ValueDrill

Overview | Installation | Structure | Usage | Citations

Smart Drilling Decision-making for Mineral Exploration

---

Overview

Drilling is the most cost-intensive component of orebody definition. Conventional drilling campaigns commonly prioritize intersecting mineralization, resulting in highly consistent drill collar locations and orientations. This drilling strategy can minimize the risk of missing ore, but it limits the ability to reduce subsurface uncertainty and resolve the broader geological context. We introduce an information-driven drilling framework that integrates geological knowledge, Bayesian stochastic modeling, and the value of information (VOI). Given an ensemble of prior geological model realizations, this package computes the VOI for each candidate drillhole — the expected gain in decision quality from drilling that hole before committing to an action decision. An alternative grade-thickness metric is also provided for comparison.

Authors:

Xiaolong Wei (xiaolong.wei@ubc.ca)

Haruna M. Grema (hgrema@eoas.ubc.ca)

Brian A. McNulty (bmcnulty@eoas.ubc.ca)

Lindsey J. Heagy (lheagy@eoas.ubc.ca)

Shaun L.L. Barker (shaun.barker@ubc.ca)

Acknowledgments:

NSERC

IMDEX

Ideon

---

Installation

Clone the repository and install the dependencies:

git clone https://github.com/xiaolongw1223/voi_drill.git
cd voi_drill
pip install -r requirements.txt

The example scripts run without installation — they add the package to sys.path automatically.

---

Structure

ValueDrill/
├── inputs/                           # input data (place files here before running)
│   ├── drillhole_2d.txt              # existing drillhole observations
│   ├── true_model_binary_2d.txt      # binary true model (for reference)
│   └── prior_model_geometry.txt      # prior model ensemble (n_models × n_cells)
├── outputs/                          # all output figures saved here (auto-created)
├── examples/                         # runnable example scripts
│   ├── case1_vertical.py
│   ├── case1_vertical_analysis.py
│   ├── case2_cone.py
│   └── case3_random.py
├── voi_drill/                        # core package
│   ├── __init__.py
│   ├── utilities.py
│   ├── trajectory.py
│   └── metrics.py
├── requirements.txt
└── pyproject.toml

---

Usage

Minimal example

import numpy as np
from discretize import TensorMesh
from voi_drill import (
    simulate_trajectory_2d,
    value_function_action,
    utility_linear,
    VOI,
    grade_thickness,
)

# 130 × 60 cell mesh, 10 m cell size
dh = 10
hx = [(dh, 130)]
hz = [(dh, 60)]
mesh = TensorMesh([hx, hz], "00")

# prior_models: shape (n_models, n_cells), binary ore indicators
prior_models = np.loadtxt("inputs/prior_model_geometry.txt").T

# Define 5 vertical candidate drillholes
x = np.linspace(200, 1000, 5)
params = np.column_stack([x, np.full(5, 599.5), np.full(5, 500.0), np.full(5, 90.0)])
x_all, z_all = simulate_trajectory_2d(params, n_segments=50)
trajectories = [np.c_[x_all[i], z_all[i]] for i in range(5)]

# Compute risk-neutral VOI
voi_calc = VOI(
    value_fn=value_function_action,
    value_kwargs={"hit_value": 500, "miss_value": -100},
    utility_fn=utility_linear,
    utility_kwargs={"coefficient": 1.0, "cutoff": 0.0},
    no_action_value=0.0,
)
results = voi_calc.compute_for_candidates(trajectories, prior_models, mesh)
print("VOI:", results["voi"])

# Compute expected ore thickness
gt = grade_thickness(trajectories, prior_models, mesh, segment_length=10.0)
print("Grade-thickness (m):", gt)

Risk-adjusted VOI

Pass a non-linear utility function to encode risk aversion:

from voi_drill import utility_exponential

voi_calc = VOI(
    value_fn=value_function_action,
    value_kwargs={"hit_value": 500, "miss_value": -100},
    utility_fn=utility_exponential,
    utility_kwargs={"alpha": 20, "beta": -30, "gamma": 0.1},  # risk-averse
    no_action_value=0.0,
)

Reproducing the paper figures

Place the input data files in inputs/ then run any script from the examples/ directory. Figures are saved automatically to outputs/.

cd examples
python case1_vertical.py

Script	Output figures
case1_vertical.py	fig_case1_vertical_model_drillhole_trace.png, fig_case1_vertical_voi_thickness.png
case1_vertical_analysis.py	fig_case1_vertical_analysis_drillhole_1.png
case2_cone.py	fig_case2_cone_model_drillhole_trace.png, fig_case2_cone_voi.png, fig_case2_cone_thickness.png
case3_random.py	fig_case3_random_model_drillhole_trace.png, fig_case3_random_voi_thickness.png

---

Citations

If you are using the package, please cite:

---

License

MIT