What Is Condition Monitoring?

Back to Blog

What Condition Monitoring Measures

The workhorse technique is vibration analysis: rotating equipment like pumps, compressors, and motors produces vibration signatures whose frequency content reveals imbalance, misalignment, bearing wear, and looseness, often weeks or months before failure. Temperature monitoring, including thermography, catches overheating bearings and electrical connections. Oil analysis measures wear metals, contamination, and lubricant degradation, effectively a blood test for the machine.

Other parameters include acoustic emission and ultrasound for early bearing defects and leaks, motor current signature analysis for electrical faults, and process parameters like efficiency or pressure differential that indicate degradation indirectly. Which techniques apply depends on the asset - a reciprocating gas compressor and a centrifugal pump are watched differently.

How It Enables Condition-Based Maintenance

The point of gathering these signals is to move away from either running to failure or replacing parts on a rigid calendar. With condition monitoring, maintenance is triggered by measured evidence of degradation - a rising vibration trend crossing a set threshold - which avoids both surprise breakdowns and the waste of servicing healthy equipment. This is condition-based maintenance, and it is the direct precursor to predictive maintenance, which adds forecasting of remaining useful life.

Increasingly, condition data flows into the same SCADA and telemetry systems that carry process data. A vibration or temperature sensor on a remote compressor can report over the same industrial protocols as the process instruments, letting a central platform trend machine health alongside production. Merobix reads sensor data over protocols such as Modbus and MQTT, so condition-monitoring readings from field equipment can be historized and trended in the same web-native view as everything else on the site.

Setting Alert Levels: Baselines Beat Generic Limits

The hardest part of a new program is deciding what counts as abnormal. Standards help you start: the zone-based approach of ISO 10816 vibration severity and its successors classifies broadband levels by machine class, which gives a defensible first threshold before any history exists. But generic limits are a starting point, not a destination - two healthy machines of the same model can idle at different levels, so the stronger reference is each machine's own baseline, captured when the machine is known to be in good condition after commissioning or overhaul.

From the baseline, alerting becomes relative: flag meaningful change from a machine's own normal rather than the crossing of a universal number. A two-level scheme is conventional - a lower level that prompts investigation and trend review, and a higher level that prompts action - with the specific values set per machine, per the standards guidance and the manufacturer's documentation. Review the levels after every overhaul, because the baseline that follows a rebuild is a new baseline.

Continuous, Periodic, or Route-Based: Program Shapes

Not every asset earns permanently installed sensors. The classic split is criticality-driven: continuous online monitoring for machines whose failure stops production or creates a safety hazard, and periodic measurement for the balance of plant. The periodic tier is usually organized as route-based vibration monitoring - a technician walks a planned route with a portable analyzer, collecting the same points in the same order at a set interval, so trends stay comparable month over month.

Wireless battery-powered sensors have shifted the economics of the middle tier, making it practical to instrument machines that never justified wired channels. Whatever the tier, measurement consistency decides data quality: the same location, the same orientation, and a proper mount. Poor vibration sensor mounting quietly filters out the high-frequency content where early bearing faults live, which makes a program look healthy right up until a bearing is not.

Standing Up a Program Without Boiling the Ocean

1. Rank assets by criticality: production impact, safety consequence, repair cost, and spares lead time.
2. For the top tier, identify the dominant failure modes and pick the techniques that actually catch them.
3. Capture known-good baselines for every monitored point.
4. Set two-level alerts from standards guidance, manufacturer documentation, and the baselines.
5. Define the response path: who reviews an alert, within what time, with what authority to act.
6. Review quarterly: retire points that never inform a decision, add coverage where surprises still happen.

Step five is the one programs skip and regret. A rising trend that nobody owns is just data; the program only earns its keep when an alert reliably becomes a work order before it becomes a breakdown. Deciding in advance who acts, and on what evidence, is worth more than another sensor.

Pitfalls That Hollow Out a Program

The common failure modes are organizational more than technical. Thresholds set too tight generate a stream of alerts that train everyone to ignore them, and the one alert that mattered drowns in the noise. Baselines never get refreshed after an overhaul, so a rebuilt machine trends against the worn machine's normal and either alarms constantly or hides a new defect. Measurement points wander - a different location, a different mount, a different load condition - and the resulting trend says more about the collection than about the machine.

The defense against all of these is the same habit: judge trends, not single readings, and collect under comparable conditions. A single elevated sample at an unusual load proves little; three collections in a row climbing under the same conditions is a machine telling you something. Programs that survive their first few years are the ones that ruthlessly prune points nobody acts on, keep collection conditions consistent, and treat every confirmed catch - and every miss - as a reason to adjust what gets measured and at what levels.

Frequently Asked Questions