Eldorado - Repository of the TU Dortmund
Resources for and from Research, Teaching and Studying
This is the institutional repository of the TU Dortmund. Ressources for Research, Study and Teaching are archived and made publicly available.
Communities in Eldorado
Select a community to browse its collections.
Recent Submissions
Directional importance sampling for dynamic reliability of linear structures under non-Gaussian white noise excitation
(2024-12-10) Zhang, Xuan-Yi; Misraji, Mauricio A.; Valdebenito, Marcos A.; Faes, Matthias G. R.
Reliability analysis of dynamic structural systems and its implications for structural design have garnered increasing attention. Sample-based methods prove insensitive to the dimension of the probability integral. Nontheless, a substantial number of realizations is necessary for estimating small failure probabilities, resulting in time-consuming computations. Recently, the Directional Importance Sampling (DIS) was introduced for linear structural systems subjected to Gaussian loads, showcasing the ability to accurately estimate small failure probabilities with a reduced number of simulations. However, this Gaussian assumption on the load makes the method inapplicable for realistic loading scenarios as they might be of non-Gaussian nature. This contribution develops the DIS method for linear structural systems subjected to loading characterized as non-Gaussian white noise. To take the advantage of both linearity in physical space and simplicity of Gaussian space, directional importance sampling is conducted in Gaussian space and the failure probability is estimated with the aid of physical space. The information is dynamically exchanged between physical and Gaussian spaces with the aid of normal and inverse-normal transformation techniques. The whole procedure of the developed DIS method is straightforward, and it provides an explicit estimator of the failure probability. The application of the developed DIS method is presented through three examples, illustrating its accuracy and efficiency for dynamic reliability analysis.
A unified approach to finite element simulation of elastoviscoplastic fluid flows
(2025) Saghir, Muhammad Tayyab Bin; Turek, Stefan; Sokolov, Andriy
In this study, we present the development of a two-dimensional finite element solver for the simulation of fluids exhibiting both elastic and plastic constitutive properties. We achieve this by combining the constitutive models of the Oldroyd-B model for viscoelastic fluids and the Papanastasiou model for Bingham fluids within a single Eulerian numerical framework. Our aim within this approach is to approximate the given velocity, pressure, and elastic stresses. We employ a higher order finite element method for the velocity-stress approximation and a discontinuous pressure element. This specific element pair has proven highly effective for accurately capturing the behavior of both Oldroyd-B and Bingham fluids, including non-linear viscosity functions. Our study consists of three main steps. Firstly, we validate the numerical results for each module component of the constitutives to ensure the accuracy of the approximations and calculations. This step is crucial for establishing the reliability and robustness of our approach. Subsequently, in the second step, we apply the solver to simulate elastoviscoplastic fluid behavior in a porous medium and flow around a cylinder benchmark. Finally, we compute the drag and lift for the flow around the cylinder simulation to validate our numerical methodology against established benchmark results. Through the analysis of the above mentioned benchmark problems, we highlight the potential of the proposed solver to reliably capture the complex interplay of elastic, viscous, and plastic effects in non-Newtonian fluid dynamics.
A concept for personalised CT dosimetry using methods of machine learning
(2025) Kuhlmann, Marie-Luise; Kröninger, Kevin; Lühr, Armin
In the context of increasing number of computer tomography (CT) examinations, also the importance for fast and patient specific dose assessment grows. Machine learning (ML) offers a promising approach to achieve fast and user-friendly organ dose assessment in clinical CT workflows. Previous studies have shown that neural networks can reproduce Monte Carlo (MC)-calculated organ doses with reasonable accuracy; however, the representation of the radiation field properties, defining the spatial and energy distribution of the X-ray beam during acquisition, and the characterisation of the training data distributions vary widely. Furthermore, a comprehensive uncertainty assessment over the entire dosimetry process has not been addressed. This thesis presents a framework for personalised CT dosimetry based on ML methods, combining a validated newly implemented particle source for MC simulations, measurement-based radiation field characterisation and systematic uncertainty assessment at every stage of the process. In addition, the work investigated the influence of training data composition in regarding the role of synthetic and real patient geometry data. The proposed approach achieves accuracy comparable to previous studies, while providing a complete uncertainty assessment methodology.
Bridging scales in digital pathology
(2025) Hörst, Fabian; Kleesiek, Jens; Schneider, Matthias; Rückert, Daniel
Histopathology is a cornerstone of disease diagnosis and treatment, traditionally relying on manually assessing tissue specimens under a microscope. However, the advent of slide scanners to produce digital tissue representations, so-called whole-slide images (WSI), has enabled computational pathology to perform quantitative and automated tissue analysis. Current developments in Artificial Intelligence, particularly Deep Learning, have accelerated the progress in this field. This thesis proposes a comprehensive Deep Learning pipeline for quantitative histopathological image analysis, integrating WSI preprocessing, algorithm development for tissue and cell-level segmentation, and clinical application in an end-to-end workflow. The approach not only improves the quantitative evaluation of WSI but also extracts diagnostic and prognostic markers while automatically characterizing tissue dynamics through morphological tissue features. Segmenting entire tissue sections into classes like tumorous or non-tumorous requires the consideration of global tissue patterns as well as local cell morphologies. Following this, we introduce the Memory Attention Framework that can be incorporated into any encoder-decoder segmentation architecture. This framework enables the adaptive incorporation of tissue context during fine-grained local segmentation. The method was evaluated on two public datasets (breast, liver) and an internal kidney cancer dataset, demonstrating superiority over non-context and multiscale segmentation approaches. Notably, the approach reduced the number of false-positive tumor regions. Building on this, we applied the framework to a pancreatic cancer cohort consisting of 400 internal and 182 external patients to quantify the tumor microenvironment and correlate it with patient outcomes. In doing so, we were able to stratify patients into two risk groups based on tissue composition and spatial tumor-stroma distribution, which showed significant (p < 0.05) differences in their survival probabilities. Next to tissue analysis, segmentation on the cellular level is crucial to uncover the cellular composition of tissue samples. While convolutional neural networks have been extensively used for this task, we evaluate the capabilities of Transformer-based networks and incorporate so-called foundation models to improve accuracy compared to existing solutions. The proposed CellViT and CellViT++ models have proven to achieve State-of-the-Art results on several benchmark datasets, covering a broad spectrum of tissue types and cell classes, bringing cell segmentation solutions closer to clinical practice. The models require minimal data for fine-tuning and exhibit remarkable zero-shot cell segmentation quality. This capability allows for a considerably faster adaptation to new research hypotheses without the need for extensive development time. In summary, this work presents Deep Learning techniques for quantifying tissue at both the macro and micro levels, enhancing diagnostic workflows, and identifying prognostic markers.
Amtliche Mitteilungen der Technischen Universität Dortmund Nr. 3/2026
(Technische Universität Dortmund, 2026-02-02)