A condition monitoring program is a managed system for measuring equipment health and acting on it, selecting the assets worth watching, matching monitoring techniques to real failure modes, running repeatable measurement routes, setting alarm limits, and converting readings into planned work orders. It is a program, not a purchase: sensors without the routes, limits, and follow-through are just expensive decoration.

Most condition monitoring efforts stall because they start at the wrong end, someone buys vibration sensors, mounts them on whatever is nearby, and then nobody owns the alarms. This guide builds the program the way the international standard for it, ISO 17359, actually sequences the work: assets and failure modes first, parameters and baselines next, limits and routes after that, and diagnosis and action as the payoff.

What is a condition monitoring program?

A condition monitoring program is the organized practice of periodically or continuously measuring parameters that indicate machine health, vibration, temperature, oil condition, current, pressure, comparing them against baselines and limits, and using the result to decide maintenance. ISO 17359, the general-guidelines standard for condition monitoring and diagnostics of machines, frames it as a closed loop directed at the failure modes that actually threaten each asset, rather than a generic data-collection exercise.

The distinction from a strategy matters. Condition-based maintenance is the maintenance strategy, act when condition crosses a threshold. A condition monitoring program is the operational machinery that makes that strategy real: who measures what, how often, against which limits, and where the reading goes. You can also run a monitoring program that feeds a full predictive maintenance effort; the difference between those two is covered in condition monitoring vs predictive maintenance.

The condition monitoring program loopA program is a loop, not a purchase1 · SELECT ASSETSby criticality2 · FAILURE MODESwhat actually breaks3 · PARAMETERSmeasurable signal4 · BASELINE + LIMITSalert / action5 · MEASURE (ROUTE)repeatable points6 · DIAGNOSE + TRENDcompare to baseline7 · ACT (WORK ORDER)planned repairreview limits and asset list on a cycle, the loop never stops
ISO 17359 in shape: the program flows from asset selection through failure modes, parameters, limits, and routes to diagnosis and action, then feeds back to refine the list and the limits.

Which assets should a program cover?

Cover the assets whose failure consequence justifies the monitoring cost, not everything that spins. A condition monitoring program applied plant-wide dilutes attention and drowns the owner in low-value alarms. Rank assets by criticality (safety, downtime, repair cost, redundancy) and start monitoring at the top of the list.

The practical filter is two questions borrowed from every good reliability program: does this asset fail in a way that hurts, and is that failure detectable before it happens? A cheap, redundant, run-to-failure motor fails neither test and belongs on no route. A critical unspared gearbox with measurable bearing degradation passes both and is exactly what a program exists for.

Which techniques match which failure modes?

Choose the monitoring technique from the failure mode, never the other way around. ISO 17359 is explicit that monitoring should be directed at root-cause failure modes, you pick the parameter that changes first as that specific mode develops. The matrix below is the starting map for rotating and fluid equipment.

Failure modeParameter to monitorTechnique
Bearing wear, imbalance, misalignment, loosenessVibration velocity / accelerationVibration analysis (route or permanent sensor)
Loose electrical connection, overload, frictionTemperatureInfrared thermography
Lubricant degradation, contamination, wear metalsOil chemistry and particle countOil analysis (lab sampling)
Rotor bar, winding, load anomalyCurrent / voltage signatureMotor current analysis
Steam-trap leak, valve passing, early bearing faultUltrasound (airborne / structure-borne)Ultrasonic acoustic monitoring
Filter loading, blockage, pump wearDifferential pressure / flowProcess-parameter monitoring

Most programs start with two techniques, vibration on critical rotating equipment plus thermography on electrical gear, because that pair covers the majority of high-consequence failure modes at modest entry cost. Add oil, current, and ultrasound where the asset base earns it. Doubling as a check on your lubrication program is a common bonus: a large share of oil-analysis findings are contamination, not machine wear.

How do you set routes and alarm limits?

A route is the repeatable path and set of measurement points a technician (or a fixed sensor network) covers on a schedule; alarm limits are the alert and action thresholds that turn a reading into a decision. Repeatability is everything, the same point, same mounting, same load condition, same interval, because a trend is only meaningful when the measurement conditions are constant.

For alarm limits, start from a defensible external reference, then tune to your own history. For rotating equipment, ISO 20816 gives vibration severity zones you can anchor to before you have enough of your own data.

ISO 20816 vibration severity zonesSeverity zones give you a starting limitZONE Anewly commissionedZONE Bunrestricted long-term operationZONE Cnot for long-term run · remedial actionZONE Ddangerous · immediate actiongoodbadset ALERT near the B/C boundary, ACTION near C/D, then tune to your history
ISO 20816 severity zones A–D for broadband bearing-housing vibration. They are a defensible default: put your alert limit near the B/C boundary and your action limit near C/D, then adjust with your own baselines and failure history.

How do you build the program? A step sequence

  1. Rank assets by criticality and pick the pilot set. Take the top 10–20 critical assets where failure hurts and degradation is measurable. A program proves itself on a focused set before it earns a plant-wide budget.
  2. Identify the failure modes that actually occur. Pull them from work-order history and downtime reason codes. Monitor the modes you have, not the ones a sensor brochure lists.
  3. Match a technique and parameter to each mode. Use the failure-mode matrix above. Prefer the earliest-detecting parameter your budget supports.
  4. Establish baselines under normal load. Capture the healthy-state reading for every point. A limit without a baseline is a guess; connected machines often already hold weeks of history to baseline from.
  5. Set alert and action limits. Anchor to ISO 20816 zones or manufacturer specs, define an alert level (open an inspection) and an action level (open a work order), and expect to tune both.
  6. Build the route and assign the owner. Fix the points, interval, and method, and name the person who screens readings and closes the loop. An unowned program dies at the first ignored alarm.
  7. Wire readings to work orders, then review quarterly. An out-of-limit reading must become a planned work order, not an email. Each quarter, count catches and false alarms and re-tune limits and the asset list.

Why do condition monitoring programs fail?

Most programs die of neglect, not bad sensors. The failure patterns are predictable, and every one of them is a management problem rather than a technology problem:

The through-line is follow-through. A program that reliably turns readings into planned work, and reviews its own limits and asset list on a cycle, survives; one that stops at the dashboard does not. This is the same discipline that separates a working TPM program from a wall of laminated checklists nobody signs.

What does a working program require and return?

The evidence and economics come from primary sources, not vendor decks:

Track the program with MTBF trending up and unplanned downtime trending down on monitored assets, inside your broader maintenance KPI set. The most common failure point is not the sensing, it is that readings, baselines, work orders, and downtime logs live in separate systems, so no one sees the whole picture. Connecting machine monitoring data with work and parts records into one operational layer, the way Harmony does with no rip-and-replace, is what turns a folder of readings into acted-on decisions; see the CLS case study for unified plant data in practice.