What’s included:
- real bearing vibration signals (healthy + inner/outer race faults)
- spectrum peak detection + envelope analysis
- ISO 20816-3 severity assessment
- (baseline) anomaly detection like OneClassSVM/LOF
- local HTML reports under reports/
Looking for feedback on: MCP tool design, missing diagnostics to prioritize next, and whether the combo LLM with deterministic tooling is useful in real PdM workflows.
Thanks, that’s exactly the intent. Automotive would be interesting, but it would need real-time constraints. Happy to get feedback on what signals would be most valuable
Hi HN, I built an open-source MCP server for vibration-based predictive maintenance.
The idea: the LLM handles conversation and orchestration; the server runs deterministic vibration analysis and writes an interactive local report.
Repo: https://github.com/LGDiMaggio/predictive-maintenance-mcp
Quick try: - git clone https://github.com/LGDiMaggio/predictive-maintenance-mcp.git - cd predictive-maintenance-mcp - python setup_venv.py - python validate_server.py
What’s included: - real bearing vibration signals (healthy + inner/outer race faults) - spectrum peak detection + envelope analysis - ISO 20816-3 severity assessment - (baseline) anomaly detection like OneClassSVM/LOF - local HTML reports under reports/
Looking for feedback on: MCP tool design, missing diagnostics to prioritize next, and whether the combo LLM with deterministic tooling is useful in real PdM workflows.
This is probably the AI we need in cars instead of the one that will tell you Taylor Swift’s shoe size.
Thanks, that’s exactly the intent. Automotive would be interesting, but it would need real-time constraints. Happy to get feedback on what signals would be most valuable