Mining Pelletizing: Adaptive SPC for Less Scrap

Talk to an Expert Book a Demo

30–50%

Scrap reduction achievable with adaptive SPC and predictive quality analytics in mining pelletizing operations

8–16 mm

Target green pellet diameter range — a specification that is only consistently held when balling disc parameters adapt continuously to moisture and feed variation

20%

Efficiency improvement projected for iron ore pellet plants through AI-driven controls and adaptive process optimisation by 2026

40+

Process variables influencing green pellet quality — moisture, binder dose, disc angle, rotation speed, feed rate, particle size, and ore blend composition

Static SPC vs Adaptive SPC: What Changes at Every Stage of Pelletizing Quality Control

Process Event

Static SPC Response

Adaptive SPC Response

New ore blend arrives

False alarms on moisture and disc speed — operators learn to ignore alerts during ore transitions

Limits recalibrate to new blend baseline — only genuine drift generates an alert

Binder batch change

Binder dosage needs adjustment but static limits do not reflect the new binder efficacy — oversize rate climbs before the operator recognises the shift

Adaptive limits detect the binder-linked process shift and alert the operator to adjust dosage before oversize accumulates

Recipe grade change

Balling moisture window shifts with the new recipe — operators must manually reset expectations; static alarms fire continuously on parameters now operating at legitimate new setpoints

Recipe-aware limits load automatically when the grade change is registered — limits reflect the new normal immediately

Seasonal moisture variation

Ambient humidity changes alter the effective moisture of filter cake inputs — static limits generate seasonal false alarm patterns

Adaptive limits incorporate environmental context — seasonal moisture shifts update the baseline, genuine drifts still generate alerts

Induration temperature drift

Static limits fire when temperature deviates from the nominal setpoint — but do not correlate temperature drift with its downstream crush strength consequence

Adaptive model correlates induration temperature with fired pellet strength prediction — alerts before the crush strength test returns an off-spec result

Talk to an Expert Book a Demo

Green pellet size and wet strength are determined by the balling disc or drum parameters and the moisture content of the feed. Moisture variation of ±0.5% above the optimal balling window is sufficient to produce rapid pellet growth leading to oversize — which is recycled back through the disc, consuming capacity and energy. Moisture below the optimal window produces weak green pellets with insufficient wet strength, which spall during transport to the induration furnace. The binder dosage amplifies both effects: under-dosed binder with high moisture produces unstable large pellets; over-dosed binder with low moisture suppresses growth and creates dense, slow-growing pellets that miss the size window. Adaptive SPC tracks moisture, binder rate, and disc speed simultaneously — alerting the operator when the combination is trending toward an oversize or undersize outcome before the screen confirms it.

The induration furnace hardens green pellets through controlled drying, preheating, firing, and cooling stages — targeting temperatures of 1200°C to 1350°C. Excessive moisture in green pellets entering the drying zone causes spalling — steam pressure builds inside the pellet and ruptures it before adequate strength develops. Temperature profile drift in the firing zone produces under-fired pellets with insufficient crush strength, which fail the tumble index specification and generate fine scrap downstream. Each transition zone of the furnace — drying, preheating, firing, cooling — has its own critical parameter window. Adaptive SPC monitors temperature profiles against expected ranges that adjust for the current pellet moisture and size distribution entering the furnace — catching thermal excursions before they translate into batch-level crush strength failures.

Post-induration screening separates on-spec product (typically 9–16mm for direct reduction pellets) from oversize, undersize, and broken pellets. The crush strength and tumble index tests confirm mechanical quality. Defects at this stage are entirely preventable — they are the downstream consequence of upstream parameter deviations that occurred hours earlier. When the oversize rate exceeds target, the root cause is always in the balling moisture and disc parameters from the previous cycle. When tumble index fails, the root cause is always in the induration temperature profile from the furnace run that produced those pellets. Adaptive SPC closes this time gap by detecting the upstream deviation in real time and forecasting the downstream quality outcome before the test results confirm it.

Dashboard View 01

Live UCL/LCL With Current Regime Baseline

The control chart shows the current measurement values against the adaptive UCL and LCL — limits that have already incorporated the current ore blend, binder batch, and recipe state. The operator sees a chart where normal variation from today's inputs stays within the bands — and where genuine process drift stands out clearly because it crosses limits that are already calibrated to today's conditions. No more crying-wolf alerts on parameters that are legitimately different today than they were when the limits were last set.

Operator action: Amber trend — note it. Red breach — respond. Both are real because the limits adapt to what is normal now.

Dashboard View 02

Predictive Scrap Alert With Ranked Cause

When the adaptive model identifies a parameter combination trending toward a scrap outcome, the predictive alert fires before the product is affected. The alert shows the operator the ranked root cause: "Oversize risk elevated. Primary driver: moisture at 9.8% — 0.4% above adaptive optimal for current particle size. Reduce moisture addition by 0.3 L/min." The operator does not need to investigate. The system has done the correlation. The recommended action is specific and immediate. The operator executes, logs the action, and the alert clears when the trend reverses.

Operator action: Execute the recommended adjustment, log the action, confirm the trend reversal on the chart.

Dashboard View 03

Recipe and Blend Change Registry

Every time a recipe change, ore blend transition, or binder batch change is logged, the adaptive SPC system registers the change and begins transitioning the control limits to the new baseline using a configurable window of incoming data. The operator sees which regime is currently active and when the limits transitioned — giving them context for any process behaviour during the transition period. Recipe changes that shift balling moisture targets are applied automatically to the moisture control limit, with the old and new limits both visible during the transition window.

Operator action: Log the recipe or blend change. The system handles limit recalibration. No manual limit update required.

Dashboard View 04

Shift Scrap Summary and Event Log

At the end of each shift, the system generates an automatic quality summary: total oversize rate, any predictive alerts fired, corrective actions taken, and the Cpk for each monitored quality characteristic across the shift. Every alert, operator action, and limit change is timestamped and searchable. The shift summary satisfies internal quality record requirements without manual log entry — and provides the handover documentation the next shift needs to understand what happened and where the process currently stands.

Operator action: Review at shift end. Hand over the live dashboard to incoming operator. No paper logbook required.

Adaptive SPC uses two complementary mechanisms to make this distinction. First, when a recipe change, ore blend transition, or binder batch change is registered in the system — either manually by the operator or automatically from a DCS event — the limit recalculation begins immediately using a sliding data window that updates to the new regime. The system flags this period as a transition window, so alerts during it are contextualised as expected variation, not defect risk. Second, the adaptive algorithm uses a statistical test to distinguish between common-cause variation (which shifts the baseline) and assignable-cause events (which generate an alert against the new baseline). An assignable cause is identified when the magnitude or pattern of the deviation is inconsistent with the natural variation of the new process state. This is the distinction that static SPC cannot make — and it is the reason adaptive limits reduce false alarm rates by 50–70% while maintaining or improving detection sensitivity for genuine defects. Talk to an expert about configuring the transition logic for your recipe and ore blend change frequency.

Yes — and this is one of the highest-value capabilities for pelletizing operators. Crush strength testing occurs after induration is complete, often hours after the pellets have been produced. By that point, the batch is committed and any failure requires segregation or downgrading. The adaptive SPC model correlates historical induration temperature profiles and green pellet input characteristics with crush strength outcomes, building a predictive relationship between what happens in the furnace and the test result that will follow. When current induration temperature data matches the profile historically associated with below-specification crush strength, the system generates a predictive quality alert before the test — giving the operator time to adjust the firing zone parameters while the batch is still being processed, or to flag the batch for mandatory testing before release. Book a Demo to see crush strength forecasting demonstrated on real pelletizing process data.

iFactory supports both configurations. The highest-value deployment integrates directly with the DCS or SCADA historian — pulling process variable data automatically at the sensor polling rate, typically every 1 to 5 seconds, without operator data entry. This enables real-time adaptive limit calculation and predictive alerts at the timescale that pelletizing operations require. For sites where direct historian integration is not immediately available, a manual data entry mode supports operator-entered readings from key parameters — moisture, disc speed, binder rate, oversize rate — which feed the adaptive model at the measurement frequency the operator uses. Integration scope is confirmed during the deployment assessment. Most modern pelletizing plants with ABB, Siemens, or Rockwell control systems have standard OPC-UA or REST API connectivity available for historian data export. Talk to an expert about data connectivity options for your control system configuration.

The time lag between a balling parameter deviation and the downstream quality consequence — oversize rate, crush strength, tumble index — is one of the core challenges in pelletizing process control. A moisture excursion at the disc translates into oversize rate change at the screen 15 to 30 minutes later, and into crush strength variation after the full induration cycle. iFactory's adaptive model uses a configurable lag correlation window that maps the upstream parameter signal to the downstream quality outcome based on the process residence time. The model learns the specific lag for each process variable pair at your plant from historical data — which may differ from published averages due to equipment configuration and throughput. When a predictive alert fires, it fires at the upstream measurement, not after the downstream consequence appears — giving the operator the full lag window as intervention time. Book a Demo to see lag-corrected adaptive SPC configured for your pelletizing circuit timing.

Talk to an Expert Book a Demo