Avalanche Simulations using Fracture Mechanics on the GPU
Snow is an extremely complex material due to the structure of snow crystals and how snow behaves when it is settled within snow layers. These factors makes it hard to accurately simulate how snow layers are affected by external factors like temperature, sun radiation, and several others. And the type of snow layers, and the bonding strength between them are crucial when calculating the danger of avalanches.
In this thesis we apply fracture mechanics in order to calculate where fractures are propagating in the snow layers, and by these calculations, we can try to predict where it is high danger of avalanches. This is accomplished by using the Finite Element Method which is used to model deformation in the snow layers based on the self weight of the snow, strain and stresses is then further derived which is used to calculate the so called Energy Release Rate. The energy release rate is then compared to the Critical Energy Release Rate to determine out any fracture propagation.
The Graphical Processing Unit (GPU) is utilized to speed up the calculations due to the vast amount of data which is required to accurately simulate fracture propagation in the snow layers. And due to time limitations, optimizations has been left out. However, it was found out that even though optimization was left out, the GPU is performing about 5x faster than a parallelized CPU version.
In the simulations, data found by Christian Sigrist has been used. Sigrist performed fracture testing on different kind of snow specimen within a laboratory and in the field, and he found crucial parameters for both homogeneous (stable) and heterogeneous (unstable) snow, and our simulation shows that homogeneous snow does not show any fracture propagation, and the heterogeneous snow shows a lot of uncontrolled fractures.
The results in this thesis has been obtained by different kind of visualization methods that have been implemented in this project, where we can in real-time change the visualization method and also have the possibility of pausing/resuming the simulation to obtain more detailed analysis. The types of parameters that is possible to visualize is; snow density, normal stress, shear stress, a so called energy ratio, and the lengths of the fractures.
Place, publisher, year, edition, pages
Institutt for datateknikk og informasjonsvitenskap , 2014. , 194 p.
IdentifiersURN: urn:nbn:no:ntnu:diva-26811Local ID: ntnudaim:11604OAI: oai:DiVA.org:ntnu-26811DiVA: diva2:751719
Elster, Anne Cathrine, Førsteamanuensis