What is condition monitoring?

The mechanics

Three maintenance strategies exist, and condition monitoring is the input to the third.

• Run to failure. Cheapest until it isn't. Rational for a $400 fan with a spare on the shelf; irrational for the only compressor feeding a production line.
• Time-based. Rebuild every 8,000 hours whether the machine needs it or not. This replaces healthy components and, because every intervention introduces installation error, occasionally causes the failure it was meant to prevent.
• Condition-based. Measure something that correlates with deterioration, trend it, and act when it crosses a threshold.

For rotating equipment — pumps, fans, motors, compressors, gearboxes — the primary measurement is vibration. A machine in good condition has a characteristic vibration signature. Imbalance, misalignment, looseness, bearing wear and gear defects each change that signature in a characteristic way.

ISO 20816: the zone framework

ISO 20816 (which supersedes and consolidates ISO 10816 and ISO 7919) evaluates machine vibration using broadband velocity RMS measured across 10–1,000 Hz, and it classifies the result into four zones:

Zone	Meaning
A	Newly commissioned machines
B	Acceptable for unrestricted long-term operation
C	Unsatisfactory for long-term operation; acceptable for a limited period while a repair is arranged
D	Sufficient to cause damage

The critical detail that gets skipped: the numerical boundaries depend on the machine's power group and on how it is mounted. ISO 20816-3 separates large machines (above 300 kW) from medium ones (15–300 kW), and rigid mounting from flexible mounting, and the thresholds differ substantially between them. Applying one number across a mixed asset population produces false alarms on the machines that are fine and silence on the ones that are not.

Indicative boundaries for Group 2 machines (15–300 kW) on rigid foundations: Zone A/B at 1.4 mm/s, B/C at 2.8 mm/s, C/D at 4.5 mm/s. For Group 1 (above 300 kW) on rigid foundations the same boundaries sit at 2.3, 4.5 and 7.1 mm/s. Flexible mounting raises all of them. Always apply the table for the actual machine.

Worked example

A 75 kW centrifugal pump on a rigid foundation — Group 2.

• Commissioning baseline: 1.9 mm/s velocity RMS → Zone B, acceptable indefinitely
• Six months later: 2.9 mm/s → just into Zone C. Not an emergency; a trend worth watching and a reason to check alignment at the next shutdown
• Two months after that: 4.8 mm/s → Zone D. This machine is doing damage to itself now, and the repair should be scheduled rather than awaited

The value is not in the absolute number. It is in the rate of change: 1.9 to 4.8 mm/s in eight months on a machine that had been stable for three years is the finding. The zone table tells you when to stop discussing it.

The honest boundary: what overall RMS can and cannot do

This is where most of the category over-claims, and being straight about it is worth more than the claim.

What broadband velocity RMS gives you: severity. It tells you a machine has already moved from acceptable into unsatisfactory or damaging. For a plant with no monitoring at all on most of its rotating equipment — which is the normal condition — that is a real and valuable ring of coverage, cheap to install and cheap to run.

What it does not give you: early warning of a bearing defect. Incipient bearing damage produces very low-energy, high-frequency impacts that are buried in the broadband signal. Detecting them requires time-waveform capture at 25,600 samples per second or higher with envelope (demodulation) analysis, plus the bearing's geometry and the machine's running speed at the moment of acquisition — and on a variable-frequency-drive machine, running speed changes constantly.

A vendor promising "months of advance notice" from an overall-RMS sensor is describing a capability that measurement class does not have.

On motor current signature analysis: MCSA infers mechanical faults from the motor's current spectrum and is genuinely useful for rotor bar and eccentricity faults. It requires the motor at 40% load or higher, is masked by the switching harmonics of a VFD, and is less sensitive to bearing faults than vibration is. Where MCSA is the right tool, it should be used, not approximated.

What condition monitoring is not

It is not predictive maintenance. Condition monitoring is the measurement. Predictive maintenance is a programme built on it: thresholds, a named owner, a work-order route, a parts strategy and a feedback loop that records whether the prediction was right. Buying sensors and calling it predictive maintenance is the most common way these programmes fail.

A severity trender is not a diagnostic tool. It says the machine is in Zone C. It does not say whether that is imbalance, misalignment, looseness or a bearing, and those have different repairs with different costs.

More data is not more warning. Continuous high-rate waveform capture on every asset multiplies cost and storage and produces no additional finding on machines that are in Zone A. Overall RMS trended continuously, with high-rate capture on demand or on alarm, covers the same ground for a fraction of the cost.

Widely quoted failure-cause statistics should be checked. The percentage most often cited for the share of motor failures attributable to bearings traces back to studies published in the early 1980s and is routinely misquoted, so OptimizeOS does not use it. The defensible statement is simpler and holds up: bearings are a leading cause of rotating-equipment failure, and vibration is the measurement that sees them.