Data-Driven Automation: SPC Control Loop

Process Controls

Process drift 0.30

Sensor noise 0.40

Automation gain K_p 0.60

Spec tolerance ±3.0

Live Data Stream

oldestσ ≈ 0.00now

Current value

0.00

Rolling mean

0.00

Yield

100%

Auto-rejects

How it works

Each tick a hidden process value drifts randomly and is read through sensor noise, exactly like a real manufacturing measurement or streaming metric:

process[t] = process[t-1] + drift·N(0,1)·dt + correction
reading[t] = process[t] + noise·N(0,1)

The automation loop is a proportional feedback controller — it never sees the true process value, only the noisy readings, and continuously analyzes a rolling window of them:

error      = mean(last 30 readings) − target
correction = −K_p · error · dt      (applied only while Automation is ON)

A reading counts as out of spec when |reading| exceeds the tolerance band (the red dashed lines on the chart and the belt's inspection gate). With automation ON, the robotic arm removes those items and the controller's correction term pulls the drifting process back toward zero. Switch automation OFF to watch the same drift go uncorrected and unfiltered — the exact failure mode a real automated QA/monitoring system exists to prevent.

Drift — how fast the underlying process wanders off target.
Noise — measurement uncertainty added on top of the true value.
K_p — how aggressively the automated loop corrects drift from its running analysis.
Tolerance — the spec window; anything outside it is flagged for automated rejection.

In-spec item Out-of-spec item Inspection gate / arm

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