Maintaining the channels and infrastructure for large commercial ports is costly and complex. Port authorities traditionally rely on bathymetry surveys to guide their dredging operations, which limits their ability to be proactive. This research aims to develop machine learning (ML) methods to predict sedimentation rates (SR) by analysing patterns in hydrological and meteorological (hydro-meteo) conditions and dredging data. By investigating the possibility of SR prediction models, this research can contribute to efficient maintenance operations.

The study focuses on the Botlek, a harbour in the Port of Rotterdam situated in the Rhine-Meuse estuary. Three data types are integrated to capture the dynamic interplay of saline and riverine factors within the estuary. The data consists of Multibeam bathymetry surveys, hydro-meteo variables (such as salinity, river discharge, and tidal variation), and dredging logs. The surveys provide the net sediment accumulation, which is assumed to result from a specific period of hydro-meteo conditions and dredging. Due to the limited availability of surveys, the number of features had to be managed. To avoid having more features than samples, the hydro-meteo variables are aggregated into daily, weekly, and
monthly means.

The ML algorithms evaluated in this research are Linear Regression (LR), Random Forest Regression (RFR), and Support Vector Regression (SVR). The feature importance scores from the RFR and the accuracy on small datasets of the SVR were decisive in this selection. All algorithms were tested and refined over multiple development phases to determine their predictive accuracy and robustness across different data configurations. It was found that the ML algorithms can reasonably predict SR. In addition to dredging data, incorporating hydro-meteo variables enhanced the predictive accuracy and consistency. Specifically, a feature set of dredging volumes, salinity, discharge, and tidal variation improved model performance. Training the model on dredging data only resulted in an inferior performance. The monthly aggregation of these hydro-meteo data was the most beneficial configuration. Among the tested algorithms, SVR consistently outperformed both LR and RFR, with its best configuration achieving a mean R2 score of 0.69 over four different dataset splits.

The ML models do not have to be able to predict sediment volumes with a small confidence interval to be practical for the port authorities. Instead, the models should be able to predict trends in sediment accumulation and provide actionable insights to enable proactive dredging. The research shows promising results, as most predictions fall within the acceptable range. However, predictive accuracy strongly varied across the different dredging areas in Botlek. Further development is required to provide real-time, reliable SR predictions before the models can be integrated into the maintenance operations of ports.

Currently, inland waterway safety assessments rely heavily on historic accident data and expert opinions, often lacking comprehensive qualitative and quantitative information. Addressing this gap, this thesis introduces a method based on Automatic Identification System (AIS) data to identify anomalous vessel behaviour for enhancing safety assessments.

The proposed methodology establishes definitions of normal vessel behaviour and identifies deviations from these norms as anomalies. Analysing vessel trips recorded in AIS data logs, various features—including speed, acceleration, direction, manoeuvrability, and positional attributes—are extracted to define vessel behaviour.

The Uniform Manifold Approximation and Projection (UMAP) algorithm reduces the multidimensional features into a two-dimensional embedding to condense the vessel behaviour into a manageable form. This reduction technique preserves the inherent behaviour while representing similarities within a 2-dimensional space. Subsequently, the K-means clustering algorithm is applied to group vessels displaying similar behavioural patterns. The hyperparameters for clustering are determined using the elbow method and multiple scoring metrics.

Application of this methodology to the Moerdijkbrug at the Hollands Diep and Schellingwouderbrug at the IJ reveals clusters with similarities in vessel direction and paths. Several atypical patterns were observed and further investigated, analysing less than 1% of the data set in both cases, revealing two distinct patterns classified as probable accidents in the IJ case. These findings demonstrate the potential of the proposed method in identifying specific vessel behavioural anomalies with implications for safety assessment on inland waterways.

This thesis introduces a new algorithm for optimising shipping routes within a dredging project. Highly dynamic and time-dependent hydrodynamic features influence shipping routes. Due to the complex interactions between the horizontal tide, vertical tide, stratifying forces, wind-driven forces, and limited water-depth, shipping routes were previously only optimised for large scale routes (order of 1000 km). This study presents an algorithm that can optimise shipping routes that are influenced by these small scales (order of 10 km) hydrodynamic features. This algorithm uses graph theory to solve for the time-dependent fastest path between start and destination. Graph theory searches for the optimal path through a set of nodes that are connected with edges. This study uses the time-dependent shortest path algorithm which accounts for the FIFO-criteria (Waiting criteria) and can solve the non-convex nature of the problem

The input of this algorithm is a hydrodynamic model. These models are Computational Fluid Dynamic (CFD) models that calculate currents and water levels in a specific domain. The domain is discretised into cells and nodes to calculate these hydrodynamic features. This study uses the nodes of this hydrodynamic model as the vertices of the graph. However, for some cases, the hydrodynamic model has too many nodes for the shortest path algorithm. This study presents a method for reducing the number of nodes without reducing the spatial resolution. The nodes are reduced based on a combination of the vorticity and the magnitude of the flow.

This algorithm is implemented in a python software package named Hydrodynamic Algorithm for Logistic enhancement Module (HALEM). HALEM can determine the optimal shipping route for a given hydrodynamic model. Defining different cost functions results in different optimisation purposes. This thesis presents cost functions for the fastest route, shortest route, cheapest route and least polluting route. This software is then implemented in the OpenCLSim software so that this combination of software can optimise routes of entire projects. A case study simulates a beach-nourishment at Schouwen Westkop Noord to demonstrate the practical use of HALEM and OpenCLSim. For this project, 425,500 m3 sand should be dredged offshore and pumped onto the beach. Due to the narrow gullies and tidal changes in hydrodynamic features, the routes were hard to predict. The simulation with HALEM and OpenCLSim shows an increase in the production with 21 % compared to the simulation with just OpenCLSim.

Vector-based graphics are not directly compatible with the raster-based structure of dedicated graphics hardware. They suffer from a locality problem where, in order to check if a specific display pixel lies inside or outside the shape, the entire shape needs to be taken into account. This negates the advantage that the graphics pipeline’s parallel rendering structure offers. Current solutions rasterize vector graphics into a discrete raster-based texture in an extra render pass, every time the required texture resolution changes.
This work presents an approach to store vector-based shapes into discrete raster-based textures, optimized for parallel rendering. This is done by encoding the shape as piece-wise cubic Bézier curves, structured in a quad-tree. Additional data is added to remove aliasing effects in case of minification. These textures are then rendered by a custom fragment shader.
This work discusses the implementation of both the serialization and deserialization steps, as well as rendering regular vector shapes with a comparable quality.