Most mechanical monitoring programs fail not because of wrong technology selection but because of wrong deployment: sensors on non-critical assets, no established baselines, alarms that no one responds to, and no connection to the maintenance workflow. Choose the system after you have defined how the program will operate.
Step 1: Identify Your Failure Modes First
Before selecting any technology, list the dominant failure modes for your asset fleet. This step is skipped more often than any other, and it is where the most money gets wasted.
Bearing fatigue, expressed as ball-pass fault frequencies (BPFO, BPFI), is best detected through vibration analysis. Imbalance and misalignment are also vibration problems. Lubrication condition is better assessed through ultrasound and oil analysis. Gear wear shows up in both vibration spectra and oil particle counts. Each failure mode has a preferred detection technique.
Building your technology selection around the actual failure modes in your fleet prevents a common outcome: purchasing a system that monitors the wrong parameters and generates no actionable information.
| Failure Mode | Preferred Detection Technique | Notes |
|---|---|---|
| Bearing fatigue (BPFO, BPFI) | Vibration analysis | Fault frequencies calculated from bearing geometry and RPM |
| Imbalance | Vibration analysis (1X component) | Strong signal at 1X running speed |
| Misalignment | Vibration analysis (2X component) | Strong 2X, often with axial component |
| Lubrication condition | Ultrasound, oil analysis | High-frequency friction precedes vibration signature |
| Gear wear | Vibration plus oil analysis | Gear mesh frequency in spectrum; particles in oil |
| Cavitation (pumps) | Vibration, pressure monitoring | Broadband noise floor increase in high-frequency range |
Step 2: Classify Assets by Criticality
Not every asset needs the same monitoring intensity. Criticality classification is the tool that determines what each asset actually needs.
Three practical tiers:
- Tier 1 (continuous monitoring): Single-train assets with no installed redundancy, high consequence of failure (safety, regulatory, or production loss above a defined threshold). These need permanent sensors and real-time data.
- Tier 2 (periodic route-based monitoring): Semi-critical assets with some tolerance for failure between inspection intervals. Spare capacity or short lead-time parts reduces the consequence of an undetected fault.
- Tier 3 (minimal or no monitoring): Non-critical assets with easy replacement, low consequence, and short repair time. Run-to-failure may be the correct strategy.
Most industrial plants have 10 to 30% of their assets in Tier 1. Deploying continuous monitoring across the entire fleet is rarely cost-justified. Not deploying it on Tier 1 assets is the more common mistake, and it is the more expensive one.
Complete the criticality classification before contacting vendors. It determines your sensor count, your budget split, and your deployment priorities, and it prevents vendors from simply recommending coverage of every measurement point.
Step 3: Choose the Sensing Approach
Three approaches are in widespread use. Each has a defined use case.
Continuous wireless sensors
Permanently mounted. Data transmits automatically on a defined interval or continuously. No measurement gaps between manual routes. Modern systems range from basic temperature-and-overall-vibration units to full-spectrum FFT sensors with automated fault frequency analysis. The correct choice for Tier 1 critical assets where continuous awareness of machine condition is required.
Route-based portable measurement
A technician carries a handheld analyzer and visits each measurement point on a defined schedule. Lower capital cost per point. Requires trained personnel and a defined route schedule. The critical limitation: the measurement interval defines the maximum detectable fault progression rate. A monthly route cannot reliably detect faults that develop and fail within three weeks. For slowly progressing failure modes on assets with adequate consequence tolerance, this is appropriate. For fast-developing faults on critical assets, it is not.
Hybrid deployment
Continuous sensors on the top 10 to 30 Tier 1 assets; route-based portable measurement on Tier 2. This is the most common cost-effective approach for mid-size plants. It concentrates continuous awareness where the consequence of a missed fault is highest.
Step 4: Evaluate Sensor Specifications for Your Asset Types
Sensor datasheets contain specifications that directly determine whether the sensor will detect the faults you care about. These are not marketing numbers.
Frequency range
Bearing fault detection requires the sensor to capture the fault frequency band. On an asset running at 1,800 RPM with a 9-ball bearing, the ball-pass outer-race frequency (BPFO) is approximately 90 Hz. A sensor rated to 1,000 Hz covers that comfortably. A sensor rated to only 200 Hz will also cover it. But on high-speed machinery above 3,600 RPM, bearing fault frequencies shift proportionally higher. A motor running at 7,200 RPM with the same bearing generates BPFO near 360 Hz. At that speed, a sensor with a flat response only to 200 Hz will produce attenuated or unreliable readings at the fault frequency.
Map your asset running speeds and bearing models to fault frequencies before specifying sensor bandwidth. The math is straightforward and takes less time than a vendor demo.
Mounting method
Stud mount (threaded insert drilled and tapped into the machine housing) provides the highest resonance frequency and most reliable high-frequency response. Use it on Tier 1 assets where permanent installation is practical. Adhesive mounting works for locations where drilling is not feasible; the resonant frequency is lower and high-frequency accuracy is reduced. Magnetic mounting is practical for route-based portable measurement where the technician moves between points.
Environmental ratings
Confirm the operating temperature range against the actual surface temperature at the mounting location, not the ambient temperature. Verify IP rating against the wash-down, moisture, and dust exposure the sensor will experience. If the asset is in a classified hazardous area, intrinsic safety (ATEX, IECEx, or NEC Class/Division) certification is a hard requirement, not an optional feature.
Step 5: Evaluate the Analytics Platform
The sensor captures data. The platform determines whether that data becomes a maintenance decision. Sensor selection and platform selection are equally important; in practice, the platform receives far less scrutiny.
Five questions to ask every vendor:
- Does the system calculate bearing fault frequencies automatically from the bearing designation, or does the user enter expected frequencies manually? Manual entry requires trained personnel and introduces entry errors. Automatic calculation from bearing catalog data scales to large fleets without proportional analyst effort.
- Does it trend amplitude at fault frequencies over time, or only compare to a fixed threshold? Fixed thresholds generate alerts only when damage is already advanced. Trending from an established baseline detects fault initiation earlier and gives more lead time for planned intervention.
- Does the system establish a baseline from initial measurements and alarm on deviation from that baseline? A machine running rough from installation will not alarm on an absolute threshold but will show deviation from its own normal pattern if baseline trending is in use.
- Does the system distinguish fault types (bearing degradation, imbalance, misalignment) from the spectrum automatically, or does interpretation require a trained analyst? If all faults generate the same generic alert, the alert loses credibility quickly. Maintenance teams stop responding to alerts they cannot interpret.
- Does the platform integrate with your CMMS or EAM to generate work orders when an alert fires? If the alert sits in the condition monitoring platform and requires a separate manual step to generate a work order, alarms will be missed during high-workload periods. Integration closes the loop between detection and action.
Step 6: Pilot Before Full Deployment
No vendor demonstration will predict how a system performs in your specific facility. Network interference, ambient vibration, temperature cycling, and the workflows of your actual team are not reproducible in a lab.
Install continuous sensors on 5 to 10 assets representing the variety in your fleet. Run the pilot for 3 to 6 months. Evaluate against these specific questions:
- Are alarms generating actions that the maintenance team considers credible and useful?
- Is the false alarm rate low enough that alerts are not being ignored?
- Does the interface fit how your team actually works, or does it require workflow changes that are unlikely to stick?
- What is the connectivity reliability in your facility? Dead zones, signal interference, and network configuration issues surface in the first weeks of a pilot, not in a vendor demo.
- What is the actual installation time and commissioning effort per sensor? Vendor estimates are typically optimistic.
Use pilot results to set realistic expectations for full-fleet deployment cost and timeline. Vendor claims made before a pilot are projections; pilot results are evidence.
Common Evaluation Mistakes
- Evaluating price per sensor point without accounting for total deployment cost. Installation labor, cable or wireless infrastructure, commissioning, and initial training can exceed the sensor hardware cost on a per-point basis. Get fully-loaded cost estimates, not component prices.
- Selecting by sensor count rather than by which assets most need monitoring. A proposal for 200 sensors across 200 points sounds comprehensive. If those 200 points include 160 Tier 3 non-critical assets and miss 10 Tier 1 assets without redundancy, the coverage is inverted.
- Ignoring the analytics capability. A low-cost sensor that generates overall vibration RMS and nothing else produces data, not decisions. The ability to detect specific fault types from frequency domain data is what generates actionable alerts; that capability lives in the analytics layer, not the sensor hardware alone.
- Not defining the response protocol before deployment. An alert with no defined owner, no connection to the work order system, and no escalation path will be missed. Before the first sensor goes on a machine, the workflow for responding to an alert should be documented and tested.
See Continuous Monitoring in Practice
Tractian's platform combines continuous vibration, temperature, and electrical monitoring with automated FFT analysis, fault frequency trending, and direct CMMS integration.
See Mechanical Monitoring