Portable Predictive Maintenance Assistant
A portable industrial maintenance platform built around Raspberry Pi 5 and AI HAT+ 2. The system combines thermal imaging, vibration monitoring, maintenance knowledge retrieval, AI-assisted troubleshooting, and field documentation into a single handheld tool intended for predictive maintenance of motors, bearings, gearboxes, conveyors, and other mechanical assets.
Problem
Industrial maintenance technicians often carry multiple tools for thermal inspections, vibration measurements, troubleshooting, documentation lookup, and note taking. Existing tools can be expensive, disconnected from one another, and often lack onboard intelligence for analysis and recommendations.
Solution
Create a portable edge-AI device that combines:
- Thermal imaging
- Vibration analysis
- Acoustic monitoring
- Maintenance documentation retrieval
- AI-powered troubleshooting
- Meeting and conversation summarization
The device is designed to operate locally using a Raspberry Pi access point without requiring enterprise Wi-Fi access.
User Experience
A technician:
- Powers on the device.
- Connects directly to the Raspberry Pi access point.
- Opens a local dashboard.
- Captures thermal scans.
- Records vibration measurements.
- Receives AI-assisted maintenance insights.
- Searches repair instructions and SOPs.
- Documents findings and work order notes.
Future versions may include a dedicated mobile application.
Current Status
In active prototype development.
Thermal imaging hardware has been acquired and successfully connected. Initial thermal frame acquisition from the FLIR Lepton 3.5 through the PureThermal Mini Pro has been validated.
Vibration sensing hardware selection has been completed but full integration remains in progress.
Tech Stack
Hardware
- Raspberry Pi 5
- Raspberry Pi AI HAT+ 2
- FLIR Lepton 3.5
- PureThermal Mini Pro JST-SR
- ADXL357 Evaluation Board
- Future microphone integration
Software
- Linux
- Python
- OpenCV
- FFmpeg
- V4L2
- Local Web UI
- LLM/RAG stack (planned)
Links
GitHub: TBD
Demo: TBD
Documentation: TBD

