π Lesson 14
D5
From Blasthole Assay to Mill Feed Reconciliation
It's the process of matching the metal grade measured in blasthole samples to the actual grade of ore that ends up in the mill, so we know if our mining and sampling are accurate.
π― Learning Objectives
- β Calculate grade reconciliation error (GRE) and its confidence interval using paired assay data
- β Analyze sources of grade variance by applying mass balance equations across mining-to-mill unit operations
- β Design a blasthole sampling protocol that meets ISO 3082:2020 composite sampling precision requirements
- β Apply geostatistical estimation variance models (e.g., KRIGING vs. nearest-neighbor) to quantify expected grade misrepresentation in muckpiles
- β Explain how blast-induced fragmentation and muckpile segregation impact mill feed grade fidelity
π Why This Matters
A 5% grade reconciliation error at a $150/ton copper operation can cost over $3M annually in missed recovery or incorrect blending decisions. When blasthole assays say '0.85% Cu' but mill feed assays average '0.72% Cu', it signals hidden losses β from poor sampling representativity, undetected dilution, or inaccurate geological modeling. This lesson bridges the gap between drilling geologists and metallurgists, turning reconciliation from a reporting exercise into a predictive control tool.
π Core Principles
Reconciliation rests on three pillars: (1) Sampling Theory β blastholes provide discrete, spatially clustered samples; their representativity depends on spacing relative to geological continuity (range of influence), drill deviation, and compositing method. (2) Mass Balance β every tonne mined must be accounted for in muckpile volume, haul truck payloads, stockpile inventories, and mill feed tonnes, with grade weighted accordingly. (3) Error Propagation β assay uncertainty (Β±0.03% Cu at 95% CI), sampling bias (e.g., core loss in weak zones), and blending inefficiency compound non-linearly. Modern workflows use conditional simulation (e.g., SGSIM) to model grade uncertainty forward from blasthole grid to mill feed.
π Grade Reconciliation Error (GRE)
GRE quantifies the percentage difference between predicted (blasthole-based) and actual (mill feed) grade, normalized to the mill feed grade. It isolates systematic bias when calculated over statistically valid time periods (β₯30 consecutive shifts).
Grade Reconciliation Error (GRE)
GRE (%) = [(G_pred β G_actual) / G_actual] Γ 100Measures systematic grade bias between geological model prediction and verified mill feed grade.
Variables:
| Symbol | Name | Unit | Description |
|---|---|---|---|
| G_pred | Predicted grade | % metal | Weighted average grade from blasthole assays and block model, expressed as metal content. |
| G_actual | Actual mill feed grade | % metal | Composite assay of mill feed material, validated by certified reference materials and duplicate sampling. |
Typical Ranges:
Well-controlled open-pit operation: Β±1.5% to Β±3.5%
Deep underground with complex geology: Β±4.0% to Β±8.0%
π‘ Worked Example
Problem: Over a 30-shift period, the block model predicts mill feed grade = 0.92% Ni (based on 2.5m Γ 2.5m blasthole assays, composited by bench). Actual mill feed assay (composite of 120 primary crusher samples) = 0.84% Ni. Assay lab QA/QC shows Β±0.018% Ni (95% CI) for both datasets.
1.
Step 1: Identify G_predicted = 0.92% Ni, G_actual = 0.84% Ni
2.
Step 2: Apply GRE = [(G_predicted β G_actual) / G_actual] Γ 100 = [(0.92 β 0.84) / 0.84] Γ 100
3.
Step 3: Calculate: (0.08 / 0.84) Γ 100 = 9.52%. Compare to typical operational tolerance of Β±3.5% β this 9.5% error exceeds safe limits and triggers root cause investigation.
Answer:
The result is 9.5%, which exceeds the acceptable reconciliation tolerance of Β±3.5% and indicates significant grade misrepresentation upstream.
ποΈ Real-World Application
At Valeβs Voiseyβs Bay NiβCoβCu Operation (Labrador), a persistent +6.2% GRE in cobalt was traced to preferential sampling of high-grade fracture zones during blasthole drilling. Drillers avoided fractured ground due to bit wear, causing under-sampling of low-grade matrix rock. After implementing randomized drill pattern offsets and mandatory core recovery logging (per ASTM D5084), GRE reduced to +0.9% within 4 months β directly improving mill throughput stability and reducing reagent overuse by 11%.