Examensarbeten för masterexamen
Länka till denna samling:
Browse
Senast publicerade
Visar 1 - 5 av 641
- PostMicromechanics-based Artificial Neural Networks and Transfer Learning for Modeling Short Fibre Reinforced Composites in Automotive Applications(2023) Cheung, Hon Lam; Chalmers tekniska högskola / Institutionen för fysik; Chalmers University of Technology / Department of Physics; Mirkhalaf, Mohsen; Mirkhalaf, MohsenFor the automotive industry, improving efficiency is crucial while weight reduction is one key factor for high efficiency. Short Fiber Reinforced Composites (SFRCs) provide superior performance at reduced weight and are suitable for mass production, making them attractive materials for the industry. However, the mechanical modeling of SFRCs poses challenges. A full field analysis may take multiple trials to generate a proper realization, and subsequent analysis can take hours or days to finish. Furthermore, the mechanical response is influenced by fiber orientation and volume fraction, which can have countless configurations. Therefore, data-driven models for SFRCs have gained popularity. Previous work utilized mean-field analysis results to train a recurrent neural network, aiming to predict elasto-plastic stress response of SFRCs with different strain paths and properties. This study enhanced the mean-field network with a limited amount of full-field data, aiming to improve the network’s prediction accuracy to a full-field level.
- PostControlling Active Clusters Using Wave-Shaped Light Patterns(2023) BERGSTEN, ALFRED; Chalmers tekniska högskola / Institutionen för fysik; Chalmers University of Technology / Department of Physics; Volpe, Giovanni; Callegari, AgneseColloidal systems appear in various contexts. In some of these systems, ther mophoretic forces can arise around otherwise passive particles when they are il luminated, leading to the emergence of complex behaviours. These types of systems has been extensively studied under constant, uniform light where the emergent be haviours are simply activated and deactivated. The aim of this project is to show that the emergent behaviour can not only be activated and deactivated, but also controlled by employing more complex light patterns. The model used in this project includes Brownian motion and thermophoretic forces, with collisions between particles being resolved by a volume exclusion method. The thermophoretic forces are activated by employing travelling wave light patterns to affect the behaviours of different clusters formed as a result of these forces. Two different patterns are then superimposed to show that more complex light patterns can induce more complex behaviours. This study is mostly qualitative in nature and only conducted in simulations. While the parameter space has only been roughly explored and the study needs to be val idated through physical experiments, the results of the project indicate that a more comprehensive exploration of the parameter space for a broader range of clusters can be of interest.
- PostGuiding Column Generation using Deep Reinforcement Learning Trainee and Training Device Optimization(2023) Lindén, Anton; Chalmers tekniska högskola / Institutionen för fysik; Chalmers University of Technology / Department of Physics; Granath, Mats; Wojciechowski, AdamMany optimization problems can be formulated as a Integer Linear Program (ILP), which is an optimization problem that involves minimizing or maximizing a linear objective function subject to linear constraints and integrality requirements. Some examples include train scheduling, airline crew scheduling and production planning. ILP models with an exponentially growing set of variables are often solved using an algorithm known as Column Generation (CG). CG iteratively improves the objective function value by generating new variables, or columns, without considering every possible variable in the ILP model. This thesis was performed together with Jeppe sen, a Boeing company, and investigated the possibility of using Deep Reinforcement Learning (DRL) to generate new variables in CG for a scheduling problem for air line pilots. Results show that it is possible to teach an agent a policy that slightly improves the quality of the generated variables in this specific problem. However, it is still unclear whether or not the benefits of using DRL outweighs the extra effort of setting up and training an agent.
- PostDetailed Microstructure and the Influence of Post-Treatment on CVD TiAlN Wear Resistant Coatings(2023) Mead, Monica Audrey; Chalmers tekniska högskola / Institutionen för fysik; Chalmers University of Technology / Department of Physics; Halvarsson, Mats; Schmitz, Guido; Bäcke, OlofThe deposition of TiAlN by chemical vapour deposition (CVD) for the application as wear-resistant coatings for cutting tools has been the subject of research interest for numerous years, yet a comprehensive understanding of its growth mechanism and intricate microstructural characteristics remains incomplete. Furthermore, although the positive effect of blast-treatment on the stress state of wear-resistant coatings has received attention in scientific studies, there remains a relative sparsity of research investigating its influence on the microstructure. In this work, the detailed microstructure of nano-lamellar low-pressure CVD TiAlN coatings on cemented carbide substrates and the influence of post-treatment on the microstructure was investigated by scanning electron microscopy (SEM), scanning transmission electron microscopy (STEM) and transmission Kikuchi diffraction (TKD). SEM imaging revealed two distinct types of surface morphologies corresponding to specific grain orientations unveiled by TKD on thin foil cross-sections. As previously shown [1], pyramidal surface morphology is connected to growth in a <111> direction, additionally, a ridge-like surface morphology was connected to growth along <110> directions. Both growth directions enable fast growth with low-energy {100} facets. Furthermore, truncation of pyramidal and ridge-like surface morphologies was observed. A growth mechanism was proposed aiming to explain the characteristics of the truncated grain morphology. Here, strongly facetted surfaces emerging on the truncated grains increase the surface reaction kinetics, leading to an increased Al/Ti-ratio in the core region of the grain compared to its sides with large {100} facets. Blast-treatment of the TiAlN coatings with corundum particles led to plastic deformation up to a few hundred nanometres in depth. Impact jet wear introduced a high defect density and crack or void formation beneath the surface. Plastic deformation of the near-surface region of the coating led to the bending or disappearance of the Ti- and Al-rich lamellae typical in TiAlN coatings prepared by CVD and a continuous lattice rotation in three dimensions. The available data indicates that grains which have grown along one of their <110> directions exhibit a more pronounced lattice rotation when compared to grains that have grown along a <111> direction.
- PostDrone safe to launch system using machine learning(2023) Sahlberg, Elina; Krook Willén, Björn; Chalmers tekniska högskola / Institutionen för fysik; Chalmers University of Technology / Department of Physics; Volpe, Giovanni; Nilsson, VictorSvenska Sjöräddningssällskapet has partnered with Infotiv to develop a prototype drone and launcher system for search-and-rescue missions at sea. In this thesis project, we investigate the possibility of automating parts of the launch sequence of a seaborne surveillance drone. The main goal is to train a neural network model to use a camera feed to determine if it is safe or not to launch the drone in a given direction, and then integrate this solution with the graphic user interface through an application programming interface. Monocular depth estimation using transfer learning and the KITTI data set is evaluated. The KITTI data set does not contain maritime scenery, leading to an unsatisfactory monocular depth estimation model. U-net and convolutional neural network models are trained on the MaSTr1325 data set, which contains semantically segmented maritime imagery. We collect additional data for the semantic segmentation models and create a post-processing step that evaluates if it is safe to launch or not. These models yielded satisfactory results, and the convolutional neural network will be used by the drone operator as an extra safety measure during launch.