Predicting Tailings Dam Failures: Are Numerical Models Up to the Task?

Tailings dams, essential for storing mining byproducts, represent one of the most significant geotechnical challenges in engineering. These structures are exposed to diverse factors like material variability, unsaturated conditions, particle segregation, and environmental interactions. Despite the reliance on numerical models to simulate their behavior, the limitations of these models are increasingly evident, particularly after recent disasters like Mount Polley (Canada) and Brumadinho (Brazil). Such events underscore the complexity of tailings dams and the urgent need for improved risk prediction methods.

Tailings materials are highly variable, and their geotechnical properties can change spatially. Moreover, unsaturated conditions, segregation during deposition, and material degradation add layers of uncertainty. These challenges make it critical for engineers to refine tools and techniques for better accuracy in predicting dam behavior under extreme conditions.

The Limitations of Numerical Models

While numerical simulations are invaluable for understanding tailings dams, they have notable shortcomings. Key issues include:

1. Model Simplifications:

   • Geometry and boundary conditions often oversimplify real-world scenarios.
   • Input parameters like shear strength and permeability can carry significant uncertainties.

2. Complex Phenomena:

   • Liquefaction, a major cause of failures, remains difficult to model accurately, especially for materials with high fines content.
   • Debris flows during dam failures pose additional challenges due to the interaction of multiphase materials with varied topography.
   • Progressive deformation over time is not easily captured by most numerical models.

3. Input Data Limitations:

   • Spatial variability of tailings properties.
   • Measurement errors in laboratory or in-situ testing.
   • Scarcity of comprehensive data for precise modeling.

Toward Reliable Modelling and Safer Tailing Dams

To enhance the reliability of numerical simulations, it's imperative to adopt a multifaceted approach:

1. Validation and Back-Analysis:

   • Models must be validated against field monitoring data and past failure cases.
   • Sensitivity analyses help identify critical input parameters and their influence.

2. Technological Integration:

   • Leveraging machine learning can improve failure prediction by identifying patterns in large datasets.
   • Advanced models incorporating multiphysics simulations enable a more holistic analysis of dam behavior.

3. Field Experience and Collaboration:

   • Geotechnical observations, in-situ testing, and engineering judgment complement numerical approaches.
   • Collaboration between engineers, hydrologists, and geologists ensures a comprehensive risk assessment.

By addressing the limitations of numerical models and integrating technological and field insights, the mining industry can take significant strides toward ensuring tailings dam safety. This proactive approach can prevent disasters and build public trust in geotechnical practices.