The IIoT monitoring architecture
Each layer has a job:
- Sensor node: the transducer, a small processor, a radio and a battery. Many nodes compute overall values and simple features on board and only send raw waveforms on a schedule or on alarm.
- Gateway: receives radio messages from many nodes and forwards them over Ethernet, Wi-Fi or cellular. Some networks are meshes where nodes relay for one another.
- Edge computing: an industrial PC or the gateway itself calculates spectra, features and anomaly scores near the machine, cutting data volume and latency.
- Platform: stores time series and waveforms, runs analytics and models, shows dashboards, and manages alarms and users.
- Integration: findings go to the maintenance management system as notifications or work orders, and operating context (speed, load, product) comes back from control systems and historians. Open protocols such as MQTT and OPC UA keep these interfaces vendor-neutral.
Wired or wireless?
| Consideration | Wired online | Wireless online | Portable route |
|---|---|---|---|
| Typical use | Machinery protection and critical assets | Balance-of-plant, remote and hard-to-reach assets | Accessible, lower-tier assets; analyst diagnostics |
| Data rate | Continuous, full bandwidth | Periodic: minutes to hours for overall values, hours to days for waveforms | Monthly or quarterly |
| Automatic trip | Yes | No (not designed for protection) | No |
| Installed cost per point | High (cabling, junction boxes, racks) | Low to moderate | Lowest hardware cost, highest labour |
| Power | From the system | Battery (years) or energy harvesting | Instrument battery |
| Data quality | Best | Good with correct sensor and mount; limited by bandwidth and battery | Good, but depends on the technician and mount |
| Safety | No exposure after installation | No exposure after installation | Technician near running machinery |
Choosing a wireless network
No single wireless technology is best. The choice is a trade-off between range, data rate, power and how much infrastructure you want to own.
| Technology | Range (indicative) | Data rate | Power | Good for | Watch out |
|---|---|---|---|---|---|
| Bluetooth Low Energy (BLE), including mesh | 10–100 m | Moderate | Very low | Dense clusters of sensors, phone-based commissioning | Short range in metal-rich plants; needs many gateways or mesh |
| Industrial mesh (IEEE 802.15.4 based, e.g. WirelessHART, ISA100.11a) | 50–200 m per hop | Low to moderate | Low | Process plants; self-healing networks; existing industrial wireless infrastructure | Limited throughput for full waveforms |
| Proprietary sub-GHz (868/915 MHz) | 100 m to 1 km+ | Low to moderate | Low | Good penetration in plants, fewer gateways | Ties you to one supplier's ecosystem |
| LoRaWAN and similar LPWAN | 1–10 km | Very low (tens to hundreds of bytes per message) | Very low | Large sites, remote assets; overall values and edge features | Not suited to raw waveforms; duty-cycle limits |
| Wi-Fi | 30–100 m | High | High | Mains-powered sensors, full waveforms, existing networks | Battery life; IT network policies |
| Cellular (LTE-M, NB-IoT, 5G including private 5G) | Wide area | Low to high | Low to moderate | Distributed assets (pipelines, pumping stations, wind), no plant network needed | Subscription costs; coverage indoors and underground |
Before committing, run a radio site survey. Steel structures, tanks, concrete and moving equipment block and reflect signals. Plan gateway positions with line of sight to as many sensors as possible, mount antennas high, and allow margin: a link that works on installation day may fail when scaffolding or a new tank appears.
Battery life and data richness
Most battery-powered sensors run for several years, but only if the measurement schedule is sensible. Taking and transmitting a raw waveform costs far more energy than sending an overall value. The illustration below uses a simple energy model to show the trade-off.
The model is illustrative only: real battery life depends on the sensor's electronics, radio conditions, temperature (cold reduces capacity) and firmware. Use the supplier's battery calculator for your actual schedule, and ask how battery life was tested.
Practical schedules
- Overall values and temperature every 10–60 minutes for trending and alarms.
- Spectra or features several times a day.
- Full raw waveforms once or twice a day, plus on alarm or on request, for diagnosis and machine learning.
- Trigger on state: only measure when the machine is running, ideally within a defined speed and load window.
- Energy harvesting (vibration, thermal or solar) can extend life in suitable locations; check the harvested power against your schedule.
Edge analytics
Sending every raw waveform to the cloud is expensive in energy, bandwidth and storage. Edge processing calculates features (RMS, peak, crest factor, kurtosis, band energies, envelope spectra) near the machine and sends those instead. A good design keeps the ability to request raw data when a feature moves, because analysts and machine learning models both need waveforms to confirm and learn. See data and feature engineering and edge AI.
Beyond vibration: other wireless sensors
| Sensor | What it adds | Notes |
|---|---|---|
| Surface temperature (often built into vibration nodes) | Late-stage safety net, context for vibration | Measures housing, not bearing, temperature |
| Clamp-on current sensors | Load, running hours, motor current signature analysis | Installed in the MCC; check sampling rate and record length for rotor bar analysis |
| Contact ultrasound | Lubrication condition and slow-speed bearings | Useful with automatic lubricators to confirm grease delivery |
| Online oil sensors | Particle count, water, dielectric, wear debris | Large gearboxes, hydraulics, turbines |
| Fixed thermal cameras | Continuous thermography of critical panels, kilns, conveyors | Mains-powered; often wired or Wi-Fi |
| Pressure, flow, speed | Operating context, performance monitoring | Often already available from control systems |
Cybersecurity
Every wireless sensor and gateway is a network device in your operational technology environment. Treat it that way:
- Follow the IEC 62443 series for industrial automation security: zones and conduits, least privilege, defence in depth.
- Keep monitoring networks segmented from control networks; data should flow out of the control zone, not into it.
- Require encryption on the radio link and in transit, unique device credentials, and signed firmware updates.
- Plan how firmware and certificates will be updated over the sensor's lifetime.
- Agree with IT where data is stored, who can access it, and how long it is kept.
Hazardous areas
Many pumps, compressors and fans sit in areas where flammable gas, vapour or dust may be present. Sensors there must be certified for the zone or division, typically as intrinsically safe equipment, under schemes such as IECEx and ATEX (zones 0, 1, 2 for gas; 20, 21, 22 for dust) or the North American class and division system. Check the certification, gas group and temperature class of the whole node, including the battery, which may need to be replaced only outside the hazardous area or with a certified battery pack.
Questions to ask before you choose a system
- What are the sensor's bandwidth, noise floor, range and mounting options? Is the data sheet specific?
- Can I get raw waveforms, at what Fmax and lines, how often, and can I export them in an open format?
- What battery life is achievable with my schedule, at my site temperatures?
- Which wireless technology is used, and how many gateways will my site need?
- How does the system integrate with my maintenance management system and historians (open APIs, MQTT, OPC UA)?
- What security certifications and update processes does it have?
- Is it certified for my hazardous areas?
- Who owns the data, and can I take it with me if I change systems?
Key takeaways
- Wired systems protect critical machines; wireless systems monitor the balance of plant economically.
- Choose the wireless technology for range, data rate and power, and run a radio site survey.
- Raw waveforms dominate battery use: send features often, waveforms less often or on alarm.
- Treat every sensor as an OT network device: segment, encrypt and plan updates.
- In hazardous areas, the whole sensor node, including the battery, must be certified.