Research Overview

My research focuses on data analysis, modeling, prediction, and validation using experimental data. I am interested in advanced data analysis techniques, uncertainty and sensitivity analysis, and the behavior of materials and nuclear fuel, particularly in dry storage. A list of my publications can be found here.

Fuel Rod Behaviour
One of the challenges of long-term dry storage of used nuclear fuel is the potential for degradation of the fuel cladding, which can lead to the release of radioactive materials into the cask. Cladding degradation can occur due to a variety of factors, including corrosion, creep, and hydride formation. The conditions under which the fuel is stored, including temperature, humidity, and the presence of other materials, can also affect the rate of degradation. In the research project LEDA (Long-term Experimental Dry Storage Analysis) we are investigating the cladding performance under dry interim storage conditions, with a focus on the significance of hydrogen. The experiments will involve studying different fuel rod segments under prototypical conditions representative of dry storage in Germany. The goal is to use the data generated to extend and validate models and methods for predicting fuel rod integrity, and to review and analyze the criteria for the exclusion of systematic cladding failure in the context of hydrogen behavior over storage times of more than 40 years.

Muon Radiography
Cosmic-ray muons can be used to non-invasively image spent nuclear fuel in sealed dry storage casks. By analyzing the scattering data of the muons after they pass through the cask, it is possible to determine the materials that they have penetrated. We use simulations to investigate the feasibility of detecting single missing fuel rods in a sealed cask using this technique. We simulate a vertically standing cask model loaded with fuel assemblies from a pressurized water reactor and muon detectors above and below the cask. By analyzing the scattering angles we find that missing rods can be reliably identified in a reasonable time period, depending on their position in the assembly and cask, and on the angular acceptance criterion of the incoming primary muons. We are working to design an optimal experimental setup for use in an interim dry storage facility. Our goal is to experimentally prove our theoretical predictions within the next 3-5 years.

Computational Benchmarks
In the field of computational science, benchmarks are essential tools for evaluating and comparing the performance of different systems and components. Whether seeking to optimize an existing system or design a new one, a thorough understanding of benchmarking techniques is crucial for achieving reliable and meaningful results. As a researcher, I have participated in the design and analysis of benchmarks for a variety of applications, including criticality safety, cladding behavior, and temperature prediction. These benchmarks are used to assess the accuracy, speed, and other characteristics of algorithms and software programs, as well as to compare the performance of different computer systems and architectures.

Calculation Methods and Data Analysis
In my research, I use advanced statistical methods such as Monte Carlo techniques and Bayesian statistics to analyze large heterogeneous data sets. These data sets include a variety of experimental and documented data on characteristics of used nuclear fuel rods, including criticality safety, cladding behaviour, and loading documentations of casks. For example in criticality safety assessments, it is necessary to accurately predict the effective neutron multiplication factor (keff) with a sufficient safety margin. This prediction is made using validated calculation methods and computer codes (such as criticality codes) to calculate keff for a specific case. I worked on how to generate (e.g. link, link, link) and handle integral experimental covariance matrices in code validation procedures. The determination of correlations is also influenced by how experimental data is translated into calculation models. Correlated data can occur when different experiments share parts of their experimental setups, measurement systems, or other relevant parameters. We analyzed some publicly available experimental data sets (e.g. link, link) in greater detail. Some of our findings went into a software we developed used now by the German Federal Office of the Safety of Nuclear Waste Management (link).