Coming dissertations at Uppsala university
-
The Thin-foil Proton Recoil neutron spectrometer for DT plasmas
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-526257
Recent advancements in plasma physics are intensifying the demand for advanced diagnostic techniques in fusion research, particularly for the upcoming ITER fusion reactor. The ITER fusion reactor is projected to be ten times more powerful than its predecessors, imposing higher constraints on operational parameters. To meet ITER's requirements, such as the fuel ion ratio nt/nd and fuel ion temperature Ti, a High Resolution Neutron Spectrometer System (HRNS) has been proposed.
This thesis focuses on the Thin-foil proton recoil (TPR) spectrometer, an integral part of the HRNS, with an emphasis on its application and validation within the ITER context. The research encompasses two main areas: spectrometer simulations and experimental validation. Through a combination of custom transport code and Geant4 simulations, the study investigates the optimization of the TPR spectrometer's design in terms of efficiency and energy resolution. Additionally, selected design performance under ITER-like conditions has been investigated. These simulations are critical in assessing the spectrometer's capabilities and limitations during operation at ITER. Subsequent experimental validation, conducted using a DT neutron generator and a TPR spectrometer prototype, verified the existing simulation framework in terms of energy resolution and background discrimination methods.
We examined a Tandem neutron spectrometer, used in fusion plasma diagnostics at JET to further investigate TPR spectrometer diagnostic possibilities. Tandem spectrometer was operational during JET's first DT campaign, the spectrometer shares the neutron detection principles of the TPR. The fuel ion ratio nt/ntot was determined using the Tandem data together with inputs from PENCIL or TRANSP, for previously not analysed JET discharges. Our findings indicate that estimation of nt/ntot is feasible using either PENCIL or TRANSP. Furthermore, the research demonstrates that TPR based neutron spectrometers can be effectively used in fuel ion ratio determination.
In conclusion, this research significantly advances fusion plasma diagnostics. It validates the TPR spectrometer's design in terms of energy resolution and efficiency for ITER, predicting a signal-to-background ratio of approximately 550 and a maximum count rate of 120kHz. The results from the TPR prototype experiment, replicated with the Geant4 simulation, along with comparative analysis with the JET's Tandem spectrometer, highlight the TPR spectrometer's broad applicability in fusion diagnostics, marking a major advancement in the field.
-
Understanding and predicting methane formation and bubble emission from lakes
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-526222
The role of lakes and reservoirs as significant emitters of the potent greenhouse gas methane (CH4) to the atmosphere is well established, but several uncertainties remain particularly due to knowledge gaps in the relationship between CH4 dynamics and sediment characteristics, which limit our capabilities to provide robust estimates of CH4 fluxes from lakes.
In my thesis, I investigated CH4 formation and CH4 bubble emissions (ebullition) in lakes, with a particular focus on sediment characteristics that control these processes, using different approaches such as laboratory experiments, field surveys and process-based modelling.
Sediment incubation experiments have revealed that CH4 formation rates can be predicted based on the age and total nitrogen content of the sediment. This relationship holds true across a wide range of sediment types found in various climates, ranging from alpine tundra to tropical regions, and can be effectively estimated by a common empirical model. Additionally, the supply and the quality of organic matter play a crucial role in determining the extent of CH4 formation in the sediment. Moreover, frequent additions of organic matter to surface sediment enhance the speed of CH4 formation rates. Importantly, the relationship between organic matter supply, its quality and frequency of addition with CH4 formation rates can be predicted with a logistic model.
Field surveys conducted in a small eutrophic lake revealed that the spatial variability in CH4 ebullition is regulated by site-specific sediment characteristics: high CH4 ebullition rates were observed in areas characterized by elevated organic matter density in surface sediment and temporary sediment deposition, although no correlation was found with sediment accumulation rates.
The use of a 1D model to simulate CH4 fluxes from a lake demonstrated its ability to simulate fairly well the whole-lake average CH4 ebullition fluxes observed in the field, despite experiencing some interannual variability in model performance. However, when dividing the lake into smaller sections to simulate the observed longitudinal spatial variability in CH4 ebullition, the model systematically overestimated the fluxes.
Overall, this thesis highlights the critical role of sediment characteristics and sedimentation regime on regulating the CH4 formation and ebullition fluxes, providing advancements in understanding CH4 dynamics in lakes.
-
On Deep Learning for Low-Dimensional Representations
Link: http://urn.kb.se/resolve?urn=urn:nbn:se:uu:diva-526130
In science and engineering, we are often concerned with creating mathematical models from data. These models are abstractions of observed real-world processes where the goal is often to understand these processes or to use the models to predict future instances of the observed process. Natural processes often exhibit low-dimensional structures which we can embed into the model. In mechanistic models, we directly include this structure into the model through mathematical equations often inspired by physical constraints. In contrast, within machine learning and particularly in deep learning we often deal with high-dimensional data such as images and learn a model without imposing a low-dimensional structure. Instead, we learn some kind of representations that are useful for the task at hand. While representation learning arguably enables the power of deep neural networks, it is less clear how to understand real-world processes from these models or whether we can benefit from including a low-dimensional structure in the model.
Learning from data with intrinsic low-dimensional structure and how to replicate this structure in machine learning models is studied within this dissertation. While we put specific emphasis on deep neural networks, we also consider kernel machines in the context of Gaussian processes, as well as linear models, for example by studying the generalisation of models with an explicit low-dimensional structure. First, we argue that many real-world observations have an intrinsic low-dimensional structure. We can find evidence of this structure for example through low-rank approximations of many real-world data sets. Then, we face two open-ended research questions. First, we study the behaviour of machine learning models when they are trained on data with low-dimensional structures. Here we investigate fundamental aspects of learning low-dimensional representations and how well models with explicit low-dimensional structures perform. Second, we focus on applications in the modelling of dynamical systems and the medical domain. We investigate how we can benefit from low-dimensional representations for these applications and explore the potential of low-dimensional model structures for predictive tasks. Finally, we give a brief outlook on how we go beyond learning low-dimensional structures and identify the underlying mechanisms that generate the data to better model and understand these processes.
This dissertation provides an overview of learning low-dimensional structures in machine learning models. It covers a wide range of topics from representation learning over the study of generalisation in overparameterized models to applications with time series and medical applications. However, each contribution opens up a range of questions to study in the future. Therefore this dissertation serves as a starting point to further explore learning of low-dimensional structure and representations.