Master Assignment
Avalanches and land sliding are hardly “predictable” and “controllable” issues that result in severe damages and losses. Therefore, understanding the possible causes and technology development for their prevention and control are challenging yet indispensable tasks for engineers and scientists. In this project, experimental lab-scale data will be used to train the computer to predict the avalanche behavior of granular media for any combination of particle sizes and viscosities of the interstitial liquid that might be expected in an avalanche or landslide.
Fig. 1 – Schematic of machine learning framework for avalanche prediction. Experimental data (left) to be injected into the machine learning algorithm (middle), and schematic of an avalanche (right).
A lab-scale prototype was used to mimic wet granular avalanches and study the effect of viscosity and particle size on the avalanche behavior. In this project, we will feed the obtained experimental data such as the angle of repose, velocity gradient and avalanches amplitude (obtained from a large amount of post-processed images) to the machine learning algorithm and test its capability for fast and accurate prediction of avalanche occurrence and flow behavior for different liquid viscosities and particles sizes. Experiments may also be performed to obtain additional relevant data to increase the accuracy of the predictive algorithm.
Tasks to be done:
- Basic postprocessing
- Basics of avalanches
- Literature review on learning and prediction of avalanches
- Use extracted features that are already extracted and feed the network with that
- Reduce the total data set to something memory efficient (“raw” data set)
- Use “raw” data set to predict movements of particles in time
Contact Person(s)
