Key takeaways
  • Condition monitoring detects deterioration before failure, so intervention is planned rather than forced.
  • Vibration, thermography, oil analysis and ultrasound cover most common failure modes between them.
  • A single reading is close to worthless. The value is entirely in the trend, which means consistent measurement points and intervals.
What makes a programme work
Consistency, not instrument quality

The same point, the same method, the same interval, recorded and trended. A modest handheld used consistently outperforms an expensive instrument used sporadically, because deterioration shows up as change and change needs a baseline.

Vibration analysis

Detects bearing wear, misalignment, imbalance, looseness and gear defects in rotating equipment. Bearing faults typically appear well before failure, giving a useful planning window. Start with overall vibration levels trended over time; spectral analysis is a later refinement once the trend shows something.

Infrared thermography

Detects electrical connection faults, unbalanced loads, overheating bearings, insulation defects, steam trap failures and refrigeration problems. Electrical panel surveys are usually the highest-return application, because a loose connection heats long before it fails and the failure is often a fire risk.

Oil analysis

  • Wear metals indicate which component is degrading, since different alloys point to different parts.
  • Contamination by water, dirt or fuel indicates ingress or a seal problem.
  • Additive depletion and viscosity change indicate the oil itself is finished.
  • Sampling technique matters more than the laboratory: same point, same conditions, clean bottles, machine at temperature.

Ultrasound

Detects compressed air and gas leaks, steam trap failure, electrical arcing and early bearing faults. Leak surveys are frequently the fastest payback available in a facility, because compressed air leaks are inaudible, continuous and often account for a large share of the compressor's output.

Starting a programme

  • Choose assets from the criticality ranking, not from what is convenient to reach.
  • Define the measurement points physically and mark them, so the same spot is used every time.
  • Set intervals from how quickly the failure mode develops, not from convenience.
  • Record every reading, including the normal ones. Normal readings are what make the abnormal one obvious.
  • Define what happens when a reading exceeds a threshold, including who decides and what action follows.

The most common failure

Readings collected and filed without anyone trending them. If nobody looks at the trend and nobody is accountable for acting on it, the programme is a data collection exercise. Assigning a named reviewer and a monthly review slot costs almost nothing and is what turns readings into avoided failures.