The primary flood defences (dikes) of the Netherlands need to be sufficiently safe. Safety assessments are thus regularly carried out by water boards and Rijkswaterstaat, most recently for the safety norms of 2050. It is important that safety assessments are conducted thoroughly and reliably. Imagine an incorrect assessment; a safe dike is assessed as unsafe and millions of euros are spent on an unnecessary reinforcement, or an unsafe dike is assessed as safe and lives are at risk during extreme river discharges. However, conducting the assessment for all dike sections in the Netherlands requires the development of complex modelling software, which takes much input data and takes a long time to compute.
This year, a new round of safety assessments is starting, which will last until 2035. As a part of the initial phase of the project, research into new methodologies for the assessment of dikes is included, such as machine learning techniques. Machine learning models can find patterns in large amounts of data, allowing them to make predictions for similar but new situations. The Ministry of Infrastructure and Water Management is curious if and how these techniques can be used for safety assessments.
The objective of the project is to research how a machine learning model for the assessment of dike safety can be developed, utilizing data of river dikes and previous assessments in the Netherlands. You will dive into the relevant characteristics of river dikes for failure mechanisms such as overtopping, piping and macro instability. You will learn and discover information processing techniques required for machine learning, and apply them to develop your own models. You will evaluate the models and see what you and the ministry can learn from them, giving out an advice on the suitability of these techniques for supporting such important policy decisions.
The assignment leaves room for finding out which aspects of safety assessments are to be included in the scope. Data will be made available by the Ministry of Infrastructure and Water Management, Rijkswaterstaat and water boards of the Netherlands. Knowledge of programming is required, preferably in Python, though Matlab offers some tools as well.
References
- Zhu et al. (2023) "Research on Safety Evaluation of Yangtze River Embankment Based on Fuzzy Neural Network," 2023 3rd International Conference on Consumer Electronics and Computer Engineering, doi: https://doi.org/10.1109/ICCECE58074.2023.10135298
- Flynn et al. (2021). “Data-Driven Model for Estimating the Probability of Riverine Levee Breach Due to Overtopping”. Journal of Geotechnical and Geoenvironmental Engineering, 148(3). url: Data-Driven Model for Estimating the Probability of Riverine Levee Breach Due to Overtopping
- Jamalinia et al. (2021). A Data-Driven Surrogate Approach for the Temporal Stability Forecasting of Vegetation Covered Dikes. Water, 13(1), 107. doi: https://doi.org/10.3390/w13010107