How five condition-monitoring methods inform one machine decision

The real constraint

The hard part is matching evidence to a comparable machine state.

In many connected plants, permanent sensors, handheld routes, oil-laboratory reports, thermal surveys, historian tags, motor-control data and maintenance records already exist. A practical constraint is deciding which records describe a comparable machine state, which ones do not, and whether the combined evidence changes the action.

A lubrication problem may affect friction, temperature, vibration and wear debris. A process disturbance may change motor load, vibration and temperature without creating a local defect. Two unrelated mechanisms may also appear in the same week. Integration makes each observation testable instead of allowing the first alarm to become the story.

A useful review separates machine change from measurement or operating-state change, exposes evidence against the first fault label and identifies the safest next check that can alter the work scope.

Five disciplines

Each method measures a different physical effect.

The strength of an integrated programme comes from those differences. Vibration can help localize dynamic mechanical behaviour. Thermography compares apparent surface-temperature patterns. Ultrasound detects high-frequency emissions associated with friction, impacts or leakage under a controlled method. Motor current signature analysis adds electrical and driven-load evidence. Oil analysis adds evidence associated with lubricant condition, contamination and wear in a sampled compartment. None is a complete diagnosis on its own.

The method scopes in this comparison follow primary guidance for vibration measurement, thermography, airborne and structure-borne ultrasound, online electrical signature analysis, and tribology-based monitoring.

These are ConditionCore's five current disciplines, not an exhaustive catalogue of condition-monitoring methods. Integration does not mean applying all five to every asset at every interval. Select methods against credible failure modes, asset criticality, access and safety, warning interval, and whether the result can change a maintenance decision.

The fifth discipline

Motor current tests motor, supply and driven-load hypotheses.

ConditionCore uses Motor Current Analysis as the interface label. The online method is motor current signature analysis, commonly abbreviated MCSA. ISO 20958 sets guidelines for online techniques based on electrical signature analysis of three-phase induction motors. This is distinct from de-energized motor circuit analysis, which is also shortened to MCA.

Rotor asymmetry, air-gap effects and changing load torque can modulate stator current. A feature can also change with slip, load, supply distortion, motor design or VFD and control state. That makes MCSA useful for screening and hypothesis testing, but weak as a standalone locator. When voltage is available, current and voltage together can help separate motor response from supply influence.

The motor must run for current data to represent its operating state. Installing or connecting sensors is a separate task. De-energize, lock out and verify absence of voltage before exposure to live parts unless qualified personnel perform an authorized energized test under the applicable electrical-safety procedure. Online data does not waive electrical or mechanical guarding. Other motor and drive types require method-specific acquisition and interpretation.

What integration requires

Build one shared machine record before looking for agreement.

Two trends that rise on the same calendar day are not automatically related. Before the team tests a common mechanism, every file or stream needs enough context to show where it belongs and whether the comparison is fair.

Provenance is not clerical overhead. It lets another reviewer challenge an apparent agreement instead of inheriting it. If the review begins with one alarm, use the five-question signal method before bringing in the wider evidence set.

Alignment does not require simultaneous acquisition. A vibration waveform may represent seconds, a thermal image one operating moment, motor current a matched run and oil analysis a slower compartment trend. Define a relevance window appropriate to the mechanism and operating state instead of pretending unlike observations share one timestamp.

Reasoning without alarm votes

Interpret each channel within its physical scope.

Three elevated channels are not three votes for the same diagnosis. State what each method observes, what it can localize, which explanation it supports and what it leaves unresolved. Then test three possibilities: one plausible common mechanism, multiple independent mechanisms, or evidence that is not yet sufficient. Agreement is useful when the roles fit. Disagreement is useful when it changes the next test.

• Map the machine and its state.Keep the train, compartments, feeder, points, speed, load, process state and maintenance events together.
• Test competing explanations.Show what supports the leading hypothesis, what contradicts it, what remains unresolved and whether an independent mechanism also fits.
• Choose a discriminating action.Prefer the safest repeat measurement or inspection most likely to change the work scope or action window.