🎓 Lesson 7
D4
Sensor Selection and Placement Strategy
Choosing the right sensors and putting them in the best places to accurately measure ore grade while blasting or mining in real time.
🎯 Learning Objectives
- ✓ Analyze trade-offs among sensor types (XRF, LIBS, PGNAA) for specific ore matrices and operational environments
- ✓ Design sensor placement layouts that satisfy Nyquist–Shannon sampling criteria for grade variability at bench, muckpile, and crusher feed scales
- ✓ Calculate minimum detectable concentration (MDC) and precision limits for a given sensor configuration under field conditions
- ✓ Explain how sensor drift, matrix effects, and particle size distribution impact real-time grade estimation uncertainty
📖 Why This Matters
In modern mines, real-time grade control isn’t just about faster decisions—it’s about preventing dilution, reducing regrind costs, and meeting smelter penalties. A $500k sensor installed in the wrong location or mismatched to the ore type delivers zero ROI. This lesson shows how engineering judgment—not just vendor specs—determines whether your grade control system guides operations—or misleads them.
📘 Core Principles
Effective sensor strategy rests on three interdependent pillars: (1) Measurement physics—understanding how each sensor interacts with ore (e.g., XRF penetration depth vs. LIBS plasma volume); (2) Geostatistical representativity—ensuring sampling frequency and position capture grade heterogeneity at relevant scales (bench > muckpile > ROM bin); and (3) Operational integration—accounting for dust, vibration, moisture, conveyor speed, and maintenance access. Poor placement violates the 'sampling unit equivalence principle': if the sensor sees only 10% of the material stream’s grade variance, no algorithm can compensate.
📐 Minimum Sampling Frequency (Nyquist–Shannon Adaptation)
To resolve grade variation without aliasing, sensor sampling rate must exceed twice the dominant spatial frequency of grade heterogeneity along the material flow path. This adapted Nyquist criterion ensures temporal resolution matches geological variability scale.
💡 Worked Example
Problem: A copper-gold porphyry ore exhibits dominant grade variogram range of 1.8 m along conveyor belt direction; belt speed = 2.4 m/s; sensor dwell time per sample = 0.3 s.
1.
Step 1: Convert spatial grade variability into temporal frequency: f_spatial = 1 / 1.8 m ≈ 0.556 m⁻¹
2.
Step 2: Multiply by belt speed to get temporal frequency: f_temporal = 0.556 m⁻¹ × 2.4 m/s ≈ 1.33 Hz
3.
Step 3: Apply Nyquist: f_sample ≥ 2 × f_temporal = 2.66 Hz → max sampling interval ≤ 1 / 2.66 ≈ 0.376 s
4.
Step 4: Compare with sensor dwell time (0.3 s): 0.3 s < 0.376 s → sampling is sufficient
Answer:
The sensor’s 0.3 s dwell time satisfies the Nyquist criterion (0.376 s max), ensuring grade variability is resolved without aliasing.
🏗️ Real-World Application
At Newcrest’s Telfer Mine (Western Australia), an inline XRF analyzer was initially mounted 1.2 m upstream of the primary crusher discharge chute. High dust loading and coarse particle segregation (>150 mm) caused 22% grade bias vs. lab assays. Relocation to a vibratory feeder downstream—where material was homogenized and <75 mm—combined with dual-angle XRF geometry, reduced bias to <3% and improved Cu recovery forecasting accuracy by 17% over 3-month trial.