How VibeSentry Works
From sensor to prediction in four simple steps.
Sensor Captures Vibration Data
Each sensor puck contains an industrial-grade MEMS accelerometer that captures tri-axis vibration data from your appliance motor. On-chip processing runs a Fast Fourier Transform (FFT) to extract a compact feature vector — the vibration "fingerprint" of your appliance at that moment.
The sensor transmits this 100-byte fingerprint every 15 minutes over Zigbee 3.0, using barely any battery.
Hub Runs Local Analysis
The VibeSentry hub receives data from all your sensors and runs local anomaly detection. It stores 12 months of vibration history and can provide basic alerts even without a cloud connection.
For Pro subscribers, the hub uploads data to the cloud for fleet-scale analysis.
Cloud Matches Against the Fleet
This is where VibeSentry's real power lives. Your appliance's degradation trajectory is compared against thousands of other appliances of the same type. Using dynamic time warping and Bayesian probability modeling, we match your trajectory to known failure patterns.
When your washer's vibration trajectory matches 47 other washers that had bearing failures 3 months later — we know what's coming.
You Get Actionable Predictions
The VibeSentry app tells you exactly what's happening:
Estimated remaining life: 2–4 months of normal operation
Estimated repair cost: $280
Schedule service?
No vague "something might be wrong" alerts. Specific failure modes, timelines, and costs — so you can plan repairs on your schedule, not in an emergency.
The Technology Behind It
MEMS Accelerometers
Industrial-grade tri-axis vibration sensing at consumer prices. The same core technology Augury uses for $2,000+/machine industrial monitoring — now in a $13 sensor puck.
Zigbee 3.0 Mesh
Ultra-low-power wireless protocol with mesh networking. Proven in millions of smart home devices. Sensors relay through each other for whole-home coverage.
Edge FFT Processing
On-chip Fast Fourier Transform extracts vibration features at the sensor, sending only a compact fingerprint — not raw data. This is what enables 2+ year battery life.
Fleet Learning
Every confirmed failure labels the preceding trajectory, making predictions better for every future user. The more sensors deployed, the smarter the system gets.