University of Twente Student Theses

Accelerating a detailed 1D2D hydrodynamic model : Optimising the calculation time of a detailed 1D2D Hydrodynamic model by resolving numerical instabilities and applying numerical simplifications

Diggelen, P. van (2024) Accelerating a detailed 1D2D hydrodynamic model : Optimising the calculation time of a detailed 1D2D Hydrodynamic model by resolving numerical instabilities and applying numerical simplifications.

PDF
3MB

Abstract:	Large parts of the Netherlands are located below sea level and face the threat of fluvial flooding. On the other hand, the intensity of extreme precipitation is expected to increase leading to a larger probability of pluvial flooding. To successfully incorporate potential future threats in the current water management strategy, it is important to be able to generate accurate predictions of the effects of these events. Hydrodynamic simulation models provide a solution to see the effects of extreme events on the water system but have long computation times, making them unsuitable for real-time use and evaluation of a large set of events. The objective of this study is to get insight into methods that can accelerate these hydrodynamic simulation models to make them better suitable for application in water management practices. The research is executed with the help of a case study from HydroLogic. This case study includes a highly detailed D-HYDRO model of the study area of Hoogheemraadschap de Stichtse Rijnlanden (HDSR), which is a Dutch water board. HDSR will use the model in the future to identify bottlenecks in the water system. In this procedure, several thousands of different rainfall events must be simulated. At the start of the research, the model computed 15 times slower than reality, making it unsuitable for simulating all these events. Two approaches are researched that can accelerate the model. First, the effect of numerical instabilities on the computation time is examined. Numerical instabilities arise if numerical errors grow and the output starts to diverge. This limits the time step that can be taken, resulting in long computation times. Indicators for numerical instabilities are high flow velocities, water depths of zero meters and waterways that fill or empty quickly. Several of these locations are present in the original model of the case study and are examined. This resulted in a list of seventeen causes of numerical instabilities. Most of these numerical instabilities are related to poor data quality and incorrect model construction. By obtaining missing data and improving the model construction, most numerical instabilities are solved. As a consequence, the computation time is reduced by 96%. Furthermore, by correcting the data and enhancing the model construction, the optimised model is also more accurate. Since resolving numerical instabilities can drastically advance the computation time while improving the accuracy, it is advised to always check for numerical instabilities in detailed 1D2D hydrodynamic models and solve potential issues. Secondly, numerical simplifications are implemented in the optimised model to obtain surrogate models. Surrogate models approximate the detailed hydrodynamic model and are therefore faster, but come with a loss in accuracy. The tested numerical simplifications constitute a reduced 2D grid resolution, a reduction in the number of 1D calculation points and an increased maximum Courant number. Reducing the 2D grid resolution and number of 1D calculation pointsreduces the computation time by 2% to 57% but also introduces significant errors in the simulated water level of more than 5 cm. Increasing the maximum Courant number is the most efficient and can reduce the computation time of the optimised model by more than 90% while having an error in the water level that is less than 0.5 cm. A maximum Courant number of 5, the default is often 0.7 in detailed hydrodynamic models, seems to be the best in the trade-off between computation time and accuracy. Overall, it can be concluded that resolving numerical instabilities is an effective first step in reducing the computation time of complex 1D2D hydrodynamic models since the accuracy will often improve as well during this step. If a larger reduction in computation time is required, the maximum Courant number can be increased between 1 and 5 depending on the required accuracy and computation time savings for the specific project.
Item Type:	Essay (Bachelor)
Faculty:	ET: Engineering Technology
Programme:	Civil Engineering BSc (56952)
Link to this item:	https://purl.utwente.nl/essays/104041
Export this item as:	BibTeX EndNote HTML Citation Reference Manager

Show download statistics for this publication

Repository Staff Only: item control page