Predict
Dual-MCU edge monitoring for rotating machines
Predict is an industrial IoT sensor designed to monitor motors, pumps, and fans through continuous multi-physics observation. Its architecture separates network and power management from acquisition and edge analytics so that the device can learn normal behavior on-site, quantify anomalies locally, and only ship the right payload through the right transport.
Dataflow
Product Context
Industrial monitoring context
Predict targets rotating assets where failures often announce themselves through subtle changes in vibration, acoustics, temperature, or magnetic behavior. Instead of transmitting raw measurements continuously, the product emphasizes embedded interpretation so it can stay efficient, autonomous, and usable at scale.
How the Embedded Learning Loop Works
Installation and context setup
- The machine and installation context are known, but no baseline exists yet.
- The objective is to position the sensor, validate the environment, and prepare the first acquisition routines.
- Autonomous behavior modeling has not started at this stage.
Electronic, Data, and Software Architecture
Electronic & Embedded Stack
A dual-MCU design separates networking from sensing and edge processing.
- A master controller manages wake-up, energy, and wireless links.
- A dedicated STM32 handles acquisition and local analytics.
Connectivity & Data Path
The communication path changes with payload size and usage mode.
- LoRaWAN sends lightweight health indicators and alerts.
- BLE is used for local diagnostics, live measurements, and raw data retrieval.
Mobile & Cloud Layer
The smartphone is both a field interface and a bridge to backend services.
- The app supports pairing, diagnostics, and heavy-data relay.
- Backend services separate databases, storage, and cloud processing.