Files
LemyStorme_59372100_2026.pdf
Open access - Adobe PDF
- 32.57 MB
Details
- Supervisors
- Faculty
- Degree label
- Abstract
- Mudflows are deadly gravity-driven hazards on Earth: unpredictable dense torrents of water, soil, and rock that travel at tens of metres per second and bury everything in their path. Predicting where they go, how fast, and where they stop is a central and hard challenge for risk mitigation. Their complex behaviour, which combines the physics of water flows and granular media, makes them difficult to model and simulate. This master’s thesis tackles a simple question: can Watlab, a finite-volume shallow-water solver built at UCLouvain for river and flood flows, be turned into a reliable predictive tool for viscoplastic mudflows, and which rheology best captures the behaviour of real muds? The starting point is a single-phase, two-dimensional, depth-averaged description of the flow. A quick order-of-magnitude analysis exposes the core difficulty: in water, inertia dominates friction, but in mud the two are comparable. Therefore, the solver’s original operator-splitting strategy breaks down in the friction-driven run-out phase. Three targeted fixes restore it: a temporary state variable, a vector-valued stopping criterion, and a semi-implicit treatment of friction. On top of this, four viscoplastic laws, Bingham, Herschel-Bulkley, Casson and quadratic, are implemented. The basal shear stress is recovered both by inverting the velocity law and through a faster, explicit slope-based formulation. The model is first stress-tested on a Carbopol dam- break benchmark, then turned loose on a real natural-mud experiment from an experimental campaign performed at Politecnico di Milano. The final verdict shows that the approach based on Watlab is well suited for such applications. The slope-based closure matches the experimental data accurately, mainly for the steady flow phase, and is substantially less expensive than the velocity-based closure. On real mud, the model captures the overall shape and downstream extent of the deposit, but misses some behaviours of the experiment and the calibration remains delicate. The outcome is a version of Watlab able to simulate viscoplastic free-surface flows, but not yet a ready-to-use predictive tool for natural mudflows. The perspectives for future work are numerous and promising for the development of a reliable tool for risk mitigation.