Condition Monitoring in Bearing Maintenance

For most of the industry's history, bearing maintenance ran on the calendar: relubricate every so many hours, inspect every so many months, replace at the first sign of trouble. Condition monitoring changed the input. Instead of asking how long a bearing has run, it asks how it is running right now, and the answer decides the action.

This guide covers the four measurements that carry almost all the value - vibration, temperature, lubricant condition and acoustic emission - what each one detects earliest, how to calculate the bearing defect frequencies from geometry and speed, how to read the ISO 10816 severity zones as a decision rather than a number, and the routine that avoids both missed failures and needless strip-downs.

From calendar-based to condition-based maintenance

The calendar method is not wrong so much as blind. It replaces a bearing that has years left, or leaves one running that is weeks from seizure, because it cannot see the difference. Condition monitoring replaces the clock with three questions: is there a defect, where is it, and how fast is it growing?

The commercial case is straightforward. A monitored bearing can often be changed on a planned stop, while an unmonitored one changes the plant's schedule instead. The technical case is that most bearing failures announce themselves long before they become functional failures, and the announcement is measurable. ISO 15243 classifies the failure modes; monitoring is how you catch them in the early stage rather than the terminal one.

The four measurements that matter

Each technique has a different earliest warning, and using only one of them means missing the failures that the others see first.

What each technique sees first

The table below is the one we use when a plant asks where to start. Read it from left to right: the technique you spend money on should be the one whose earliest warning arrives in time for you to act.

The order matters. A raceway defect shows in the vibration envelope first and in the temperature last, which is why reacting to temperature alone means reacting late. Our grease lubrication intervals guide explains why lubricant state is a leading indicator rather than a lagging one.

Defect frequencies you can calculate

A damaged raceway generates a repeating impact each time a rolling element passes over it. The repetition rate depends only on the geometry and the running speed, so it can be calculated in advance and used to look for a specific fault rather than a generic rise.

The geometry behind BPFO and BPFI

Write the number of balls as n, the ball diameter as d, the pitch diameter as D, the contact angle as a, and the shaft speed in hertz as f. The four characteristic frequencies are then simple products.

A worked example

Take a deep groove ball bearing with 8 balls, a ball diameter of 10 mm, a pitch diameter of 60 mm and a contact angle of zero, turning at 1,500 rpm, which is 25 Hz. The ratio d/D is 10/60, or 0.1667. BPFO then comes to 83.3 Hz, BPFI to 116.7 Hz, FTF to 10.4 Hz and BSF to 72.9 Hz. The value of the calculation is that these are exact targets: when a peak appears at 83.3 Hz with sidebands at the shaft speed, you are looking at an outer raceway defect rather than a gear fault or an unbalance. Our defect frequency reference tabulates the formulas for other arrangements.

Vibration severity zones as a decision table

ISO 10816, continued as ISO 20816, defines vibration severity by machine class and by zone. The zones are not an alarm to be argued with; they are a decision about what to do next.

The zones are defined on overall velocity in mm/s rms for a given machine class, and the class is set by the machine's power and mounting rather than by the bearing. Read the standard for the exact boundaries, because a rigid-mounted 15 kW motor and a flexible-mounted 300 kW fan do not share them.

Trending versus alarm limits, and portable versus permanent

An alarm limit is a single number; a trend is a direction. Both are needed, and the trend is usually the more valuable of the two. A machine sitting at the top of zone B for two years is healthy; the same machine rising through zone B in three months is not, even though the number is identical today.

Portable route or permanent sensor

The hardware follows the same logic. A portable route visits a fixed list of points on a schedule and gives broad coverage at low cost, which suits a plant with many similar machines. Permanent sensors on a critical machine give continuous data and catch the fast-developing failure that a monthly route would miss, which suits an unspared machine that stops the line. Many plants run both: portable routes for the population, permanent sensors for the critical few.

Avoiding the classic false alarms

Most bad maintenance decisions come from a measurement that was real but misread, so check the known confounders before you act on any spectrum.

• Sensor mounting. A probe held by hand or a magnet on a rough casting gives a different reading every time. Mount it properly and consistently.
• Resonance. A structural resonance can amplify a harmless frequency into a frightening amplitude; check the machine's natural frequencies first.
• Loose foundation. Looseness looks like bearing damage in the spectrum until you find the soft foot.
• Variable speed drives. A drive changes speed and load continuously, so a fixed defect frequency is wrong and the trend is noisy. Use speed-synchronous tracking.

What to do when a defect frequency appears

A defect frequency is a hypothesis, not a diagnosis. Before committing to a strip-down, verify it against the two causes that masquerade as bearing damage. Sample and analyse the grease or oil for water and wear metals, because a lubrication failure produces the same frequency as a raceway spall. Measure the internal clearance if the machine allows it, because a loss of clearance from heat can copy a defect signature. Then review alignment and fit, since misalignment and a creeping outer ring both generate harmonics that mimic a raceway fault. If the trend is confirmed and rising, plan the replacement on a stop rather than waiting for the failure, because the remaining life under ISO 281 is no longer the calculated value.

Step-by-step condition-monitoring programme

1. Inventory the machines. List every rotating machine with its speed, power and criticality.
2. Classify by criticality. Decide which machines get permanent sensors and which get a portable route.
3. Establish a baseline. Record vibration, temperature and lubricant state while the machine is healthy.
4. Calculate the defect frequencies. Work out BPFO, BPFI, BSF and FTF for each bearing from its geometry and speed.
5. Set the alarm strategy. Use the ISO 10816 zones for severity and a rate of change for the trend.
6. Run the route. Take readings on schedule and at a consistent mounting point.
7. Analyse the exceptions. Investigate any point outside its baseline or its zone, not every point.
8. Confirm with a second measurement. Check lubrication and clearance before deciding on a strip-down.
9. Act and record. Replace or repair, then update the baseline so the next cycle starts from truth.

Common mistakes

• Reacting to temperature alone. It is a late, indirect indicator and it will let a raceway defect reach the terminal stage.
• Ignoring the grease. Lubricant condition is often the earliest signal and the cheapest to act on.
• Replacing the bearing without fixing the alignment. The new bearing inherits the cause and fails on the same schedule.
• Trusting a single alarm limit. Without a trend, a limit cannot tell a stable machine from a failing one.

Monitoring is only as good as the action behind it. The failure modes it is meant to catch are catalogued in our failure analysis primer, the same discipline applied to a specific drive train is in the motor bearing field guide, and the environment that decides which failures dominate is set out in the industry pages. If you are building a programme for a specific machine, send the duty and the arrangement; if you are choosing the bearings as well, the deep groove range and the self-aligning roller range are the two families that carry most of the planet's monitored positions, and both are cross-referenced in the interchange lookup against brands such as SKF.