The Bearing That Failed Between Preventive Maintenance Visits
A customer had a conveyor drive bearing that was scheduled for replacement every 6 months (preventive maintenance). It failed at month 4. The machine was down for 8 hours while they replaced it. If they’d been monitoring the bearing, they would have seen it was degrading weeks before the failure. The vibration levels had been creeping up since month 2 — a sign of rolling element damage. But nobody was measuring it. The PM schedule was calendar-based, not condition-based. The failure wasn’t a surprise to the bearing — it was a surprise to the maintenance team because they weren’t listening.
Machine vibration monitoring and predictive maintenance is about listening to the equipment before it fails. Instead of replacing parts on a schedule (which is either too early or too late), you monitor the vibration, temperature, and current, and you replace the part when it actually needs it. This article is how I add condition monitoring to machines so failures don’t happen unexpectedly.
Preventive vs. Predictive Maintenance
Preventive Maintenance (Time-Based)
Replace parts on a schedule (every 3 months, every 2,000 hours). Simple to plan, but two problems:
- Too early: The part still has life left. You’re throwing away good components. Wasted cost and labor.
- Too late: The part fails before the next PM interval. Unplanned downtime, which is far more expensive than scheduled maintenance.
Predictive Maintenance (Condition-Based)
Monitor the component’s condition. Replace it when the data shows it’s degrading. You replace it just before it fails — not too early, not too late.
| Approach | Trigger | Cost | Downtime |
|---|---|---|---|
| Preventive (calendar) | Time / hours | Higher (parts wasted) | Planned, but failures happen between |
| Predictive (condition) | Vibration / temp / current | Lower (replace only when needed) | Planned, no surprise failures |
| Run-to-failure | It broke | Low upfront, high failure cost | Unplanned, expensive |
What Vibration Tells You
A healthy bearing, gear, or motor vibrates in a characteristic pattern. As the component degrades, the vibration changes. The vibration signature tells you what’s wrong.
Common Faults and Their Signatures
- Bearing defect (rolling elements): High-frequency vibration spikes at specific frequencies (the ball pass frequency, the cage frequency). A worn bearing shows increasing high-frequency vibration.
- Misalignment: 1× RPM (the shaft rotates once per revolution, and the misalignment causes a vibration cycle at that frequency).
- Unbalance: 1× RPM, but high amplitude. A heavy spot on the rotor causes vibration once per revolution.
- Wear (looseness): Multiple harmonics (2×, 3× RPM). Loose components vibrate at multiple frequencies.
- Gear damage: Sidebands around the gear mesh frequency. A chipped tooth shows a spike at the mesh frequency modulated by the shaft RPM.
You don’t need to diagnose every fault. You need a baseline (the healthy vibration level) and an alarm threshold. When the vibration exceeds the threshold, the component is degrading. Schedule a closer look.
Sensors: What to Install
Condition monitoring starts with sensors. You don’t need expensive systems for every component — focus on the critical ones.
Vibration Sensors (Accelerometers)
An accelerometer measures vibration (acceleration, velocity, or displacement). Mount it on the bearing housing (magnetic mount or stud). For rotating equipment (motors, pumps, gearboxes), this is the standard sensor.
Two approaches:
- Portance (periodic): A technician walks around with a vibration meter every month. Reads the levels. Simple, cheap, but not continuous.
- Permanently mounted (continuous): Sensors wired to a monitoring system. Data is collected 24/7. Trends are tracked. More expensive, but catches developing faults in real time.
Temperature Sensors
A rising temperature is a lagging indicator (the component is already hot). But it’s simple and cheap. A thermocouple or RTD on the bearing housing or motor winding alarms at a setpoint. Good as a backup to vibration.
Current Monitoring (Motor Amps)
A motor’s current draw changes as the load increases. A pump that’s losing efficiency draws more current. A bearing that’s seizing draws more amps. Monitoring motor current (via a current transformer on the supply) is cheap and catches mechanical problems indirectly.
| Sensor | What It Detects | Cost | Best For |
|---|---|---|---|
| Accelerometer | Bearing, misalignment, unbalance, gears | Medium | Rotating equipment (motors, pumps, gearboxes) |
| Temperature | Overheating (lagging) | Low | Bearings, motor windings |
| Motor current | Load increase, mechanical binding | Low | Pumps, fans, conveyors |
Where to Monitor: Critical Assets Only
Not every component needs monitoring. Focus on the ones where failure causes unplanned downtime.
Monitor These
- Main drive motors: A failed motor stops the machine.
- Gearboxes: Expensive to replace, slow to source.
- Main bearings: Conveyor drives, spindle bearings.
- Pumps (hydraulic, coolant): A failed pump overheats the system.
Don’t Bother With These
- Small, cheap cylinders and sensors: If they fail, replace them. Monitoring costs more than the part.
- Components with spare redundancy: If a standby pump is ready, the main pump can fail without downtime.
- Components that are already PM’d: If a bearing is replaced every 6 months anyway, vibration monitoring may not add much (unless it’s failing between PMs).
The 80/20 rule: 20% of the components cause 80% of the unplanned downtime. Put sensors on those 20%. Don’t instrument every bearing in the machine. The monitoring system that covers every sensor is expensive, and the data becomes noise. Monitor the critical few.
Setting Alarm Thresholds
A sensor reading is meaningless without a threshold. Set them based on baseline and standards.
Baseline Measurement
When the machine is new and healthy, measure the vibration. This is the baseline. Compare future readings to this baseline. A 2× increase from baseline is the first warning. A 4–5× increase is critical — schedule maintenance now.
Standards (ISO 10816)
The ISO 10816 standard gives vibration velocity limits based on machine size and type. For a general industrial motor (medium size), the alarm is around 4.5 mm/s RMS. Above that, investigate. Use the standard as a starting point, then adjust based on the baseline.
Data Integration: Where Does the Alarm Go?
The sensor data should reach the maintenance team, not just sit in a PLC.
- Machine HMI: The operator sees a “bearing A vibration high” alarm. They flag it for maintenance.
- Maintenance system (CMMS): The alarm creates a work order automatically. The maintenance team gets a notification.
- Trending software: The vibration data is logged over time. You can see the trend (is it slowly increasing, or spiking suddenly?). A slow increase means gradual wear. A sudden spike means an acute failure.
The ROI: What You Save
The cost of a monitoring system (sensors + wiring + software) is small compared to one unplanned downtime event.
- Unplanned downtime: A bearing failure on a critical line costs $5,000–50,000 per hour (lost production, expedited parts, overtime).
- Predictive monitoring: The system alerts you 1–4 weeks before failure. You schedule the replacement during planned downtime (a weekend, a changeover).
- Extended part life: You replace the bearing when it’s actually worn, not at 50% life. The parts cost goes down.
For a machine that runs 24/7, a monitoring system pays for itself in one avoided unplanned event. For a single-shift machine, the ROI is lower — the downtime cost is less urgent.
A Condition Monitoring Checklist
- Which components are critical? (Top 20% that cause downtime.)
- What sensors fit each? (Vibration, temperature, current?)
- Has a baseline been measured (healthy machine)?
- What are the alarm thresholds? (Baseline × 2, ISO 10816?)
- Does the alarm reach maintenance (HMI, CMMS, work order)?
- Is the data trended over time? (Not just current reading?)
- Who responds to the alarm? (Maintenance on call?)
- How are the sensors mounted? (Stud, magnetic, permanent?)
- Is there a periodic walk-down (portable vibration meter)?
- Has the ROI been calculated? (Cost of one unplanned event vs. monitoring system?)
The Bottom Line
Predictive maintenance and condition monitoring isn’t a fancy IIoT project. It’s putting a vibration sensor on the critical bearing, measuring the baseline, and alarming when it degrades. The bearing that failed between PM visits wasn’t unpredictable — it was telling you (through rising vibration) for weeks. Listen. Put sensors on the 20% of components that cause 80% of the downtime. Set thresholds based on baseline. Route alarms to maintenance. The machine that never has a surprise failure isn’t lucky. It’s being monitored.
