Yes. For industrial HVAC, I’d build this as a condition-monitoring + predictive-maintenance platform, rather than simply putting vibration alarms on motors.
ASHRAE describes predictive maintenance as using measured equipment condition/performance and historical benchmarks to determine when maintenance is actually needed; vibration monitoring is one of the established techniques.
Designing a predictive maintenance (PdM) solution for industrial HVAC systems (such as large centrifugal chillers, cooling tower fans, air handling unit [AHU] blowers, and large compressors) using vibration analysis requires a robust 4-tier architecture: Sensors → Edge/Gateway → Cloud/Platform → Actionable Insights.
What are some predictive maintenance tools for industrial HVAC units that rely on vibration sensing?
What vibration sensor systems are available for predictive maintenance on industrial HVAC units?
Don't rely solely on a single RMS vibration number. ISO 20816 specifically distinguishes broadband severity from changes in vibration magnitude, and notes that acceleration can be particularly useful for higher-frequency diagnostic information.
Belt problems, pulley misalignment, mechanical looseness
Vibration is particularly appropriate for rotating HVAC equipment; industrial condition-monitoring systems commonly target motors, pumps, compressors, fans, blowers and gearboxes.
The important part: don't build the ML model first
I'd initially establish a normal operating fingerprint for every asset.
For example:
AHU-17 Supply Fan
Normal:
RPM: 1,780
RMS velocity: 2.1 mm/s
Bearing temp: 47°C
1× RPM amplitude: normal
Bearing envelope: normal
Current:
RMS velocity: 3.4 mm/s ↑ 62%
Bearing temp: 54°C ↑
1× RPM: stable
High-frequency: ↑↑
Diagnosis:
Possible bearing degradation
Trend:
degradation rate increasing
Action:
Inspect bearing during next planned outage
This approach is already used commercially: Banner, for example, describes establishing a machine-specific vibration baseline and automatically deriving warning/alarm thresholds, while other systems use AI to identify abnormal machine patterns and component-level faults.
3. Use operating context
This is critical for HVAC.
A fan running at 100% speed during a hot afternoon shouldn't be compared directly with the same fan operating at 35% speed.
Your model should therefore ingest:
VFD frequency/RPM
Motor current
Supply/return-air temperature
Outdoor temperature
Static pressure
Chilled-water temperature
Differential pressure
Valve/damper position
Operating state
Start/stop events
BMS alarms
That lets you calculate something closer to:
"Is vibration abnormal for this machine under these operating conditions?"
rather than simply:
"Is vibration above 3 mm/s?"
That's a major difference between a useful PdM system and an alarm generator.
4. Analytics stack
I'd use three layers.
Layer A — deterministic condition monitoring
Start with engineering features:
RMS velocity
RMS acceleration
Peak
Crest factor
Kurtosis
Temperature
FFT peaks
Bearing-envelope energy
1× / 2× rotational frequency
Harmonics
This provides explainable alerts.
Layer B — anomaly detection
Build an individual baseline for each machine.
Possible models:
EWMA/CUSUM
Isolation Forest
One-Class SVM
Autoencoder
PCA
Gaussian/robust statistical models
For a first deployment, I'd favor statistical baselines + engineered vibration features over immediately deploying a deep-learning model.
Layer C — fault classification
Once you accumulate labeled failures, train models to distinguish:
Normal
│
├── Imbalance
├── Misalignment
├── Mechanical looseness
├── Bearing degradation
├── Belt problem
├── Cavitation
└── Other anomaly
Commercial systems already follow this general pattern. Honeywell, for example, describes an architecture of anomaly detection → fault diagnosis → time-to-maintenance prediction → action.
Sensor/platform options worth evaluating
You don't necessarily need to build the hardware yourself.
Fluke/Azima is interesting if you want sophisticated vibration diagnostics and expert analysis. Its current wireless system uses triaxial vibration, temperature and AI-based diagnostics.
Banner Engineering is attractive for an industrial automation/IIoT approach, with wireless vibration/temperature sensors and gateways capable of local connectivity and machine-learning-based baseline monitoring.
KCF Technologies offers a purpose-built wireless vibration platform with sensors, gateways and predictive-maintenance analytics.
Honeywell Forge Performance+ for Buildings is worth considering if the HVAC environment already uses Honeywell/BMS infrastructure; its predictive-maintenance offering specifically targets building equipment and incorporates vibration sensing.
Acoem is another option if vibration analysis is central to your maintenance program, with wireless sensors, cloud/on-premise software and AI-oriented diagnostics.
There are also more open/DIY-oriented systems. For example, NCD's current industrial sensor combines 3-axis vibration/FFT with temperature and motor current and can forward data to systems such as MQTT/AWS/Azure.
What I'd recommend building
If this is a new product/project, I'd use:
Sensor
→ 3-axis accelerometer + temperature
Communications
→ LoRaWAN or industrial Wi-Fi, depending on plant conditions
Edge gateway
→ Linux industrial PC
Protocol
→ MQTT
Data
→ Time-series database
Analytics
→ Python + engineered vibration features + anomaly models
Dashboard
→ Grafana/custom web application
Integration
→ BACnet/Modbus/OPC UA → BMS
→ REST/API → CMMS
Alerts
→ Email/SMS/Teams + CMMS work order
A particularly good MVP
Don't start with every HVAC component.
Pilot on 20–50 rotating assets:
10 AHU/supply fans
5 pumps
5 cooling-tower fans
5 compressors
5 miscellaneous motors
Run it for 3–6 months, establish baselines, deliberately capture maintenance events, and correlate vibration changes with actual failures.
Then measure:
false alarms/asset/month
detection lead time
failures detected before conventional maintenance
maintenance hours avoided
unplanned downtime avoided
sensor uptime
percentage of alerts that resulted in confirmed faults
That gives you the evidence needed to decide whether sophisticated ML is actually improving the system.
One standards caveat: ISO 20816-3 is useful for industrial machinery including motors, rotary compressors and fans, but its applicability depends on machine type, size, support and operating conditions. It shouldn't be treated as a universal HVAC alarm-threshold table.
If you tell me (1) the HVAC equipment—e.g. chillers/AHUs/cooling towers, (2) approximate number of machines, and (3) whether you want a commercial solution or to build the system yourself, I can turn this into a concrete sensor → gateway → cloud → ML architecture with specific hardware, protocols, database schema, and estimated per-asset cost.
Target Assets: Focus on rotating equipment—motor bearings, fan shafts, gearboxes, and compressor housings.
Sensor Choice: Deploy triaxial MEMS or piezoelectric accelerometer vibration sensors. They should measure overall Velocity (RMS in mm/s or in/s for general unbalance/misalignment) and High-Frequency Acceleration (for bearing faults, gear mesh issues, and cavitation).
Connectivity: Use wireless industrial protocols (e.g., Wi-Fi, BLE, LoRaWAN, or WirelessHART) to avoid costly cable runs across existing ductwork and plant architecture. Hardware like the NCD Industrial IoT Wireless Gen4 Sensor offers integrated FFT (Fast Fourier Transform) vibration and temperature monitoring.[1]
2. Data Acquisition & Edge Processing
Sampling Rate & FFT: Raw time-waveform data is heavy. Configure the local gateway or smart sensor to run onboard FFT to convert raw vibration data into frequency spectrums. This isolates specific fault frequencies (e.g., inner/outer race bearing frequencies, 1X/2X RPM unbalance peaks).
Transmission Strategy: Instead of streaming continuous raw data 24/7 (which chokes bandwidth and storage), program nodes to transmit aggregated metrics (RMS, peak-to-peak, kurtosis) hourly, and trigger high-resolution spectrum transmission only when a vibration threshold is breached.
3. Analytics & Platform Layer
Baseline Establishment: Run the HVAC equipment for 1–2 weeks under normal operating loads to map out a "healthy" operational baseline, adjusting for variable frequency drive (VFD) speed changes and seasonal ambient loads.
Diagnostic Rules Engine: Apply ISO 10816/20816 standards for mechanical vibration severity as an initial guardrail. Combine this with machine learning anomaly detection to flag shifts in spectral energy distribution.
Root-Cause Mapping:
1X RPM Peak → Mechanical unbalance (e.g., dust build-up on fan blades).
2X RPM Peak → Angular or parallel misalignment.
High-frequency non-synchronous peaks → Rolling element bearing degradation.
Sub-synchronous peaks → Fluid/airflow turbulence or belt resonance.
4. Integration & Action Workflow
CMMS/BMS Integration: Tie the platform via API (REST/MQTT) into your Computerized Maintenance Management System (e.g., Fiix, IBM Maximo) or Building Management System (BMS). When an anomaly score crosses a critical warning level for >48 hours, automatically generate a prioritized work order with the predicted failure mode attached.
To help tailor this architecture, tell me:
What are the specific HVAC assets you are targeting (e.g., rooftop AHUs, cooling towers, centrifugal chillers)?
Do you prefer a cloud-hosted SaaS platform or an on-premises/local PLC integration for data security?
Vibration sensor systems for industrial HVAC predictive maintenance
For industrial HVAC equipment, vibration monitoring can help identify developing failures in supply and return fans, air-handling units (AHUs), exhaust fans, cooling-tower motors, pumps, compressors, and bearings before they cause unplanned downtime.
The main systems available fall into three categories: wireless sensors with automated analytics, permanently installed online monitoring systems, and portable vibration analyzers. For most HVAC facilities, wireless sensors are a practical starting point because they can be installed on existing rotating equipment without extensive wiring. Critical chillers and large fan systems may justify more advanced continuous monitoring.
Leading systems to evaluate
1. ifm VVB3 — best for integration with building or plant controls
(HVAC-focused option)
A compact, wired three-axis sensor that measures vibration and surface temperature. Designed for applications including HVAC fans, motors, pumps and belt drives. Supports condition indicators for bearing wear, imbalance, friction and looseness. IO-Link integration makes it attractive when you want to connect monitoring to a PLC or industrial automation system.
Best fit: facilities with controls engineers and an existing PLC/BMS integration strategy.
2. Emerson AMS Wireless Vibration Monitor — best for scalable wireless monitoring
A wireless system that measures vibration in three axes, temperature and other machine-health indicators. High-resolution waveforms, spectral analysis and PeakVue Plus analytics help diagnose bearing and lubrication problems. Uses WirelessHART and integrates with Emerson's machinery-health software.
Best fit: large campuses, multiple AHUs, remote mechanical rooms and facilities where adding signal cable is expensive.
3. Fluke Reliability Azima Accel 310 — best for automated diagnostics
Wireless, triaxial vibration sensing paired with WATCHMAN analytics and diagnostic support from vibration specialists. The system is designed to provide component-level fault identification and prioritized repair recommendations. The manufacturer specifies approximately three years of battery life.
Best fit: maintenance teams that want actionable diagnoses without having to employ a dedicated vibration analyst.
4. Augury Machine Health — best for a managed predictive-maintenance program
Combines continuous sensing, AI-based diagnostics and expert validation. Its stated equipment coverage includes motors, fans, pumps, compressors and cooling towers. It is more of an end-to-end machine-health service than a standalone sensor purchase.
Best fit: organizations seeking maintenance recommendations and expert support across many assets.
5. SKF Enlight Collect IMx-1 — worth evaluating for bearing and rotating-equipment monitoring
A wireless condition-monitoring option within SKF's machinery reliability ecosystem. Consider it when the goal is to track vibration trends and identify developing mechanical faults across motors, fans and pumps. Verify the current sensor/software configuration and HVAC-specific diagnostic coverage with SKF before selecting it.
Low-complexity monitoring of standard motors and fans
ifm VVB3
Wireless retrofit across many HVAC assets
Emerson AMS Wireless
Automated fault diagnosis with specialist support
Fluke Azima Accel 310
AI-driven, managed predictive maintenance
Augury
Existing SKF condition-monitoring ecosystem
SKF Enlight Collect
These are shortlist recommendations based on published capabilities, not a verified head-to-head performance ranking. Pricing, sensor counts, gateways, software subscriptions and integration costs need vendor quotes.
What should you monitor?
AHU and exhaust fans
Monitor the motor and fan bearings. Detect imbalance, belt wear, misalignment, looseness and developing bearing defects.
Chilled-water and condenser-water pumps
Track motor and pump vibration to help identify bearing wear, coupling problems, looseness and possible cavitation.
Chillers and compressors
Use a more capable diagnostic system where downtime is costly. Check sensor bandwidth, sampling strategy and compatibility with the compressor type.
Cooling-tower fans
Monitor motor and gearbox bearings, shaft alignment, imbalance and mechanical looseness, accounting for outdoor exposure and variable-speed operation.
Important selection criteria
Before purchasing, confirm these five items:
Diagnostic depth: Does the system provide only overall vibration thresholds, or can it analyze frequency spectra and waveforms to identify specific faults?
Sampling strategy: Can it capture useful data during changing fan speeds, VFD operation and variable loads?
Connectivity: Does it integrate with your existing BMS, such as BACnet or Modbus, or with your PLC, CMMS or maintenance platform? Do not assume a sensor's native protocol supports direct BMS integration.
Installation: Can sensors be securely mounted at the motor and fan bearings? Are enclosures, temperature ratings and wireless connectivity suitable for the actual environment?
Alarm workflow: Will the system provide useful fault severity, trends and recommended actions—not just a vibration threshold alarm?
My initial recommendation: For standard industrial HVAC motors, pumps and fans, compare the ifm VVB3 for wired integration with Emerson AMS Wireless for wireless coverage. Add Fluke Azima or Augury to the evaluation if automated diagnostics and expert recommendations are a priority.
