In a cone crusher running copper ore, the particle size distribution shifts every time the feed hardness changes — and it changes constantly. Fixed UCL/LCL control limits set during commissioning don't know that. They sit there generating false alarms when the ore gets harder and missing real drift when material gets softer. The result for operators: alarm fatigue, ignored signals, and a scrap rate that never improves no matter how closely the team watches the charts. Adaptive SPC limits fix this by doing what static limits cannot — recalculating UCL and LCL dynamically as the process baseline changes, applying Western Electric pattern rules against a moving reference, and surfacing root-cause ML explanations that tell operators exactly which variable crossed the line. Mining crushing teams running adaptive SPC are cutting scrap 30–50%. Here's how it works on the floor.
Why Static SPC Fails in Crushing Circuits
Traditional SPC was designed for controlled batch manufacturing — pharmaceutical filling lines, automotive stamping cells, semiconductor fab bays — where the process baseline is stable and control limits can be set once from historical data. A crushing circuit is none of those things. Feed ore arrives with variable hardness, moisture, and fragment size distribution that no blending program fully smooths out. Liner wear progresses continuously over weeks, shifting the gap and the particle size output even when the operator hasn't touched a setting. Recipe changes between ore types reset the expected PSD range entirely. Static UCL and LCL limits set at commissioning become meaningless within a shift. The real damage shows up in two failure modes operators know well: alarms that fire constantly when nothing is actually wrong (false positives that train teams to ignore charts), and real drift that slides under the fixed limits undetected until downstream screens report oversize or fines that should have been caught upstream.
How Adaptive SPC Limits Work: What Changes for the Operator
Adaptive SPC doesn't replace the Shewhart chart — it replaces the static UCL/LCL calculation with a dynamic model that updates as process conditions change. From the operator's perspective, the chart still looks the same. What changes is that the limits actually reflect the current process state rather than a months-old commissioning baseline.
Western Electric Rules in Crushing: What Each Rule Catches
Each Western Electric rule detects a different type of process instability. In a crushing circuit, different rules correspond to different real-world failure modes. This is the translation operators need to understand why the system is alerting and what action to take.
Static SPC vs Adaptive SPC: The Operator's Daily Experience
The technical difference between static and adaptive limits matters less to operators than the practical difference in what their shift looks like. Here's the same crushing circuit, same ore, same team — with and without adaptive SPC running.
| Situation | Static SPC | Adaptive SPC |
|---|---|---|
| Harder ore seam arrives mid-shift | Zone A alarm fires — operator checks, nothing wrong, resets, continues | Limits widen to reflect known harder-ore variance — no alarm unless a real exceedance occurs within that context |
| Liner wear shifts PSD over 3 weeks | Each individual reading within limits — drift undetected until downstream reports fines | Rule 4 (8 points same side) fires at day 7 — liner wear flagged as primary cause, work order generated |
| Recipe change — new ore blend | Old limits apply to new material — continuous false alarms or missed shifts depending on direction | Recipe change event triggers limit recalculation — new baseline established within first 30 readings |
| Alarm fires on operator screen | Red number — operator must investigate which of 5–6 variables caused it | "Liner wear: 62% contribution" displayed — one action, no investigation required |
| End of shift scrap report | 12% fines generated — source unclear, report filed, nothing changes | Fines rate 4–6% — two Rule 3 alerts caught the drift at hour 3, CSS adjusted before damage compounded |
| Management review of SPC data | Charts show constant alarms — team appears unresponsive; limits appear meaningless | Alarm rate down 47%, scrap trending down 30–50% — clear evidence SPC is working |
We had been running SPC on the secondary crusher for two years. The charts were always in alarm. The team had learned to ignore them completely — the SPC screen was just background noise on the workstation. When we switched to adaptive limits, alarm volume dropped by more than half in the first week. The team started looking at the charts again because suddenly the alarms actually meant something. By month three, fines generation was down 38%. That's what trust in a system looks like when the system is actually calibrated to your process.
— Process Control Engineer, Copper Concentrator — Tertiary Crushing Circuit, 3 cone crushersConclusion: The Limit That Adapts Is the Limit That Gets Respected
The 30–50% scrap reduction adaptive SPC delivers in mining crushing isn't from better operators or better ore — it's from control limits that mean something again. Dynamic UCL/LCL recalculation eliminates the false alarms that erode trust in SPC systems. Western Electric pattern rules applied against those adaptive limits catch the drift patterns — liner wear, recipe change, feed hardness progression — that static single-point alarms never see. And root-cause ML attribution turns an anonymous alarm into an actionable instruction the operator can execute in seconds. The technology isn't new. What's new is having it applied to a process as variable as a live crushing circuit.
iFactory's Adaptive SPC platform connects to your existing DCS and historian to deploy dynamic control limits, Western Electric rule detection, and root-cause attribution across your crushing circuit — without replacing your instrumentation or retraining your team. Book a Demo to see what adaptive limits would have caught on your last 90 days of data, or Get In Touch to start your deployment assessment.
