Industrial IoT is only as good as its sensors. Yet sensor selection is often an afterthought—delegated to procurement or defaulted to whatever the integrator recommends. This guide provides a systematic approach to choosing sensors that actually deliver the data you need, in the environment where you need it.
Start with the Question, Not the Sensor
Before evaluating sensors, be clear about what you're trying to learn:
- What decision will this data inform?
- What accuracy do you actually need?
- How fast do things change?
- What are the consequences of bad data?
A temperature sensor for comfort monitoring has very different requirements than one for process control. An over-specified sensor wastes money; an under-specified one provides useless data.
Key Sensor Categories
Vibration Sensors
The workhorse of condition monitoring. Vibration reveals bearing wear, imbalance, misalignment, and mechanical looseness.
Types:
- Accelerometers: Most common. Measure acceleration, derive velocity and displacement. Good for high frequencies.
- Velocity sensors: Direct velocity measurement. Better for lower frequencies typical of rotating machinery.
- Proximity probes: Measure shaft displacement directly. Used for journal bearings and critical machinery.
Key specifications:
- Frequency range: Match to expected fault frequencies. Bearings need high frequency (10-20kHz); imbalance is low frequency.
- Sensitivity: mV/g. Higher sensitivity for low-vibration equipment; lower for heavy machinery.
- Dynamic range: Must accommodate both normal operation and fault conditions.
- Mounting: Stud mount is best; magnetic acceptable for portable; adhesive for low-vibration permanent.
Temperature Sensors
Ubiquitous in industrial applications. Options vary widely in accuracy, response time, and robustness.
Types:
- Thermocouples: Wide temperature range, rugged, inexpensive. Lower accuracy. No external power needed.
- RTDs: High accuracy and stability. More expensive, slower response, requires excitation current.
- Thermistors: High sensitivity over narrow range. Fast response. Non-linear output.
- Infrared: Non-contact measurement. Good for moving or inaccessible targets. Affected by emissivity.
Key specifications:
- Range: Must cover expected conditions plus safety margin.
- Accuracy: Often specified at reference temperature; degrades elsewhere.
- Response time: Critical for process control; less important for trending.
- Stability: Drift over time. RTDs best; thermocouples vary by type.
Pressure Sensors
For hydraulics, pneumatics, process vessels, and flow inference.
Types:
- Piezoelectric: Dynamic pressure changes. Good for hydraulics and acoustics.
- Strain gauge: Static and dynamic. Most common industrial type.
- Capacitive: High accuracy for lower pressures.
Key specifications:
- Range: Include pressure spikes, not just operating range.
- Proof pressure: Survival rating above operating range.
- Media compatibility: Wetted materials must survive the process fluid.
- Output type: 4-20mA most common; voltage or digital also available.
Current and Power Sensors
For energy monitoring and motor health.
Types:
- Current transformers (CTs): Non-invasive, clamp-on installation. AC only.
- Hall effect: AC and DC measurement. Isolated.
- Shunt resistors: Direct measurement. Invasive installation.
- Power meters: Integrated measurement of voltage, current, power, energy.
Flow Sensors
For liquids and gases in pipes and ducts.
Types:
- Electromagnetic: Conductive liquids only. No pressure drop. High accuracy.
- Ultrasonic: Clamp-on possible. Good for large pipes. Affected by pipe condition.
- Coriolis: Mass flow. High accuracy. Expensive.
- Differential pressure: Proven technology. Creates pressure drop.
- Thermal: Good for gases. Measures mass flow directly.
Environmental Considerations
Industrial environments are harsh. Your sensors must survive.
Temperature
- Operating range vs. storage range
- Temperature cycling effects
- Electronics vs. sensing element limits
Moisture and Ingress
- IP rating: IP67 for washdown; IP68 for submersion
- Condensation inside enclosures
- Cable entry points (often the weak link)
Vibration and Shock
- Mounting hardware fatigue
- Connector reliability
- Electronic component survival
Electromagnetic Interference
- VFD-driven motors generate significant EMI
- Shielded cables and proper grounding essential
- 4-20mA more noise-immune than voltage outputs
Chemical Exposure
- Cleaning agents and process chemicals
- Lubricant compatibility
- Corrosive atmospheres
Connectivity Options
Wired
Advantages:
- Reliable, no battery concerns
- Higher data rates possible
- No RF interference
Disadvantages:
- Installation cost (especially retrofit)
- Cable routing in existing facilities
- Limited flexibility
Wireless
Advantages:
- Lower installation cost
- Flexibility for rotating or moving equipment
- Easier retrofit
Disadvantages:
- Battery life limitations
- RF interference in industrial environments
- Network infrastructure required
Wireless protocols:
- Wi-Fi: High bandwidth but power-hungry
- Bluetooth/BLE: Short range, lower power
- Zigbee: Mesh networking, low power
- LoRaWAN: Long range, very low power, low bandwidth
- WirelessHART: Industrial-grade, process applications
- Cellular (LTE-M, NB-IoT): Wide area, carrier-managed
Data Output and Interface
Analog Outputs
- 4-20mA: Industry standard. Noise-immune. Self-powered loop possible.
- 0-10V: Simple but more susceptible to noise.
- mV output: Requires signal conditioning near sensor.
Digital Outputs
- Modbus RTU/TCP: Universal support, simple protocol.
- HART: Superimposed on 4-20mA. Digital and analog together.
- IO-Link: Point-to-point smart sensor standard.
- CANbus: Common in mobile and automotive applications.
Common Selection Mistakes
Over-Specifying
Buying laboratory-grade accuracy for industrial trending. A $50 sensor often works as well as a $500 sensor for the actual application.
Ignoring Installation Costs
A cheap sensor requiring complex mounting or wiring can cost more than an expensive sensor with easier installation.
Neglecting Maintenance
Sensors in harsh environments need calibration and replacement. Budget for ongoing costs, not just initial purchase.
Forgetting About Cabling
Long cable runs affect signal quality. Factor in cable cost, installation labor, and signal conditioning requirements.
Single Sourcing
Relying on one vendor for a critical sensor creates supply chain risk. Identify alternatives early.
Ignoring Future Needs
Choosing a sensor that just meets today's requirements leaves no room for expanded monitoring.
The Selection Process
Step 1: Define Requirements
- What physical quantity?
- What range and accuracy?
- What sample rate?
- What environmental conditions?
Step 2: Survey Options
- Multiple vendor options
- Different technology types
- Wired vs. wireless
Step 3: Evaluate Total Cost
- Unit cost
- Installation materials and labor
- Infrastructure requirements
- Maintenance and calibration
- Replacement cycle
Step 4: Pilot Test
- Test in actual environment
- Verify data quality
- Assess installation challenges
- Evaluate maintenance requirements
Step 5: Standardize
- Reduce variety for easier maintenance
- Leverage volume purchasing
- Simplify spare parts inventory
The Platform Advantage
Sensor-agnostic platforms like Cohera decouple sensor selection from system architecture. Benefits:
- Best sensor for each application: Not limited to one vendor's ecosystem
- Mix and match: Different sensors on same platform
- Future flexibility: Swap sensors without changing software
- Avoid lock-in: Sensors and analytics are independent choices
The Bottom Line
Sensor selection deserves careful attention. The right sensor delivers reliable data that enables insights. The wrong sensor produces noise that undermines trust in the entire system.
Start with what you need to know, not what sensors are available. Match specifications to actual requirements—no more, no less. Consider the full lifecycle cost, not just purchase price. And pilot before committing.
The sensor is the foundation. Build on solid ground.