Semester projects
Project proposals Spring Semester 2024
The following is a list of proposals for semester projects offered at the Chair of Risk, Safety and Uncertainty Quantification. To enquire about conducting a project at the Chair, please directly contact the responsible supervisor.
Optimization of truss structures
Supervisor: M. Moustapha
Optimization is a major task in the design of structures where the analyst seeks to reduce the cost while ensuring that some performance criteria are met.
Many approaches have been developed in the literature depending on whether uncertainties are directly accounted for in the design process or not. The goal of this project is to perform a comparative study of various methods for deterministic and reliability-based design optimization (respectively DDO and RBDO).
In this project, students will design a truss structure using Abaqus as main case study. They will then proceed to optimize this structure in various configurations using UQLab, the Chair’s Matlab platform for uncertainty quantification.
Prerequisites: At least one of the following two courses
Additional information
- Group work: Yes (2)
Cross-entropy methods in structural reliability analysis
Supervisor: M. Moustapha
Structural reliability aims at assessing the reliability of a system by evaluating the probability of failure, i.e., the probability that the system fails to fulfil some performance requirements due to uncertainties in the system itself and its environment.
Various methods are available in the literature to efficiently evaluate probabilities of failure. The goal of the project is to investigate cross-entropy based importance sampling.
In this project, the student will code the respective algorithms in UQLab, the Chair’s Matlab platform for uncertainty quantification, and validate them on different examples taken from the literature.
Prerequisites (at least one)
Additional information
- Group work: Yes (2)
Python-based structural design tool
Supervisor: S. Schär
Hand calculation based structural design is still widespread in Civil Engineering practice. This prohibits the use of advanced computational models and makes the design procedure prone to human error.
In this project, the student will extend an existing python-based structural design tool for simple beam structures. The focus of the project lies on improving the interface of the ex-isting design tool and extending its design capabilities.
The implementation of the tool will be done with Jupyter Notebook and Python. The goal is to provide a tool that can be used in practical applications later on.
Prerequisites
- Vertiefung Konstruktion
Additional information
- Group work: Yes (2)
Leja sequences and experimental design for PCE
Supervisor: N. Lüthen
Polynomial chaos expansion (PCE) are a popular surrogate modelling method in the field of uncertainty quantification. The surrogate is constructed based on a limited number of evaluations of the original expensive model (experimental design). In order to construct a reliable surrogate, it is essential to use an informative set of training points.
The goal of this semester project is to implement the recently proposed experimental design method of Leja sequences using the UQ software UQLab, to investigate its potential for surrogate modeling, and to compare its performance to other established experimental design methods.
Prerequisites:
Additional information:
- Group work: No
Physics-informed polynomial chaos expansion
Supervisor: N. Lüthen
Partial differential equations are the mathematical description of physical phenomena such as the deformation of materials or the flow of liquids. Depending on the mesh size, solving them using traditional methods such as finite elements can become computationally very expensive.
The quantity of interest can also be approximated using surrogates such as neural networks or polynomial chaos expansions (PCE). However, the surrogate does not necessarily obey the physical restrictions of the phenomenon such as boundary conditions or conservation laws, unless this is explicitly required during training (physics-informed).
The goal of this project is to implement the recently proposed method of physics-informed PCE using the UQ software UQLab, to compare its performance to sparse PCE, and to analyze its potential and limitations.
Prerequisites:
Additional information:
- Group work: No
Past projects
The following is a list of past semester projects conducted at the Chair.
- Philip Schnabel, Polynomial chaos expansion for dependent inputs, 2022
- Prijanthy Panchadcharam, Extension of python-based design tool, 2021
- Raphael Sieber, Optimization of truss structures, 2021
- Matthias Schneider, Python-based structural design tool, 2021
- Pietro Parisi, Meta-ensembling: Ensemble methods for metamodelling, 2020
- Riccardo Arrigoni, Partial least squares for polynomial chaos expansions, 2019
- Marco Zumstein & Raphael Fässler, Python-based structural design tool - concrete design, 2019
- Minxiang Gao, Reliability-based design optimization of truss structures using surrogate models, 2019
- Lukas Bachmann & Alexander Hammett, Python-based structural design tool - Jupyter notebook, 2019
- Urias Morf & Mingpeng Zhu, Python-based structural design tool - FE code, 2019
- Andrin Kapser, Weighted average Monte Carlo simulation for reliability analysis, 2018
- Elena Giacomazzi, Directional importance sampling method for reliability analysis, 2018
- Viviane Rogenmoser, Reliability-based design optimization of truss structures, 2018
- Florian Schmid, Line sampling for reliability analysis, 2017
- Jochen Rieger & Stephane Tobler, Optimization of truss structures, 2017
- Kleio Sampatakaki, Asymptotic sampling for reliability analysis, 2017
- Matthias Reutimann, Calibration of partial safety factors for assessing the durability of concrete structures, 2015
- Raphael Wegmann, Reliability Analysis with Subset Simulation, 2015
- Charel Eicher, Reliability analysis of ten-story reinforced-concrete building, 2014
- Dimitrios Piskas, Kriging estimation methods for structural reliability, 2013