Key Points
- A condition monitoring program starts with a scored asset list, before any sensor is bought.
- Match each asset's dominant failure mode to the technique that detects it earliest.
- Continuous sensors go where a failure costs the most or a route reading is too far apart or unsafe.
A condition monitoring program is built in a fixed order: list the assets, score how much each failure would hurt, match failure modes to techniques, then decide which machines get continuous sensors and which stay on a route. Teams that start by buying sensors usually end up with sensors on the easy-to-reach machines instead of the ones that stop the plant.
This guide walks through that order with a worked example: 20 assets in a hypothetical food plant, scored on a simple worksheet, with the coverage decision for each one. The numbers are example data. Swap in your own asset register and the same math applies.
What is a condition monitoring program?
A condition monitoring program is the plan, people and tools a plant uses to measure machine health on a set schedule or continuously, and to turn those measurements into maintenance work before a failure. It defines which assets are monitored, with which technique, how often, who reads the data and what happens when a reading changes.
It is the execution layer under a condition-based maintenance program. CBM is the strategy of triggering work from asset condition. The condition monitoring program is the rollout: the sensors, routes, thresholds and alert owners that make that strategy run on the floor.
Step 1: Build the asset register and set the scope
Start with a list of the rotating and electrical assets in the area you plan to cover, one row per machine. A first program usually covers one production line, one utility system or one building, because a team can respond to alerts from 20 to 50 assets far more reliably than from 500.
For each row, record what the worksheet in Step 2 needs:
- Asset name and tag: the name the technicians use, plus the tag in your maintenance system so alerts can become work orders later.
- Function and redundancy: what the machine feeds, and whether an installed spare can take over.
- Failure history: breakdowns from the last few years, taken from work orders and from the supervisors who remember the ones never logged.
- Access: whether a technician can safely reach the machine while it runs. Cooling tower fans, roof units, lift station pumps and hazardous areas often fail this test.
Split drive trains into their parts. An extruder's motor and gearbox fail in different ways, so they get separate rows.
Step 2: Score asset criticality
Score each asset on the consequence of a failure and the likelihood of one, then multiply. The product ranks the asset list so the first sensors go where a failure costs the most. This is a simplified form of criticality analysis, sized for a team building its first program.
The scoring scale
Score three consequence factors and one likelihood factor from 1 to 5:
| Factor | Score 1 | Score 3 | Score 5 |
|---|---|---|---|
| Production impact (P) | Failure stops nothing; an installed spare takes over | Line slows or a short stop while a spare is swapped in | Line or plant stops with no spare in place |
| Safety and environment (S) | No safety or environmental exposure | Minor exposure, contained by existing controls | Release of a hazardous fluid, fire risk, or a reportable environmental event |
| Repair cost and lead time (C) | Cheap part held in stock, repaired within a shift | Repair takes days or needs an outside shop | Long-lead part or rebuild measured in weeks |
| Likelihood (L) | No failure on record and a light duty cycle | Failed once in recent years or runs a heavy duty cycle | Repeat failures or known bad actor |
The formula is Score = (P + S + C) x L. The highest possible score is 75: (5 + 5 + 5) x 5. Adding the consequence factors before multiplying keeps one high safety score from being hidden by low production impact, which matters for assets like an ammonia compressor.
Set the decision rules before you score, so the results do not get bent toward the machines people already like:
- 27 or higher, and the dominant failure mode shows up in vibration, ultrasound or temperature: continuous monitoring.
- 21 to 26, and the asset is hazardous or hard to reach while running: continuous monitoring, because a route cannot collect the reading safely.
- 15 to 26 otherwise: route-based monitoring with handheld tools or periodic samples.
- Under 15: preventive maintenance or run-to-failure.
Worked example: 20 assets scored
The plant below is a hypothetical food and beverage site with one extrusion line, a spiral freezer, ammonia refrigeration, a boiler and the usual utilities. The scores are example data, sorted from highest to lowest.
| Asset | P | S | C | L | Score | Dominant failure mode | Coverage decision |
|---|---|---|---|---|---|---|---|
| Extruder gearbox | 5 | 2 | 5 | 3 | 36 | Gear wear, bearing wear | Continuous |
| Boiler feedwater pump A (duty) | 4 | 4 | 3 | 3 | 33 | Bearing wear, misalignment | Continuous |
| Extruder main drive motor (250 hp) | 5 | 2 | 4 | 3 | 33 | Bearing wear, rotor bar damage | Continuous |
| Process cooling water pump (no spare) | 5 | 2 | 3 | 3 | 30 | Bearing wear, cavitation | Continuous |
| Ammonia refrigeration compressor | 4 | 5 | 5 | 2 | 28 | Bearing wear, lubrication failure | Continuous |
| Cooling tower fan gearbox | 3 | 2 | 4 | 3 | 27 | Gear wear, lubrication failure | Continuous |
| Rotary screw air compressor 1 (duty) | 4 | 1 | 4 | 3 | 27 | Bearing wear, overheating | Continuous |
| Spiral freezer fan | 4 | 2 | 3 | 3 | 27 | Unbalance from ice buildup, bearing wear | Continuous |
| Hydraulic power unit, press | 3 | 2 | 3 | 3 | 24 | Pump wear, fluid contamination | Route-based |
| Wastewater lift pump (duplex) | 2 | 4 | 2 | 3 | 24 | Clogging, bearing wear | Continuous (hard to reach) |
| Dust collector fan | 2 | 3 | 2 | 3 | 21 | Unbalance from buildup | Continuous (hard to reach) |
| Main conveyor head pulley drive | 4 | 1 | 2 | 3 | 21 | Bearing wear, belt wear | Route-based |
| Mixer gearbox | 4 | 1 | 4 | 2 | 18 | Gear wear | Route-based |
| CIP transfer pump (spare on shelf) | 2 | 1 | 1 | 3 | 12 | Seal leak | Preventive or run-to-failure |
| Vacuum pump, packaging | 3 | 1 | 2 | 2 | 12 | Vane wear | Preventive or run-to-failure |
| Boiler feedwater pump B (standby) | 1 | 4 | 3 | 1 | 8 | Bearing damage from sitting idle | Preventive or run-to-failure |
| Office AHU supply fan | 1 | 1 | 2 | 2 | 8 | Belt wear | Preventive or run-to-failure |
| Palletizer conveyor motors | 2 | 1 | 1 | 2 | 8 | Bearing wear | Preventive or run-to-failure |
| Restroom exhaust fan | 1 | 1 | 1 | 2 | 6 | Bearing wear | Preventive or run-to-failure |
| Rotary screw air compressor 2 (standby) | 1 | 1 | 4 | 1 | 6 | Bearing wear | Preventive or run-to-failure |
What the worksheet tells you
The 20 assets split into 10 for continuous monitoring, 3 for route-based monitoring and 7 for preventive maintenance or run-to-failure. A few results are worth a closer look, because they are where teams usually argue:
- The ammonia refrigeration compressor scores 28 with a likelihood of 2. Its safety and repair-cost scores carry it. Likelihood-heavy formulas would push it down the list, which is the wrong call for an asset that can release ammonia.
- Feedwater pump A scores 33 and pump B scores 8. The standby pump barely runs, so its production impact and likelihood stay low. Monitor the duty pump continuously and rotate the pair on a schedule so the standby does not sit with flat-spotted bearings.
- The wastewater lift pump and dust collector fan land at 24 and 21. On score alone they would be route assets. Both sit where a technician with a handheld collector has a hard time getting a safe reading, so they move to continuous.
- The hydraulic power unit stays on a route at 24. Its main failure mode is fluid contamination, which shows up in an oil sample before it shows up in vibration. A monthly sample covers it.
- The mixer gearbox scores 18. It hurts production when it fails, but it has no failure history. It stays on a monthly route and moves up if its likelihood score changes.
Rescore the list once a year and after any major failure. A failure is new likelihood data.
Step 3: Match each failure mode to a monitoring technique
Pick the technique from the failure mode. A gearbox that fails from gear wear needs vibration and oil analysis; a motor that fails from rotor bar damage needs a current test. The table maps common failure modes to the technique that tends to detect them earliest.
| Failure mode | Technique that shows it first | What changes in the data | Usual coverage |
|---|---|---|---|
| Rolling bearing wear | Vibration (envelope and high-frequency bands) and ultrasound | Bearing defect frequencies appear, ultrasound level rises | Continuous on critical assets, route on the rest |
| Lubrication failure | Ultrasound, then temperature | Friction raises the ultrasound level well before the bearing housing heats up | Continuous, so lubrication can follow condition |
| Unbalance, misalignment, looseness | Vibration | Amplitude grows at 1x and 2x running speed, phase shifts between bearings | Continuous or monthly route |
| Gear wear | Vibration and oil analysis | Gear mesh frequency and its sidebands grow; wear metals rise in the oil sample | Continuous vibration plus quarterly oil samples |
| Pump cavitation | Vibration and ultrasound | Broadband noise and random high-frequency energy at the pump casing | Continuous where suction conditions change |
| Broken rotor bars, electrical faults | Motor current signature analysis | Sidebands appear at twice the slip frequency on each side of line frequency | Periodic MCSA test or online electrical monitoring |
| Hot electrical connections | Infrared thermography | Temperature rise across one phase or one lug | Quarterly or annual thermography route |
| Hydraulic fluid contamination | Oil analysis | Particle count and water content climb | Monthly or quarterly sample |
Most rotating assets in the worksheet fail through bearings, lubrication, alignment or gears, which is why vibration analysis and ultrasonic testing carry the bulk of a first program. Oil analysis, infrared thermography and motor current signature analysis fill the gaps that vibration and ultrasound do not reach: wear metals, hot electrical connections and rotor faults.
Step 4: Decide continuous vs route-based coverage
Continuous monitoring fits assets where a fault can develop between two route visits, the failure is expensive, or the machine is unsafe to approach. Route-based monitoring fits assets whose faults develop slowly enough for a monthly or quarterly reading to catch them.
The test is the gap between readings compared with how fast the fault grows. A bearing defect on a high-speed fan can go from first sign to failure in a few weeks, so a monthly route can miss most of that window. Oil degradation in a hydraulic unit usually moves over months, so a monthly sample sees it coming.
For the worked example, that gives:
- Continuous (10 assets): extruder gearbox and motor, feedwater pump A, the process cooling water pump, the ammonia compressor, the cooling tower fan gearbox, the duty air compressor, the spiral freezer fan, the lift pump and the dust collector fan.
- Route (3 assets): hydraulic power unit (monthly oil sample), main conveyor head pulley drive and mixer gearbox (monthly vibration and ultrasound readings).
- Preventive or run-to-failure (7 assets): the standby units, the CIP transfer pump with a spare on the shelf, the packaging vacuum pump, palletizer motors and the building fans.
The choice of a handheld vibration analyzer or continuous monitoring comes down to the same test, asset by asset.
Step 5: Plan the sensor installation
Plan each sensor location before the install day: which bearing housing, which axis, which mounting method, and how the data leaves the machine. Getting these four right decides whether the data is usable.
- Location: mount as close to the load zone of the bearing as the housing allows. On a motor-pump set, a common layout is one sensor on the motor drive end and one on the pump bearing housing.
- Mounting: a drilled and tapped stud or a bonded base gives a solid path for high-frequency signals. Avoid mounting on thin covers, fan shrouds or painted guards.
- Connectivity: decide early whether sensors run on the plant network or their own link. Plant Wi-Fi approvals can add weeks to a rollout. Learn more about wireless vs wired condition monitoring.
- Hazardous areas: check the area classification for each asset. In the worked example, the ammonia compressor room needs sensors certified for that location.
Tractian sensors mount with adhesive or drill-and-tap and are designed to install in minutes on running equipment, and they send data over 4G/LTE, so the rollout does not wait on plant Wi-Fi or new wiring. They are IP69K sealed, ATEX/IECEx compliant and Class I Div 1 certified, with NFPA 70 Class 1, 2, and 3 (Division I) compliance.
Tractian also handles the install. CMRP/CAT-certified specialists choose sensor placement on each asset, mount the sensors and train your technicians on site, then stay connected with the field team after go-live.
Step 6: Set baselines and name an owner for each alert
Each newly monitored asset needs a baseline of normal behavior before its alerts mean much, and each alert needs a named person who acts on it. Programs that skip the second part produce data that nobody reads.
Collect the baseline across the asset's normal operating range: different loads, speeds and product runs. A spiral freezer fan reads differently with a full load of product than at idle, and a baseline taken at one state will raise false alerts at the other.
Then write the response rule for each severity level:
- Who reviews: usually the reliability engineer, or the maintenance planner where there is no reliability role.
- How fast: for example, critical alerts reviewed the same shift and warnings within two working days.
- What follows: an inspection, a planned repair or a documented decision to keep running with a closer watch.
Ranking alerts by the criticality score from Step 2 keeps a warning on the extruder gearbox ahead of a warning on the dust collector fan when both arrive on the same morning.
Tractian builds this ranking into the alerts themselves. Alerts are ranked by failure severity and asset criticality, so mission-critical machines rise to the top while low-impact noise stays out of the way.
Tractian benchmarks each asset's data against millions of historical machine hours and bearing libraries to make an accurate, explainable call, complete with evidence.
Step 7: Connect alerts to work orders
An alert should become a work order in the system your technicians already use, with the asset tag, the fault, the severity and the supporting data attached. When an alert sits in a separate dashboard, it depends on someone remembering to check it.
Map the fields before go-live: asset tag, fault description, recommended action, priority and the planner who receives it. Then close the loop after the repair. The post-repair readings show whether the fix worked, and that record is what makes the program credible with operations.
Tractian connects condition alerts to work orders in its own platform and is built to integrate with ERP and EAM systems such as Oracle, Infor and Maximo, plus BI tools through a SQL connector.
Step 8: Measure the first wave and expand
Run the first wave for a set period, such as one quarter, and measure it before adding assets. Expansion decisions should rest on what the first wave caught and what it cost the team in time.
Track a short list of program measures:
- Coverage: share of assets above your continuous threshold that are actually monitored.
- Alert response: share of alerts reviewed within the response time you set in Step 6.
- Alert to work order: share of alerts that led to an inspection or repair.
- Confirmed catches: faults found on inspection after an alert, with the part and the condition photographed.
- No-fault-found rate: alerts where the inspection found nothing, a sign that baselines or thresholds need work.
- Planned vs unplanned work: the share of maintenance hours on the monitored assets that was planned.
The second wave comes from the worksheet: the route assets that moved up after rescoring, plus the next line or utility system. To put a price on each wave, budget for the true cost of a condition monitoring program, including what it spends beyond sensors.
Mistakes that stall new programs
Most stalled programs trace back to scope and ownership decisions made in the first month.
- Monitoring what is easy to reach. The first sensors go on accessible machines near the shop, while the cooling tower gearbox on the roof stays unmonitored.
- Skipping the failure mode step. A vibration sensor on a hydraulic unit that fails from contamination will stay quiet until the pump is already damaged.
- No alert owner. Alerts land in an inbox shared by six people, and each assumes another one handled it.
- Baselines taken at one operating state. The team then spends the first month chasing alerts that are just a load change.
- Too many assets in wave one. 200 assets and two people to review them means slow responses, and slow responses erode trust in the alerts.
How Tractian fits a new condition monitoring program
Tractian gives a new program continuous coverage for the assets at the top of the worksheet without a long IT project. Each wireless sensor measures vibration, ultrasound, temperature and RPM in one device, which covers bearing wear, lubrication, unbalance, misalignment and gear faults from Step 3 on a single mount.
The data goes to the Tractian condition monitoring platform, where Auto Diagnosis maps signal patterns to specific failure modes and ranks alerts by severity and asset criticality. Each alert carries the probable root cause and recommended next steps, so the alert owner from Step 6 starts with a diagnosis instead of a raw spectrum.
- Fast setup: wireless sensors over 4G/LTE, mounted in minutes, with CMRP/CAT-certified Tractian specialists handling placement, installation and team training.
- Hazardous locations: the certifications above cover rooms like the ammonia compressor in the example.
- Integrations: alerts flow into Tractian work orders or into the ERP and EAM systems the plant already runs.
To estimate the return of a first wave before you buy, try the online condition monitoring ROI calculator.
Frequently Asked Questions
How many assets should the first wave of a condition monitoring program cover?
Size the first wave by how many alerts your team can review on time, which for many plants means 10 to 50 assets in one line or utility system. In the worked example above, 10 of 20 assets qualified for continuous monitoring. Start there, measure response times, then add the next group.
What is the difference between a condition monitoring program and condition-based maintenance?
Condition-based maintenance is the strategy of triggering maintenance from measured asset condition instead of the calendar. A condition monitoring program is how that strategy runs in practice: which assets are measured, with which technique, how often, and who acts on the data. A plant can have a CBM policy on paper and no working program behind it.
Should standby equipment be monitored?
Standby units usually score low on criticality because they rarely run, so they tend to sit on preventive maintenance. The common risk is bearing damage from sitting idle under vibration from nearby machines. Rotate duty and standby units on a schedule, and take a reading on the standby each time it runs.
How long does a new condition monitoring program take to produce useful alerts?
Continuous sensors start sending data on the day they are installed, but alert thresholds need a baseline across the asset's normal operating range. Plan for several weeks of normal running, covering different loads and products, before judging alert quality. Faults that are already present at install time can show up much sooner, since they stand out against reference data for that machine type.
Who should own a condition monitoring program?
One named person should own the program, usually a reliability engineer or the maintenance manager. That owner keeps the criticality worksheet current, sets response rules and reports program measures. Alert review can be shared with planners and senior technicians, but each severity level needs a named first responder.
Does Tractian offer installation and support?
White-glove, field-proven support from CMRP/CAT-certified specialists, shoulder-to-shoulder from pilot to scale. Tractian handles rollout and change management: sensor placement and installation, alert tuning, and hands-on training for your team.
After go-live, Tractian stays with you through sensor-health monitoring and rapid escalation when needed.
Ready to put continuous monitoring on the assets at the top of your worksheet? See how Tractian condition monitoring installs on running equipment, connects to your maintenance systems and flags failure modes early.

