Definition
Rate at which autonomous drill pattern execution drifts from designed specifications over time, measuring whether the system is developing systematic biases.
The Measurement Gap
The Measurement Gap
When operations review autonomous drill performance, they examine aggregate pattern statistics: average deviation, maximum deviation, percentage of holes within tolerance. These summary statistics treat every hole as independent. They do not ask whether the deviations are getting worse as the drill moves through the pattern.
This matters because autonomous drill AI systems carry forward contextual state. If the drill encounters hard rock at hole 5 and adapts by increasing pulldown pressure, that adaptation may persist or influence subsequent decisions at holes 6, 7, and 8. If the rock characterization model has a systematic bias in one direction, each successive adaptation compounds the previous one. The result is a drift pattern where early holes closely match the design and late holes diverge significantly, but the average deviation across all holes remains within tolerance.
Legacy quality metrics miss this sequential dependency entirely. A pattern where holes 1-10 deviate 0.05m and holes 30-40 deviate 0.35m will report the same average deviation as a pattern where all holes deviate 0.20m. But the blast engineering implications are very different: concentrated deviation in one portion of the pattern creates localized burden/spacing problems that distributed deviation does not.
Without sequential analysis, operations cannot distinguish between random scatter (acceptable) and systematic compounding (a control logic problem that will worsen as pattern sizes increase).
Decision Supported:
Are the drill AI's adaptations accumulating in one direction, and will the last holes in the pattern be significantly further from design than the first holes?
Formula
Plain Language
For each hole in the pattern, compute the composite deviation from design (combining position, depth, and angle deviations into a single distance metric). Plot these deviations in the order the holes were drilled (hole 1, hole 2, ... hole N). Fit a straight line through the points using linear regression. The slope of that line is PCDI. A positive slope means deviations are growing as the drill progresses. A slope near zero means deviations are random. A negative slope means the drill is self-correcting.
Notation
PCDI = Beta_1
Where Beta_1 is the slope from the linear regression: Deviation_i = Beta_0 + Beta_1 x hole_sequence_number_i + error_i
Variables
| Variable | Description | Source |
|---|---|---|
| Deviation_i | Composite distance between the as-drilled and as-designed position, depth, and angle for hole i (same composite metric used in DBCI, using the same unit normalization) | Calculated from drill survey and design data |
| hole_sequence_number | The order in which holes were drilled (1, 2, 3, ... N), following the actual drill path rather than the design numbering | Drill system operational logs |
| Beta_1 | Slope of the linear regression of deviation against sequence number, expressed in deviation units per hole (e.g., meters per hole) | Calculated from regression |
| Beta_0 | Intercept of the regression, representing the expected deviation of the first hole drilled | Calculated from regression |
The sign of PCDI is as important as the magnitude. Positive values indicate compounding drift. Negative values indicate convergence. The absolute value indicates the rate of change per hole.
Proposed formula, pending validation pilots.