A surrogate-assisted hybrid evolutionary framework for haulage cycle time optimisation in open-pit mining

Haulage cycle time is the primary productivity metric in open-pit mining, yet optimising fleet zone allocation remains computationally intractable by exhaustive search for realistic fleet sizes. Classical Heuristic Dispatch Rules (HDR) ignore fleet-level congestion externalities and systematically over-concentrate trucks, whereas evolutionary optimisers require expensive repeated fitness evaluations. This study addresses both limitations by training a Random Forest (RF) surrogate model on 70 operational cycle-time records from Ashaka Cement Quarry (Gombe State, Nigeria). The RF is trained on six input features including loaded and empty travel time, and achieves R² = 0.9883 and MAPE = 3.26% on held-out records, outperforming Gaussian Process Regression (GPR), XGBoost, and Multi-Layer Perceptron alternatives on this data-scarce dataset. Four dispatch strategies are evaluated over 30 independent runs on a five-zone algorithm-benchmarking scenario constructed from the empirical Ashaka data by controlled zone-feature scaling: the classical HDR, a Genetic Algorithm with RF surrogate (GA + RF), Particle Swarm Optimisation with RF surrogate (PSO + RF), and a novel two-phase Hybrid GA + PSO+RF with warm-start transfer. GA + RF reduces mean fleet cycle time by 26.2% versus HDR (19.505 vs. 26.432 s), while the Hybrid GA + PSO+RF achieves the best mean performance (19.372 s, -26.7%) and the lowest inter-run standard deviation (SD = 0.097 s). Under a common deterministic cost evaluation, the Hybrid mean solution (19.318 s) lies 0.3% above the exact enumeration optimum (19.256 s), confirming near-optimal convergence. The warm-start transfer mechanism is the principal methodological contribution: it eliminates PSO initialisation sensitivity and produces statistically reliable allocations without requiring real-time computation.

Authors

Institutions

Publication Details

Journal
Discover Geoscience
Published
2026-11-19
DOI
https://doi.org/10.1007/s44288-026-00833-7
Primary Topic
Mining Techniques and Economics
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A surrogate-assisted hybrid evolutionary framework for haulage cycle time optimisation in open-pit mining

Abdouramani Dadjé, Boroh André William, Albertine Alarba Sam-Tunsa, Mouminou Nchare et al.
Discover Geoscience
Mining Techniques and Economics
article

A surrogate-assisted hybrid evolutionary framework for haulage cycle time optimisation in open-pit mining

Abdouramani Dadjé, Boroh André William, Albertine Alarba Sam-Tunsa, Mouminou Nchare, Markous Gninaba
article en

Abstract

Haulage cycle time is the primary productivity metric in open-pit mining, yet optimising fleet zone allocation remains computationally intractable by exhaustive search for realistic fleet sizes. Classical Heuristic Dispatch Rules (HDR) ignore fleet-level congestion externalities and systematically over-concentrate trucks, whereas evolutionary optimisers require expensive repeated fitness evaluations. This study addresses both limitations by training a Random Forest (RF) surrogate model on 70 operational cycle-time records from Ashaka Cement Quarry (Gombe State, Nigeria). The RF is trained on six input features including loaded and empty travel time, and achieves R² = 0.9883 and MAPE = 3.26% on held-out records, outperforming Gaussian Process Regression (GPR), XGBoost, and Multi-Layer Perceptron alternatives on this data-scarce dataset. Four dispatch strategies are evaluated over 30 independent runs on a five-zone algorithm-benchmarking scenario constructed from the empirical Ashaka data by controlled zone-feature scaling: the classical HDR, a Genetic Algorithm with RF surrogate (GA + RF), Particle Swarm Optimisation with RF surrogate (PSO + RF), and a novel two-phase Hybrid GA + PSO+RF with warm-start transfer. GA + RF reduces mean fleet cycle time by 26.2% versus HDR (19.505 vs. 26.432 s), while the Hybrid GA + PSO+RF achieves the best mean performance (19.372 s, -26.7%) and the lowest inter-run standard deviation (SD = 0.097 s). Under a common deterministic cost evaluation, the Hybrid mean solution (19.318 s) lies 0.3% above the exact enumeration optimum (19.256 s), confirming near-optimal convergence. The warm-start transfer mechanism is the principal methodological contribution: it eliminates PSO initialisation sensitivity and produces statistically reliable allocations without requiring real-time computation.

Discover GeoscienceVol. 4(1)
University of Maroua (CM), University of Ngaoundéré (CM)
Mining Techniques and Economics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon