LLM Edge Predictive Maintenance: Bridging Industrial IoT Sensors and Large Language Models for Predictive Maintenance

Open-source reference architecture connecting industrial edge sensor hardware (STEVAL-STWINBX1 / STWIN.box) to LLM-based diagnostic assistants (Claude) via the Model Context Protocol (MCP). Includes MCP servers for sensor acquisition and vibration signal processing (FFT, envelope analysis, bearing fault detection, ISO 10816), plus Claude Skills for guided condition monitoring, fault diagnosis, and operator-friendly report generation. This is an independent open-source project, not affiliated with or endorsed by Anthropic or STMicroelectronics.