Definition
Accuracy of the autonomous drill rig MWD-derived rock characterization compared to actual geological conditions confirmed by blast results or geological mapping.
The Measurement Gap
The Measurement Gap
Autonomous drill systems generate rich real-time data about the rock they are drilling through. Rotary torque, penetration rate, vibration signatures, and pulldown pressure produce a continuous stream that the drill AI interprets as rock characterization: hard zone, soft zone, fractured zone, void. These characterizations drive real-time parameter adaptations (slower penetration in hard rock, reduced pulldown in fractured zones) and are increasingly used for geological model updates.
The problem is that nobody routinely checks whether the drill's interpretation is correct. The drill "feels" resistance and calls it hard rock, but resistance can also come from bit wear, hole deviation, clay seams, or water-bearing fractures. Without independent validation, the drill's characterization is an assertion, not a measurement.
Legacy drilling metrics focus on productivity (meters per shift, utilization) and mechanical compliance (hole position, depth). Rock characterization data is treated as a bonus output rather than a validated input. When the drill AI uses its own characterization to adapt drilling parameters, it creates a closed loop with no external validation. The system optimizes against its own assumptions rather than against ground truth.
The consequence extends beyond drilling. If the characterization data feeds into geological block models, blast design optimization, or downstream processing predictions, errors propagate through multiple decision chains without detection.
Decision Supported:
Can the drill's real-time rock sensing be trusted for parameter optimization, or is it conflating unrelated rock properties?
Formula
Plain Language
Collect the drill's rock characterization score for each hole alongside an independent measurement of rock mass quality at the same location. Compute the statistical correlation between the two datasets. A strong positive correlation means the drill's sensing aligns with reality. A weak correlation means the drill's characterization is unreliable. A negative correlation means the drill is systematically wrong, interpreting conditions inversely.
Notation
RCVS = r(Drill_hardness_proxy, Actual_RMR)
Where r is Pearson correlation (for linear relationships) or Spearman rank correlation (for monotonic but non-linear relationships).
Variables
| Variable | Description | Source |
|---|---|---|
| Drill_hardness_proxy | The autonomous drill AI's composite rock characterization score, derived from rotary torque, penetration rate, vibration spectrum, and pulldown pressure (dimensionless score) | Autonomous drill system telemetry |
| Actual_RMR | Rock Mass Rating (or equivalent: RQD, Q-system, GSI) from independent measurement at the same location as the drilled hole | Core logging, blast hole camera imagery, face mapping, or geological survey |
| r | Correlation coefficient (Pearson for linear, Spearman for rank-order), measuring the statistical association between the drill's proxy and the independent measurement | Calculated from matched datasets |
The choice between Pearson and Spearman correlation depends on the expected relationship. If the drill's proxy should scale linearly with rock quality, use Pearson. If the relationship is expected to be monotonic but not necessarily linear (harder rock always produces higher proxy scores, but not proportionally), use Spearman. Report both if uncertain.
Proposed formula, pending validation pilots.