Choosing the right sensors for condition-based monitoring
Unplanned equipment failures are a costly and frustrating source of production interruptions. Condition-based monitoring systems eliminate this problem by providing real-time insight into the condition of equipment.
It requires measuring parameters such as vibration, temperature and oil condition to detect signs of potential upcoming failure. Successful condition-based monitoring programmes combine different technologies to catch failures. A single sensor isn’t enough to cover all the possible types of machine failure.
This guide provides an overview of some of the most common sensor technologies and networking approaches to design and deploy an effective condition-based monitoring programme.
What is condition-based monitoring?
Condition-based monitoring involves technicians tracking equipment by measuring run data like vibration, heat or oil quality. This provides insights into whether machines are operating within safe baseline limits. It also means that repairs are done more efficiently only when data points cross a set limit.
This approach differs a little from predictive maintenance. Condition-based monitoring looks at the current state of a machine. Predictive maintenance uses models to predict exact failure dates. But both methods use the same basic data networks.
How to match sensors to failure modes
The P-F curve shows the window between first detecting a fault and final asset failure [1]. High-frequency signals show up first on this timeline.
Matching the tool to the failure modes means finding which signal appears earliest. Subsurface cracks produce ultrasonic waves weeks before any heat signatures appear. The wrong sensor frequency results in the alert arriving too late.
Core sensor types for condition-based monitoring
Selecting specific tracking tools requires understanding their physical limits and target applications. This includes vibration, temperature, ultrasonic, acoustic emission sensors and more.
Vibration sensors
- Industrial vibration monitoring tracks machine movement to detect defects.
- Technicians use a piezoelectric accelerometer to capture high frequencies typically effective up to 10-14 kHz. These devices find early bearing cracks.
- Velocity sensors track mid frequencies from 10 Hz to 1 kHz. This spectrum captures imbalance and shaft misalignment.
- Engineers use ISO 20816 standards to evaluate vibration severity [2]. Spikes at specific frequencies reveal the precise bearing defect frequency which shows race wear.
Temperature sensors and thermal imaging
- Changes in surface temperature are often signs of increased friction or electrical resistance.
- Fixed infrared sensors and periodic thermal imaging cameras detect these heat signatures without touching components. They show insulation faults and blocked cooling lines.
- A thermal sensor detects a fault only after heat moves to the outer casing. Heat tracking is further down the P-F curve than acoustic methods.
Ultrasonic sensors
- Non-contact ultrasonic monitoring detects high-frequency sound waves between 20 kHz and 100 kHz. These ultrasonic frequencies do not travel very far, so it is easy to pinpoint the fault source.
- Airborne sensors detect gas turbulence from vacuum leaks. Inside a housing, contact probes record sound waves from metal fatigue. These sensors alert teams before physical vibration changes happen.
Oil analysis and particle counting
- In-line fluid sensors provide continuous oil analysis by counting tiny wear particles and measuring moisture.
- This approach tracks the health of internal lubrication in gearboxes. Optical particle counters classify fluid cleanliness using the ISO 4406 standard, which codes contamination levels based on particle counts per millilitre at three size thresholds: ≥4 µm, ≥6 µm, and ≥14 µm [3].
- Chemical sensors check for oil breakdown. These tools show damage to the gear teeth before the fault causes structural vibration.
Motor current signature analysis (MCSA)
- Technicians use motor current signature analysis (MCSA) to evaluate equipment by reading magnetic fields. Current clamps attach inside electrical cabinets. This step removes the need for motor-mounted hardware.
- This condition monitoring technology finds rotor bar cracks by checking current sidebands. A unique fault signature appears when load shifts alter the motor current. MCSA flags electrical problems and mechanical imbalances.
Acoustic emission sensors
- Acoustic emission devices detect stress waves generated by micro-cracking.
- These sensors operate at high frequencies between 100 kHz and 1 MHz. They monitor slow machinery where standard accelerometers lack signal strength.
- These tools monitor crane health and slow roller bearings that operate below 10 RPM. The sensor captures energy from structural shifts. This gives immediate notification of damage before parts separate.
- Research has shown that acoustic emission techniques can detect subsurface bearing cracks as small as 0.5 mm [4].
Wired vs wireless sensor networks
Selecting a network requires balancing setup costs against data speeds. Wired networks are reliable and uninterrupted raw data streams with no signal interference. They are best for critical equipment, like high-speed turbines, that could be unsafe if they miss a data cycle.
Wireless networks are significantly less expensive and faster to install, as you don’t need to run cables across your plant. The battery-powered sensors transmit data at preset intervals. Wireless networks are a good solution for secondary machines, but battery life means you have to transmit less frequently.
Continuous vs route-based monitoring
Continuous tracking keeps sensors permanently attached to machines for live data streaming. This method works best for critical machinery where faults develop quickly. It also removes blind spots between technician inspections.
Route-based tracking uses portable handheld data collectors. A technician walks a designated route within the factory once a month to collect data. This method reduces initial equipment costs but can easily overlook unexpected breakdowns that occur between inspections.
Sensor selection criteria
Reliability teams must check the following four areas before purchasing condition-based monitoring sensors:
- Machine criticality - Place permanent, high-frequency continuous sensors on machinery that stops production lines.
- Primary fault modes - Match the device to the expected fault signature. Use accelerometers for bearings and thermal probes for electrical circuits.
- Local environment - Choose sensor housings with correct IP ratings to withstand high heat, moisture or chemicals.
- Data capabilities - Confirm networks can handle large files, or if local devices must use edge processing to compress data.
Common sensor selection pitfalls
Deploying condition-based monitoring sensors without a plan can cause data issues, including the following:
- Buying low-frequency sensors for high-speed machinery, which cuts off early fault signatures
- Installing wireless units without checking local plant radio interference
- Neglecting the cost of extra items like sensor cables and replacement batteries
- Purchasing delicate laboratory probes for heavy industrial areas where impacts occur
- Forgetting to match thread sizes to existing machine holes, delaying setup.
System connectivity and work orders
The collected signals need a clear path to the factory network. Modern condition-monitoring technology uses terminal processing to filter raw data on the factory floor. This local step compresses large files before sending summaries to cloud storage.
The final data package must link into the main facility CMMS. Connecting these systems ensures that an alert automatically opens a maintenance work ticket. Reliability teams synchronising their inventory planning with EU Automation can secure items with short lead times before the machine requires intervention.
Conclusion
Choosing the right sensors when applying condition-based monitoring effectively catches faults early. It is important to know that no single sensor type covers every possible failure mode. This means that a combination of different sensors is the best option.
Plant leaders should feed the sensor data directly into the work order system. This ensures that when a sensor identifies a potential issue, the team is already equipped with the data they need to act before a failure happens.
EU Automation provides sensing components and industrial hardware for teams building multi-sensor condition monitoring systems. This helps with the continuous operation and maintenance of critical tracking systems.
References
[1] Smith, R. (2019). Improving maintenance by adopting a P-F curve methodology. InTech, ISA.
