University of Twente Student Theses
Two data-driven approaches for predicting bed levels in 3D for the Waal River
Laar, L.I. van (2024) Two data-driven approaches for predicting bed levels in 3D for the Waal River.
|
PDF
10MB |
| Abstract: | The Dutch waterways, like the Waal River, are among Europe’s busiest navigation routes. Dredging is conducted to keep these waterways navigable. Currently, short-term dredging decisions rely on real-time bed level data. Providing forecasts on bed levels can support this decision-making process by identifying potential future bottlenecks, deeper locations for sediment disposal and areas where subsequent measurements may not be necessary. This study aimed to obtain a two-weekly bed level prediction in 3D for the Waal River, by developing and comparing two data-driven approaches. This included the development of a data-driven dune migration model, which only considers the horizontal displacement of bed features, alongside improving the bed level prediction accuracy of the machine learning model TrajGRU. The initial machine learning model TrajGRU produced highly blurred predictions and little of the bedform patterns were captured accurately. To improve these machine learning predictions, five data preprocessing and three different loss functions were tested in the machine learning model. The dune migration model showed to outperform the improved machine learning model. The improved machine learning model still provided blurred predictions, resulting a relatively high maximum bed level error of 0.77m. The dune migration model had a substantially smaller maximum bed level error of 0.30m, the locations of the maximum bed levels were captured more accurately, and the bedform patterns were captured more accurately as well. Even when manually increasing the bed level spread of the machine learning prediction, the dune migration model outperformed the machine learning model. Thus, the potential of a complicated machine learning model like TrajGRU for predicting bed levels in the Waal River shows to be small and the dune migration model proves more promising. Future research could focus more on the dune migration model and combine this method with an approach suitable for predicting vertical bed level changes as well. |
| Item Type: | Essay (Master) |
| Clients: | HKV, Netherlands |
| Faculty: | ET: Engineering Technology |
| Subject: | 56 civil engineering |
| Programme: | Civil Engineering and Management MSc (60026) |
| Link to this item: | https://purl.utwente.nl/essays/98838 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page