Smart Automation: Sensor-Driven Anomaly Detection & Auto-Response

Fault Injection

Detection & Response

Anomaly sensitivity (z-threshold) 2.5σ

Sim speed 1.0×

Automated response
off = alerts only, manual throttle

Fleet Status

Uptime

100%

Anomalies caught

Avg |z-score|

0.00

In maintenance

0 / 6

How it works

Each of the 6 stations streams a noisy sensor reading x(t) (simulated vibration/temperature). A lightweight online analytics engine tracks a running mean and variance with an exponential moving average (EMA) — the same technique real edge-analytics pipelines use because it needs no stored history:

μ ← μ + α(x − μ)
σ² ← (1 − α)·(σ² + α(x − μ)²)
z = (x − μ) / √(σ² + ε)

|z| is the anomaly score: how many standard deviations the current reading sits from the station's own recent normal. When |z| crosses the sensitivity threshold, the automated controller intervenes in a closed loop, without waiting for a human:

Throttle — cuts the station's speed (and therefore mechanical load/wear) in proportion to how far past the threshold z has gone, scaled by the response-aggressiveness of the automation.
Escalate to maintenance — if throttling isn't enough and health keeps falling, the station is pulled offline for repair, then returns to service at full health with a fresh baseline.

Turn automated response off to see the difference: alerts still fire, but nothing intervenes, so injected faults run unchecked to failure — the same before/after an SRE team sees when they enable auto-remediation on a real fleet.

Normal Throttled (anomaly) Critical In maintenance

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.