This year, the Netherlands introduced new risk standards for flood protection (Delta Programme 2017) to be implemented by 2050. This will be the world’s first national flood standard based on the quantitative estimation of flood probability, including dyke failure.
A key component of the new standards is the focus on soil properties and dyke failure risks.
Geoscientist Yajun Li and his colleagues at the Delft University of Technology studied the Dutch earth structures and analysed their properties and failure threats. Their work led to the design of a new risk assessment framework that defines ways of evaluating the stability and failure consequences of the Dutch soil constructions.
For assessing the reliability of longer dykes, typically 1000 computing jobs are needed to ensure a converged solution. High-Throughput Computing is ideally suitable for such a task and can reduce the computational time dramatically. Yajun Li
The team, led by Michael Hicks and Philip Vardon, wanted to analyse the soil properties of dykes and see if they can predict their reliability (i.e., the likelihood that a dyke is functional for a given period of time). To do that, they used a 3D analysis model, also called the 3D random finite element method (RFEM).
Then they tested the model using the computing resources provided by SURFsara (the Dutch national distributed computing e-Infrastructure) and the help of Natalie Danezi, grid & cloud services advisor.
“For assessing the reliability of longer dykes, typically 1000 computing jobs are needed to ensure a converged solution. As each job is able to be run independently on a single serial computer, High-Throughput Computing is ideally suitable for such a task and can reduce the computational time dramatically”, explains Li.
For his research calculations, Li used 500.000 core hours of compute time, 500 GB of disk storage and 500 GB of tape available via the projects.nl virtual organisation, supported by SURFsara.
“In the Monte Carlo simulation of a dyke, thousands of realisations of soil properties that satisfy the point and spatial statistics of the soil often need to be generated using a random field generator, so that the designed dyke reliability and risk consequence can be assessed. (…) Thus, running the analysis serially on a desktop computer would be unfeasible”, says Li.
One of the main results of Li’s work is that the longer the dyke, the larger the failure probability. He therefore recommends that safety standards are made according to different scales – national or regional. The results of his work are detailed in his PhD thesis, published in February 2017.
The risk assessment framework defined in Yajun Li’s PhD thesis includes the quantification of dyke reliability and risk in three dimensions, with a focus on the length effect of dykes; the analyses of large 3D dyke stability via High-Throughput Compute resources; the role of diversity in the collapse processes of dams and general consequences for the population that the structure protects.
Yajun Li used 500.000 core hours of compute time, 500 GB of disk storage and 500 GB of tape.
These resources are supported by SURFsara, the Dutch national distributed computing e-Infrastructure, and made available via the projects.nl virtual organisation.
Li, Y. 2017. Reliability of long heterogeneous slopes in 3d: Model performance and conditional simulation. Doctoral Thesis. doi: 10.4233/uuid:24ee6615-2555-4b64-8950-a77c9d969806 (full text, open source)