 - dev.png){kind=logo}
Insights from our peer-reviewed journal | Compiled by Megan Skrip - 24 September 2026
Machine learning and centrifuge data combine to model over 200 fault scenarios, pinpointing the safest angles and burial depths for tunnel stability
Courtesy of Vahid Amini
Although important for assessing the expected responses of tunnels to geological faults, physical modelling with centrifuge tests is limited by time and cost, narrowing the range of scenarios that can be efficiently tested. The study combined centrifuge test data with data generated from software simulations to define over 200 scenarios of a tunnel encountering a dip-slip fault, then identified the least hazardous and most hazardous conditions.
“Faults are among the most important and complex geological challenges designers encounter,” said lead author Vahid Amini, a doctoral student at Amirkabir University of Technology in Iran. Yet, given tunnels’ critical infrastructure functions, “it is essential to have a comprehensive understanding of the geological conditions along the tunnel route.”
The study used three machine learning methods and a range of values for fault angle, tunnel diameter and overburden to predict both fault displacement and a fragility index for a tunnel intersecting either a reverse or normal fault. Higher overburden-to-tunnel-diameter ratios were found to produce the lowest tunnel fragility, with the safest fault angles ranging between 83 and 90 degrees; fault angles between 50 and 63 degrees posed the most risk to tunnel stability.
The CIM Journal is a quarterly digital publication for peer-reviewed technical papers available to all CIM National Members as a membership benefit and to non-members for a fee. Papers cover all facets of the mining and minerals industry, including geology, mining, processing, metallurgy, materials, maintenance, environmental protection and reclamation, mineral economics, project management, health and safety, risk management, research and development, operations, and regulatory practices and issues.
Since 2020, the CIM Journal has been published by Taylor & Francis. The editorial process remains the same and CIM National Members continue to have free access to the journal. New papers, in addition to the archives, are now hosted on the CIM Journal site at Taylor & Francis. Papers published prior to 2020 can be found on the CIM Technical Paper Library or on OneMine.
To ensure you can access this CIM Member benefit, please ensure you are logged in to your CIM account.
