π Lesson 6
D4
Designing Closed-Loop Grade Control Architectures
A closed-loop grade control architecture is a system that continuously measures ore grade in real time, compares it to the target, and automatically adjusts mining or processing actions to keep the output grade on target.
π― Learning Objectives
- β Design a closed-loop control architecture for a shovel-truck-crusher-plant feed stream using appropriate sensor placement and controller type
- β Analyze loop stability and response time given sensor delay, actuator lag, and ore transit time
- β Calculate required sensor precision and sampling frequency to achieve Β±0.1% Cu grade control at 95% confidence
- β Explain trade-offs between feed-forward and feedback strategies in grade blending applications
- β Apply ISO/IEC 62443 cybersecurity principles to protect grade control network endpoints
π Why This Matters
In modern bulk copper and gold operations, even minor grade excursions β as small as Β±0.05% Cu β can cost $2β5M annually in lost recovery or penalty smelter charges. Traditional open-loop planning (e.g., pre-blast grade models) fails under rapid ore variability from geological complexity or dilution. Closed-loop grade control transforms grade from a *monitored outcome* into a *controlled input*, enabling real-time optimization of mill throughput, reagent use, and tailings quality β directly impacting net smelter return (NSR) and ESG metrics like energy per tonne of metal.
π Core Principles
Closed-loop grade control rests on four interdependent layers: (1) Sensing layer β high-fidelity, ruggedized analyzers with <2% relative error and sub-60s analysis cycle; (2) Communication layer β deterministic industrial Ethernet (e.g., TSN) with <10ms jitter to synchronize sensor, PLC, and MES timestamps; (3) Control layer β either discrete logic (for gate switching) or continuous MPC (for crusher setpoints), tuned to ore residence time (e.g., 45β90s from shovel to primary crusher); (4) Execution layer β physical actuators (e.g., variable-speed feeders, hydraulic gate diverters) with <2s response time and position feedback. Stability requires loop gain β€ 0.8 Γ critical gain (per Ziegler-Nichols), and total loop dead time must be <β
of dominant process time constant.
π Minimum Sampling Frequency Criterion
To avoid aliasing and ensure grade variance capture, the Nyquist-Shannon criterion must be adapted for geological autocorrelation length. The minimum effective sampling frequency ensures detection of grade shifts exceeding specification tolerance within one control cycle.
Nyquist-Adapted Sampling Frequency
f_min = 1 / (L_ac / (2 Γ v_conveyor) + 3 Γ Ο_sensor / Ξ_grade Γ L_ac / v_conveyor)Minimum required sampling frequency to resolve grade changes within tolerance, accounting for geological autocorrelation and sensor uncertainty
Variables:
| Symbol | Name | Unit | Description |
|---|---|---|---|
| f_min | Minimum sampling frequency | Hz | Reciprocal of maximum allowable sampling interval |
| L_ac | Grade autocorrelation length | m | Distance over which adjacent samples remain statistically correlated |
| v_conveyor | Conveyor belt speed | m/s | Linear speed of material transport |
| Ο_sensor | Sensor analytical uncertainty (1Ο) | % | Standard deviation of repeated measurements on homogeneous reference material |
| Ξ_grade | Grade tolerance half-band | % | Maximum allowable deviation from target grade (e.g., Β±0.1% β 0.1) |
Typical Ranges:
Porphyry Cu conveyor feed: 0.1 β 0.3 Hz
High-grade narrow-vein Au truck dump: 0.02 β 0.05 Hz
π‘ Worked Example
Problem: Ore exhibits grade autocorrelation over 8 m (typical for porphyry Cu skarn). Conveyor speed = 2.5 m/s. Grade tolerance band = Β±0.1% Cu. Sensor analytical uncertainty = Β±0.07% Cu (1Ο). Required confidence = 95% (z = 1.96).
1.
Step 1: Compute spatial sampling interval: max(autocorrelation_length / 2, 3 Γ sensor_uncertainty / tolerance Γ autocorrelation_length) = max(4 m, 3 Γ 0.07/0.1 Γ 8) = max(4, 16.8) = 16.8 m
2.
Step 2: Convert to time: Ξt = spatial_interval / conveyor_speed = 16.8 m / 2.5 m/s = 6.72 s
3.
Step 3: Minimum frequency = 1 / Ξt = 0.149 Hz β round up to 0.17 Hz (every ~6 s) to accommodate communication and actuation overhead
Answer:
The system must sample grade no slower than every 6 seconds (0.17 Hz), which exceeds the 10-s interval used in legacy systems β explaining frequent grade excursions at Site X.
ποΈ Real-World Application
At Rio Tintoβs Kennecott Utah Copper mine, a closed-loop architecture integrates Bruker S1 TITAN handheld XRF (calibrated to lab assays, RΒ² = 0.98), Siemens Desigo CC controllers, and hydraulic gate diverters on the primary crusher feed conveyor. When real-time Cu grade drops below 0.42%, the controller opens Gate A (high-grade stockpile) for 4.2 s while closing Gate B (low-grade ROM), adjusting blend ratio by Β±12% within 15 s. Since deployment (2021), grade standard deviation reduced from Β±0.21% to Β±0.08% Cu, increasing recoverable copper by 1.3% annually and deferring $18M in tailings storage expansion.