Industrial Edge AI
XyloZero
XyloZero is an edge AI solution that analyzes sensor signals together with operating data to detect equipment anomalies early, on site.
Operators use the selected anomaly events to set inspection priorities and decide when to schedule maintenance.
Field challenges
Equipment faults often surface only after the machine stops. A louder sound or a new vibration is not an anomaly by itself; start-ups, stops and load changes have to be read alongside the signal.
How it works
Sensor nodes attach to equipment magnetically or with bolts, collect signals and link them with PLC and drive data. XyloZero picks the intervals worth analyzing by operating state and selects anomaly events at the edge. Results go to dashboards and alerts, and set inspection priorities.
Equipment
- Motors
- Bearings
- Gearboxes
- Conveyors
- Pumps
- Fans
- Compressors
- Cranes
- Turbines
- Generators
Device
A single sensor node houses microphones, motion and environmental sensors, a processor and a battery.
- Top cover
- Sealing gasket
- MEMS microphones
- Processing module
- Main board
- Battery
- Housing
- Mounting magnet
01
Multimodal sensing
MEMS microphones, a 6-axis IMU, a magnetometer and temperature, humidity, pressure, light, CO2, VOC and NOx sensors share one board, so sound, vibration, rotation and environmental change land on the same timeline.
- Sound: 4 MEMS microphones
- Vibration and motion: 6-axis IMU
- Rotation and magnetic field: 3-axis magnetometer
- Environment: temperature, humidity, pressure, light, CO2, VOC, NOx
- Top cover
- Sealing gasket
- MEMS microphones
- Processing module
- Main board
- Battery
- Housing
- Mounting magnet
- Screws
02
Context Gating
PLC and drive data separate running, stopped, loaded and transitional states, and XyloZero evaluates only the intervals worth analyzing. Signals from transients such as start-up and shutdown are less likely to be read as anomalies.
- Running, stopped, loaded and transitional states
- Valid intervals selected for analysis
- Less noise and fewer false alarms
03
XAP and XMBP data path
Sound is compressed with XAP (Xylolabs Audio Protocol), an LC3-based audio codec, and sensor readings are packed with their timestamps into XMBP (Xylolabs Metadata Binary Protocol) batches. Both travel to the server in one HTTPS session; compression and batching keep uploads small on low-bandwidth site links.
- XAP: LC3-based audio compression
- XMBP: batched sensor readings
- HTTPS session uploads and OTA updates
From collection to response
Signal collection
Sensor signals, PLC data and drive data are collected on site.
Operating context
Operating state and site conditions are analyzed on the same timeline as the signals.
Event selection
Edge AI selects the anomaly events that matter.
Response
Results go to the control room and alerts, and set inspection priorities for follow-up.
Key capabilities
- Multimodal sensing: sound, vibration, temperature and humidity, magnetic field
- PLC and drive data integration
- Context Gating by operating state
- Edge AI inference
- Anomaly event selection
- Dashboard, alert and maintenance system integration
Deployment goals
- Early anomaly detection
- Inspection priorities
- Maintenance timing decisions
- Fewer unplanned stops and safety risks
Key specifications
| Sound | 4 MEMS microphones, 96 kHz 4-channel capture |
|---|---|
| Vibration and motion | 6-axis IMU (accelerometer, gyroscope) |
| Magnetic field | 3-axis magnetometer |
| Environment | Temperature, humidity, pressure, light, CO2, VOC and NOx |
| Power status | Voltage and current monitoring |
| Processor | Microcontroller-based processing module |
|---|---|
| Audio codec | XAP (LC3-based) |
| Data format | XMBP binary batches |
| Connectivity | Wi-Fi 2.4 GHz, HTTPS (TLS 1.3), LTE via an external router |
| Equipment integration | Industrial protocols such as CAN and EtherCAT |
| Storage | Local archive on microSD |
| Updates | Over-the-air updates |
| Power | USB-C, 1-cell lithium-ion battery charging |
|---|---|
| Battery | 10,000 mAh |
| Battery life | 1 year (varies with operating conditions) |
| Mounting | Magnet or bolts |
| Explosion-proof certification | Optional |
Frequently asked questions
- Which signals does XyloZero use?
- Sound, vibration, temperature and humidity, magnetic field and other sensor signals, together with PLC and drive data.
- Which equipment can we start with?
- We start by reviewing equipment whose signals can be collected, such as motors, bearings, gearboxes, conveyors, cranes, pumps, fans and compressors.
- Do we need to replace our PLC or control system?
- No. We check the site interfaces, take data from your existing PLCs and drives, and connect to your control-room, alert and maintenance systems.