Anomaly detection on equipment vibration telemetry that flags machines drifting out of normal behavior before they fail.
Type
Data science
Year
2026
Stack
Kusto / KQL, Azure Data Explorer, Python, Statistics, Scheduled jobs
Live
Private industry system
Illustrative vibration signal with 3σ bounds · not real data
This is industry work described in general terms. Employer names, data, and screenshots are intentionally left out, and the chart above is illustrative.
01The problem
Maintenance was reactive. By the time a machine sounded wrong, the repair was already expensive and production had already stopped.
02What I built
Statistical control bounds (3-sigma) learned from each machine’s own normal vibration behavior
Queries on Azure Data Explorer (KQL) that scan high-frequency telemetry efficiently
Scheduled jobs that score new readings and surface out-of-bounds equipment
03The result
Maintenance teams see which machines are drifting, not just which ones already broke
Simple, explainable statistics that technicians can trust and verify
Need something like this?
Tell me about the problem. I reply within one business day.