Modeling and simulation of granular matter has important applications in both natural science and industry. One widely used method is the discrete element method (DEM). It can be used for simulating granular matter in the gaseous, liquid as well as solid regime whereas alternative methods are in general applicable to only one. Discrete element analysis of large systems is, however, limited by long computational time. A number of solutions to radically improve the computational efficiency of DEM simulations are developed and analysed. These include treating the material as a nonsmooth dynamical system and methods for reducing the computational effort for solving the complementarity problem that arise from implicit treatment of the contact laws. This allow for large time-step integration and ultimately more and faster simulation studies or analysis of more complex systems. Acceleration methods that can reduce the computational complexity and degrees of freedom have been invented. These solutions are investigated in numerical experiments, validated using experimental data and applied for design exploration of iron ore pelletising systems. / <p>This work has been generously supported by Algoryx Simulation, LKAB (dnr 223-</p><p>2442-09), Umeå University and VINNOVA (2014-01901).</p>

Image segmentation through neural networks and deep learning have, in the recent decade, become a successful tool for automated decision-making. For Luossavaara-Kiirunavaara Aktiebolag (LKAB), this means identifying the amount of slag inside a furnace through computer vision.  There are many prominent convolutional neural network architectures in the literature, and this thesis explores two: a modified U-Net and the PSPNet. The architectures were combined with three loss functions and three class weighting schemes resulting in 18 model configurations that were evaluated and compared. This thesis also explores transfer learning techniques for neural networks tasked with identifying slag in images from inside a furnace. The benefit of transfer learning is that the network can learn to find features from already labeled data of another context. Finally, the thesis explored how temporal information could be utilised by adding an LSTM layer to a model taking pairs of images as input, instead of one. The results show (1) that the PSPNet outperformed the U-Net for all tested configurations in all relevant metrics, (2) that the model is able to find more complex features while converging quicker by using transfer learning, and (3) that utilising temporal information reduced the variance of the predictions, and that the modified PSPNet using an LSTM layer showed promise in handling images with outlying characteristics.