The flotation operator starts the shift and checks the SPC chart on the control room monitor. Cell three's froth depth reading has crossed the lower control limit. The alarm is flashing. The operator stops what they are doing and begins checking — reagent flow, air rate, pulp level, feed grade. Everything is within normal range for the current ore type. The reading is well within the expected range for this feed blend. But the control limit is fixed. It was calculated six months ago, for a different ore body, a different reagent regime, and a different grind target. The operator has just lost fifteen minutes to a false alarm caused by a static limit that did not move when the process moved. This is not a control system failure. It is a design limitation of fixed UCL/LCL boundaries in a process that changes hourly. AI-powered adaptive SPC limits for mining flotation replace static boundaries with dynamic limits that move with your process — eliminating false alarms, catching real deviations early, and cutting quality-driven downtime by 60% or more.
Adaptive SPC Limits · Dynamic UCL/LCL · AI Flotation Control · Downtime Elimination
Your SPC Limits Are Based on Last Year's Ore. Today's Ore Is Different. Adaptive Limits Move with Your Process — Not Against It.
iFactory's AI-powered adaptive SPC engine replaces static UCL/LCL boundaries with multivariate dynamic limits that adjust in real time to ore type changes, feed grade variation, reagent regime shifts, and grind target adjustments — eliminating false alarms and cutting quality-driven downtime by 60% or more in operating flotation circuits.
60–80%
Reduction in false alarms when fixed SPC limits are replaced with dynamic adaptive boundaries calculated from real-time multivariate process data
45 min
Average operator time lost per shift to investigating false SPC alarms that turn out to be normal process variation for the current operating conditions
100+
Flotation variables simultaneously modeled by the adaptive limit engine — froth vision, reagent flows, air rates, cell levels, feed characteristics — for each cell in the bank
60%+
Quality-driven unplanned downtime eliminated when operators act on adaptive-limit alerts instead of chasing false alarms from static boundaries
Why Fixed SPC Limits Fail in Mining Flotation — and Why Adaptive Limits Fix It
A flotation circuit does not operate under constant conditions. Ore mineralogy shifts between stopes. Feed grade changes at every mill blend change. Reagent demand varies with liberation size. Pulp density fluctuates with grind circuit throughput. Each of these changes shifts the expected operating range of every flotation variable — froth depth, bubble size, air recovery, concentrate grade, tailings grade. A fixed UCL/LCL calculated from last quarter's data cannot represent the current process capability because the process itself has moved. The result for operators is a steady stream of false alarms that destroy trust in the SPC system entirely.
Static UCL/LCL — Fixed Boundaries
UCL
LCL
Froth Depth — Cell 3
Ore type change at 14:30 shifts natural froth depth range. Static UCL/LCL does not move. Result: 3 false alarms in 90 minutes requiring operator investigation.
3 false alarms flagged as OOC — all were normal variation for current ore type
Adaptive UCL/LCL — Dynamic Boundaries
UCL
LCL
UCL
LCL
Froth Depth — Cell 3
Ore type change at 14:30 shifts natural froth depth range. Adaptive UCL/LCL recalculates from current multivariate data. Result: 0 false alarms through the transition.
0 false alarms — limits shifted with the ore type; operator focused on real process optimisation
The Adaptive Limit Engine: 4 Components That Replace Static Boundaries
The adaptive limit engine does not simply widen or narrow fixed boundaries. It replaces the concept of a static limit with a dynamic baseline that represents what the process should be doing right now — given the current feed, the current ore, the current reagent regime, and the current operating target. Every variable gets its own adaptive UCL and LCL that moves as the process moves.
01
Multivariate Baseline Model
The engine ingests froth vision, reagent flow, air rate, cell level, and feed analyser data for every cell simultaneously. A multivariate deep learning model establishes the normal operating relationship between all variables for the current ore type and circuit configuration. The baseline is not a single number — it is a high-dimensional manifold representing the acceptable operating region given current conditions.
02
Dynamic Boundary Calculator
For each variable, the engine computes the expected value and the acceptable deviation range given the current operating regime. If ore type A produces froth depths between 18–25 mm and ore type B produces froth depths between 14–20 mm under the same reagent regime, the UCL/LCL for froth depth shifts automatically when the ore type changes — with no operator action required.
03
Residual Deviation Detector
Instead of comparing raw variable values against fixed limits, the detector compares the residual — the gap between the actual value and the expected value given current multivariate conditions. A froth depth of 22 mm is above the fixed UCL for ore type B but within the adaptive limit because the engine knows the current ore type and reagent regime. The residual is near zero. No alarm.
04
Alert Confidence Classifier
Every potential deviation receives a confidence score representing the probability that it is a true special cause requiring action rather than normal process variation. The operator sees: "Froth depth deviation on cell 3 — 92% probability of special cause (reagent line blockage pattern detected)." Alerts below the configurable confidence threshold are suppressed — only actionable alerts reach the operator.
Operator Dashboard: Adaptive SPC View — What Changes on the Screen
The dashboard does not look like a traditional SPC chart with fixed horizontal lines. Every control limit on every variable is a dynamic boundary that the operator can see moving in response to changing circuit conditions. The operator is not chasing false alarms — they are watching the process and acting on real deviations only.
Dashboard Panel 01
Live Adaptive Control Chart — Limits That Move
The primary chart displays each flotation variable with its adaptive UCL and LCL as dynamic bands. When ore type changes at 14:30, the UCL/LCL band for froth depth shifts from 18–25 mm to 14–20 mm within seconds. The operator sees the process data points remaining comfortably within the shifted limits — no alarm, no investigation, no lost time. The chart includes a tooltip showing the current expected value, the contributing variables (current ore type, reagent regime, feed grade), and the confidence level of the limit calculation.
Operator action: Monitor the adaptive band — if data stays inside, the process is in control for current conditions. No action needed.
Dashboard Panel 02
Process Drift Matrix — Variable-Level Deviation View
A color-coded matrix shows every monitored variable in the circuit: green if within adaptive limits, yellow if approaching the boundary, red if exceeding with high confidence. Each cell displays the current value, the adaptive UCL/LCL range, and the residual percentage. The operator sees at a glance which variables are drifting and by how much — without needing to check each variable individually against a static number.
Operator action: Scan the matrix for red cells — investigate only the variables flagged as high-confidence deviations beyond adaptive limits.
Dashboard Panel 03
Alert History with Outcome Log — Learning from Every Event
Every adaptive limit alert is logged with its confidence score, the operator's response, the corrective action taken, and whether the deviation resolved. Over time, the model learns which alert patterns correspond to real process issues and which represent normal variation that was not captured by the current baseline. False positive patterns are automatically suppressed in future similar conditions.
Operator action: Log corrective action and outcome — model calibrates future alert thresholds based on real operator feedback.
Dashboard Panel 04
Downtime Impact Counter — Real-Time Savings Tracker
A live counter displays the operator time saved by adaptive limits vs fixed limits. It calculates: false alarms avoided × average investigation time per alarm. After one shift, the counter might read "2.5 hours saved today." After one month, "52 hours saved — the equivalent of 1.3 operator weeks returned to productive circuit optimisation." This is not an abstract metric — it is the operator's own time, tracked shift by shift.
Operator action: Review the time-saved counter — use recovered time for froth walk rounds, reagent tune-ups, and circuit fine-tuning.
Downtime Elimination: Operator Time Comparison — Fixed Limits vs Adaptive Limits
The impact of adaptive SPC limits on operator downtime is measurable from the first shift. Every false alarm eliminated is operator time returned to productive work. Every real deviation caught early is a quality excursion prevented. The table below shows the difference across a typical flotation operating day.
Daily Operator Time Allocation: Fixed SPC vs Adaptive SPC
Activity
Fixed SPC Limits
Adaptive SPC Limits
False alarm investigation
35-55 min — investigate 3-5 false OOC alerts per shift triggered by ore type changes or feed variation
Under 5 min — 0-1 alerts per shift; all are high-confidence true deviations requiring action
SPC chart review and interpretation
20-30 min — cross-reference fixed limits against shift notes; mentally adjust for known ore changes
Under 5 min — limits already adjusted; operator scans drift matrix and confirms all variables in band
Process optimisation (circuit tuning)
10-20 min — limited time after false alarm investigation and SPC review consume most of shift
40-60 min — majority of shift available for froth walk, reagent adjustment, and recovery optimisation
Quality-driven downtime events
1-2 events per week — undetected real deviations lost among false alarms; discovered via lab assay lag
<1 event per month — real deviations caught at onset by residual deviation detection before grade impact
Fixed SPC Limits
65-105 min
Operator time consumed per shift by false alarms, manual SPC interpretation, and reactive investigation
Adaptive SPC Limits
10-15 min
Operator time required per shift to review adaptive SPC dashboard and respond to high-confidence alerts only
Conclusion
Fixed SPC limits were designed for processes that stay in one operating range. Mining flotation is not one of those processes. Ore bodies change between stopes. Feed grade drifts with every mill blend. Reagent demand shifts with liberation size. Pulp density varies with grind circuit throughput. Each change moves the natural operating range of every flotation variable — froth depth, bubble size, air recovery, concentrate grade, tailings grade — yet the UCL and LCL stay exactly where they were when they were calculated months ago. The result is a steady stream of false alarms that operators learn to ignore, and a parallel stream of real deviations that get discovered too late because they were indistinguishable from the noise.
Adaptive SPC limits solve this by replacing static boundaries with a dynamic baseline that represents what the process should be doing right now — given the current feed, the current ore, the current reagent regime, and the current operating target. The operator does not need to mentally adjust for ore type changes or feed grade shifts. The limits adjust themselves. Every variable gets its own adaptive UCL and LCL that move as the process moves. False alarms drop by 60-80%. Real deviations are detected early, at the residual level, before they produce a grade excursion. And the time the operator used to spend chasing false alarms is returned to productive circuit optimisation.
iFactory's adaptive SPC limits engine is built for flotation operators who need control limits that reflect the process they are actually running — not the process that was running six months ago. Book a Demo to see how adaptive UCL/LCL boundaries respond to ore type changes, feed grade variation, and reagent regime shifts in real time, or talk to an expert about deploying adaptive SPC limits on your flotation circuit to eliminate quality-driven downtime and return operator time to process optimisation.
Frequently Asked Questions
Your SPC Limits Were Set When the Ore Was Different. Today's Flotation Circuit Needs Limits That Change as Fast as Your Feed Does.
iFactory's adaptive SPC limits engine monitors every flotation variable against dynamic UCL/LCL boundaries that shift in real time with ore type, feed grade, reagent regime, and grind target changes — eliminating false alarms, catching real deviations at the residual level, and returning hours of operator time per shift from false alarm investigation to process optimisation.
