The Haulage Module: From Simulation to Optimisation to Costing
A deep technical look at how MiningIQ integrates first-principles haulage physics, pseudo-flow pit optimisation, Numba-accelerated structure arcs, and BZ scheduling into a single vertically integrated pipeline.
MiningIQ's haulage module is not a standalone cycle-time calculator. It is one layer in a vertically integrated pipeline that begins with a geological block model and ends with a fully costed, time-sequenced mine schedule. Every block in the model carries a haulage cost computed from first-principles physics — not a lookup table, not a flat $/t assumption.
This article details the full technical architecture of the system — from road network construction through physics-based truck simulation, discrete event fleet modelling, pseudo-flow pit optimisation, and Lagrangian scheduling.
The Integration Pipeline
Seven layers, each feeding the next — from geology to costed schedule.
Road Network
Haul road profiles from Vulcan or Deswik are imported as 3D polylines. MiningIQ auto-clusters endpoints into junction nodes via KDTree, builds a NetworkX directed graph, and discovers K-shortest route candidates from any pit bench to any destination.
Physics Engine
Segment-by-segment kinematic simulation using first-principles physics. OEM rimpull and retarder curves, grade resistance, rolling resistance, corner speed caps, and anticipatory braking with configurable look-ahead.
Batch Simulation
Every combination of pit, bench, destination, and truck type is simulated through the physics engine. The fastest non-stalled route is selected. Results are stored as pre-computed cycle times for instant block-level lookup.
Block Integration
Each block in the model receives a haulage cost from its pre-computed cycle time. This feeds directly into the Mining Cost Adjustment Factor (MCAF) used by the pit optimiser — deeper benches with longer hauls have higher costs, which shifts pit shell boundaries.
Pit Optimisation
Hochbaum Pseudo-Flow (HPF) via MineFlow C++ finds the provably optimal pit. The parametric solver sweeps revenue factors to produce nested pit shells — 39 shells from a 500K block model in seconds. Constrained nesting enforces minimum cutback width, tonnes, and bench span for practical pushbacks.
BZ Scheduling
The Bienstock-Zuckerberg algorithm adds the fourth dimension: time. HPF shell sequences define mining order. The BZ Lagrangian relaxation assigns pushbacks to periods, respecting mining and processing capacity constraints, to maximise discounted NPV.
Fleet Optimisation
Schedule tonnes per period drive the DES fleet optimiser. An iterative binary search determines the minimum truck and shovel fleet to meet each period's production target — feeding directly into capital and operating cost models.
Key Technical Highlights
Physics Engine
Each truck traversal is simulated segment by segment. OEM rimpull and retarder curves interpolated via numpy, with look-ahead anticipatory braking that computes maximum safe entry speeds for every segment transition.
Pseudo-Flow Optimisation
MineFlow's C++ HPF solver finds the provably optimal pit in near-linear time. The parametric solver builds the precedence graph once and re-uses solver state across revenue factors — 39 shells in seconds, not minutes.
Numba Structure Arcs
JIT-compiled two-pass kernel generates millions of block precedence arcs from variable slope domains with Whittle-compatible bearing interpolation. A 500K model produces ~25M arcs in under 10 seconds.
BZ Scheduling
Adds the fourth dimension — time — to the pit optimisation. The HPF shell sequence defines what to mine; the BZ scheduler determines when, assigning pushbacks to periods while respecting mining and processing capacity constraints to maximise discounted project value. Critically, it assists in selecting the revenue factor shell that maximises NPV — almost always a lower RF than 1.0.
Discrete Event Simulation
SimPy-based fleet dynamics with truck-shovel queuing, stochastic distributions, operational delays from MineCost rosters, BPR road congestion model, and equipment breakdowns with MTBF/MTTR sampling.
Fleet Optimiser
Iterative DES-driven binary search for fleet sizing. Shovel-limited vs truck-limited analysis with automatic rebalancing. Annualisation from MineCost roster effective hours.
Download the Full Article
The complete article covers every layer of the system in detail — how each component works, how they connect, and how the integrated pipeline delivers a fully costed mine schedule from a single block model input.
Modular Architecture
Each layer of the system is independent and testable on its own. The physics engine has no dependency on the web framework — it is pure numerical computation. The DES fleet simulation depends only on the physics results. The pit optimiser depends only on the block model and the C++ solver.
This clean separation means components can be validated individually, and the full pipeline runs end-to-end from block model import to costed schedule in a single session.
Want to see the haulage module in action?
Book a technical demo to see how MiningIQ integrates physics-based haulage simulation with pit optimisation and scheduling.