Rolling-element bearings are the single most common point of failure in rotating machinery. They're also among the most detectable — if you know what to look for. The five failure modes that account for the majority of bearing failures in discrete manufacturing all produce measurable precursor signals before they reach the point of catastrophic damage. Catching those signals early is what separates a planned bearing replacement from a six-hour production stop.
Failure Mode 1: Fatigue Spalling
Fatigue spalling is the most common bearing failure mode in well-maintained machinery operating within its rated load range. It results from cyclic contact stress between rolling elements and raceways over the bearing's service life. Subsurface cracks initiate below the Hertzian contact zone, propagate to the surface, and eventually result in material removal — spalling — that creates pitting on the raceway surface.
The early vibration signature is subtle. Envelope analysis and high-frequency energy (HFE) measurements show elevated energy in the 2–20 kHz range before a clean spectral peak appears at the bearing defect frequency (BPFO or BPFI). This is the stage where detection matters most — the bearing still has useful life remaining, and a planned replacement can be scheduled at the next maintenance window.
As spalling progresses, a distinct peak appears at BPFO (outer race) or BPFI (inner race) and its harmonics. A kurtosis value above 6 in the high-frequency band is a reliable indicator of developing spall damage. A kurtosis above 10 indicates advanced spalling — schedule replacement within 1–2 weeks maximum. At kurtosis above 15, the risk of rapid progression to catastrophic failure is elevated; the asset should be checked at each shift start.
Detection approach: envelope analysis (demodulation) at 2–20 kHz band, combined with time-domain kurtosis trending. BPFO and BPFI frequencies are available from any bearing catalog using the bearing number.
Failure Mode 2: Contamination
Contamination ranks among the most frequent causes of premature bearing failure in manufacturing environments where airborne particles, machining chips, coolant, or water can enter the bearing housing. Abrasive particles create surface scratching on raceways and rolling elements; corrosive contaminants cause pitting and rust. Both accelerate the fatigue mechanism described above, significantly reducing actual bearing life versus the rated L10 life.
The vibration signature of contamination-damaged bearings is diffuse and wideband — elevated overall vibration without the clean defect-frequency peaks you'd see in classic fatigue spalling. This makes contamination harder to detect via spectral analysis alone and why temperature monitoring becomes especially important as a complementary channel. Contamination-damaged bearings typically run 5–15°C above their normal operating temperature due to increased friction and reduced lubricant film thickness.
Detection approach: monitor overall vibration velocity in the 10–1000 Hz band for broadband elevation, and track bearing housing temperature relative to baseline. A sustained temperature rise of more than 10°C above the established baseline warrants inspection regardless of the vibration spectral picture. In our work, contamination failure modes are detected an average of 3–6 weeks earlier when temperature data is added alongside vibration than with vibration alone.
Failure Mode 3: Lubrication Breakdown
Lubrication failure — insufficient grease quantity, degraded lubricant, or wrong lubricant specification — removes the protective film between rolling elements and raceways, causing metal-to-metal contact. The resulting surface damage looks similar to contamination-induced failure but typically progresses faster because the entire contact zone is affected simultaneously, not just areas touched by individual particles.
The earliest signature of lubrication breakdown is acoustic emission (AE) in the ultrasonic range — typically 20–100 kHz. Standard vibration sensors at 10 kHz won't catch this. Ultrasonic monitoring is the appropriate first-line tool for detecting lubrication deficiency before mechanical damage has occurred. If you don't have continuous ultrasonic monitoring, a handheld ultrasonic probe used during route-based rounds — listening for the characteristic dry, crackling sound of inadequate lubrication — can catch this mode early.
Once mechanical damage has begun, the signature transitions to wideband vibration elevation similar to contamination, with increasing overall velocity and rising temperature. By this stage, the bearing has already been damaged; early intervention avoids catastrophic failure but the bearing should be replaced at the next convenient opportunity.
Detection approach: ultrasonic monitoring as first line; temperature trending as second line; overall vibration velocity as confirmation. Also review lubrication intervals — if you're finding dry bearings on assets that were recently greased according to schedule, the interval may need adjustment for the actual operating load and speed.
Failure Mode 4: Misalignment
Bearing failure from misalignment is actually a consequence of a drivetrain geometry problem, not a bearing material problem. When a shaft is angularly or radially misaligned, the rolling elements track on an eccentric path relative to the designed contact geometry. This produces an uneven load distribution around the raceway — overloading one side while underloading the other — and causes the bearing to wear asymmetrically and fatigue prematurely.
The vibration signature of misalignment shows up at the shaft rotation frequency and its harmonics — primarily at 2X for angular misalignment, with elevated axial vibration relative to radial. The bearing itself may show accelerated BPFO or BPFI progression, but the 1X/2X pattern in the overall spectrum is the diagnostic clue that the root cause is geometric rather than a bearing material issue.
Important: replacing a bearing that has failed due to misalignment, without correcting the misalignment, produces the same failure 1–3 months later. The bearing replacement is symptomatic treatment, not a fix. A laser alignment check takes 30–60 minutes and should always follow a bearing replacement on a direct-coupled drive where misalignment is suspected.
Detection approach: look for 2X elevation in the radial spectrum, elevated axial readings relative to radial, and asymmetric bearing loading visible in infrared thermography (one side of the bearing housing runs hotter). Correct the misalignment and verify vibration returns to baseline before closing the work order.
Failure Mode 5: Overloading
Overloading occurs when the bearing carries loads beyond its rated capacity for extended periods. Common causes in discrete manufacturing include process changes that increased line speed or output requirements without a corresponding engineering review, worn tooling that increases process forces, or incorrect bearing selection during a replacement (wrong load rating, wrong size, wrong clearance class).
Overloaded bearings show characteristic fatigue patterns — spalling concentrated at the maximum load zone rather than distributed around the raceway — and typically fail faster than expected given their nominal operating parameters. The vibration signature is progressive BPFO or BPFI energy increase similar to classic fatigue, but the degradation rate is accelerated. A bearing that would normally show three to six months of progressive spectral changes before reaching critical threshold may go from first anomaly detection to failure in four to eight weeks under overload conditions.
Detection approach: standard bearing defect frequency monitoring catches overloading-induced fatigue, but the faster degradation rate requires higher monitoring frequency. If you detect early bearing defect signatures on a recently replaced bearing on the same asset where the last bearing also failed prematurely, don't just order another replacement — request a load analysis. Check whether process parameters have changed, whether the bearing specification matches the current operating conditions, and whether adjacent components (coupling, shaft, housing bore) are within specification.
Building a Detection Program Around These Five Modes
Each failure mode has a preferred detection method. A table of mode-to-method mapping makes program design straightforward:
| Failure Mode | Primary Detection | Secondary Detection | Typical Lead Time |
|---|---|---|---|
| Fatigue spalling | Envelope analysis / BPFO / BPFI | Kurtosis trending | 4–12 weeks |
| Contamination | Wideband velocity + temperature | HFE monitoring | 2–6 weeks |
| Lubrication breakdown | Ultrasonic (AE) | Temperature rise | Days to weeks |
| Misalignment | 1X/2X spectrum + axial ratio | Infrared thermography | Immediate on survey |
| Overloading | Accelerated BPFO/BPFI progression | Load analysis / bearing spec check | 4–8 weeks (faster than fatigue) |
Gearcadence's ML models are trained across all five of these failure modes. The platform classifies the failure mode alongside the TTF window, so maintenance teams don't just know that something is wrong — they know which failure mechanism is active and what the appropriate corrective action is. For bearings specifically, classification accuracy across the first three modes — fatigue, contamination, and lubrication — consistently reaches 82–88% at the early-stage detection window in our validation data. That's meaningful because the corrective action differs: a fatigue-spalling bearing needs replacement, a contamination case needs a root cause investigation, and a lubrication case may only need a re-lubrication event to extend bearing life significantly.
Five failure modes. Each with distinct signals. All detectable well before catastrophic failure. The investment in understanding the differences pays back directly in reduced parts costs, avoided production stops, and maintenance work that gets done on schedule instead of on emergency callout.
