by RMS (XP/Xrms). Its advantages are that it is not affected by bearing size, speed, or load, nor by sensitivity changes in sensors or amplifiers. It is suitable for diagnosing pitting-type faults. Monitoring the trend of XP/Xrms over time enables early warning and reflects fault progression:
   • Healthy bearings have a small, stable XP/Xrms.
   • When damage begins, impact signals increase peak values while RMS remains relatively unchanged, so XP/Xrms increases.
   • As the fault expands and RMS increases, XP/Xrms decreases toward its healthy value.

2.2 Precision diagnosis

Bearing vibration contains rich frequency components across low and high bands, and specific faults correspond to specific frequency signatures. Precision diagnosis uses signal processing to extract these characteristic frequency components.

• Low-frequency analysis
  Low-frequency refers to vibration below 8 kHz. Acceleration sensors are commonly used for bearing vibration, but low-frequency analysis often uses velocity. Acceleration signals are converted to velocity via integration after charge amplification, then passed through a low-pass filter with an upper cutoff around 8 kHz to remove high-frequency content before frequency analysis.
• Mid- and high-frequency demodulation analysis
  Mid-frequency ranges from 8 to 20 kHz and high-frequency from 20 to 80 kHz. For these bands, acceleration can be analyzed directly. After charge amplification, signals are passed through a high-pass filter to remove low-frequency components, then demodulated and subjected to frequency analysis to identify characteristic frequencies.

3. Temperature analysis

Bearing temperature can usually be inferred from the outside of the bearing housing; measuring the outer ring temperature through an oil hole is preferable when possible. Bearing temperature typically rises gradually after startup and stabilizes within one to two hours. Normal bearing temperature depends on machine thermal capacity, dissipation, speed, and load. Improper lubrication or installation can cause rapid temperature rise and abnormal overheating, requiring shutdown and corrective action.

High temperature often indicates abnormal conditions and also degrades lubricants. Prolonged operation above 125°C will reduce bearing life. Causes of overheating include inadequate or excessive lubrication, lubricant contamination, excessive loading, bearing damage, insufficient clearance, and high friction from seals.

Continuous temperature monitoring is therefore advisable for bearings and other critical components. Under steady operating conditions, any temperature change indicates a potential fault. Regular measurements can be taken with thermometers, for example a digital handheld thermometer that displays in °C or °F. For critical bearings whose failure would cause a shutdown, consider installing temperature detectors. Note that bearings show a natural temperature rise after initial lubrication or relubrication that may persist for a day or two.

4. Lubricant analysis

Lubricant analysis uses ferrography to identify and predict rolling fatigue. A lubricant sample is drawn and passed through a high-gradient magnetic field, causing particulate contaminants to deposit on a glass slide by size. Examining particle shape, size, color, and composition reveals wear types and helps detect incipient faults.

Ferrography primarily identifies ferrous materials but can also detect nonferrous metals, sand, organic matter, and seal fragments. Detection of spherical ferrous particles 1–5 μm in diameter indicates the initiation of fatigue microcracks. Fatigue spall particles with a length-to-thickness ratio around 10:1 and length greater than 10 μm indicate abnormal fatigue wear; lengths exceeding 100 μm indicate bearing failure. Thin fatigue flakes with length-to-thickness ratios around 30:1 and lengths of 20–50 μm, often with voids, increase noticeably at fatigue onset and, together with spherical particles, serve as markers of fatigue development.

5. Acoustic emission testing

Acoustic emission (AE) refers to the release of strain energy as elastic waves when a material deforms or a crack propagates. AE testing detects and analyzes these signals to infer source events in the structure. The technique identifies elastic waves generated by relative motion of internal particles to evaluate internal condition.

AE signals are classified as transient (discrete) or continuous. Transient AE signals are pulses distinguishable from background noise; continuous AE consists of many pulses that are too dense to separate. In practice, continuous AE is formed by many small transient events that are not individually resolvable.

Both transient and continuous AE can occur in malfunctioning rolling bearings. Transient AE arises from localized events such as contact impacts, surface cracks, wear, indentations, scoring, seizing, lubricant contamination particles creating hard edges, and electrical pitting from current through the bearing. Continuous AE mainly results from global issues such as lubricant film breakdown, contamination in grease causing oxidative wear, high temperature, and multiple local faults producing a large number of transient AE events in a short time.

Faults in rolling bearings, whether surface damage, cracks, or wear, produce elastic impact events at contact surfaces that emit AE signals rich in rubbing and impact information. Therefore, AE monitoring can be used to detect and diagnose rolling bearing faults.