Search in the TU Delft Repository Collections
Recently Added Records
As rail networks face increasing pressure to improve punctuality and capacity under mixed-traffic conditions, digital support tools such as Connected Driver Advisory Systems (C-DAS) are increasingly deployed. Yet, empirical evidence of their operational effects in daily practice remains limited. This paper presents a data-driven two-level approach based on macroscopic and microscopic event graph representations, constructed from the same underlying corridor and observed train movements, to evaluate the performance of C-DAS. It fuses timetable event records, track section occupation and release data, and C-DAS usage logs to enable comparison between trains operated with and without C-DAS at timetable point and block section levels. Derived from the event graphs, differences in timetable deviation, running time deviation, and the occurrence of unplanned yellow signal aspects are assessed using non-parametric statistical tests. The approach is demonstrated through its application to RouteLint, a C-DAS deployed for freight operations in the Netherlands, using one year of data from a mixed-traffic mainline corridor comprising 23,573 freight train runs. The results show modest but consistent differences in operational performance between trains operated with and without RouteLint. Statistically significant differences in timetable deviation are identified at multiple key locations, with RouteLint-equipped trains showing approximately 30–70 s lower deviations at major stations and junctions. At the microscopic level, running time deviations exhibit substantial local variability, while lower deviations are observed on selected block sections for RouteLint-equipped trains. In addition, RouteLint-equipped trains show approximately 6% lower overall odds that an observed yellow signal aspect is unplanned, while the association varies across operating contexts.
...
As rail networks face increasing pressure to improve punctuality and capacity under mixed-traffic conditions, digital support tools such as Connected Driver Advisory Systems (C-DAS) are increasingly deployed. Yet, empirical evidence of their operational effects in daily practice remains limited. This paper presents a data-driven two-level approach based on macroscopic and microscopic event graph representations, constructed from the same underlying corridor and observed train movements, to evaluate the performance of C-DAS. It fuses timetable event records, track section occupation and release data, and C-DAS usage logs to enable comparison between trains operated with and without C-DAS at timetable point and block section levels. Derived from the event graphs, differences in timetable deviation, running time deviation, and the occurrence of unplanned yellow signal aspects are assessed using non-parametric statistical tests. The approach is demonstrated through its application to RouteLint, a C-DAS deployed for freight operations in the Netherlands, using one year of data from a mixed-traffic mainline corridor comprising 23,573 freight train runs. The results show modest but consistent differences in operational performance between trains operated with and without RouteLint. Statistically significant differences in timetable deviation are identified at multiple key locations, with RouteLint-equipped trains showing approximately 30–70 s lower deviations at major stations and junctions. At the microscopic level, running time deviations exhibit substantial local variability, while lower deviations are observed on selected block sections for RouteLint-equipped trains. In addition, RouteLint-equipped trains show approximately 6% lower overall odds that an observed yellow signal aspect is unplanned, while the association varies across operating contexts.
Bachelor thesis
(2026)
-
Joseph Abraham Thomas, H.S.J. van der Zant, R. Conte, A.A.F.M. Artaud, H.X. Lin, Alexis Derumigny
Single-molecule switches are a promising technology which can result in advances in nanoscale memristors and neuromorphic computing. While resistance-switching memristor-like phenomena induced by voltage bias have been documented in molecular junctions, the fundamental mechanisms governing these processes are not yet fully understood. IV-traces measured of 5 different rigid conjugated molecules (1-SAc, 2-SMe, 3-meta, 4-Hmeta, 5-Hpara) in a mechanically controlled break junction consist of complex features such as small conductance peaks which cannot be explained using the single-level model. Resistance-switching phenomena is studied to disentangle switches caused by mechanical origin related to the positioning of the molecule within the break junction, and those caused by a hypothetical proton-hopping mechanism which is expected to be a cause for switches in 4-Hmeta and 5-Hpara molecules. In this study, the molecule is modelled as a dipole consisting of 2 coupled sites with capacitive coupling, resulting in more accurate fits of the IV-traces compared to when using the single-level model. These small conductance peaks are found to be related to a very strong coupling Γ to the electrodes or a large coupling τ between the 2-sites within the dipole model. The dipole model is found to be insufficient to explain the entire shape of these IV traces and are found to be only partially able to fit the entire trace. Model inadequacies of the proposed dipole model are discussed extensively. Corrections to the dipole model are suggested such as in the form of a parallel resistance and inclusion of non-rectangular tunnelling barriers. The results of the curve fits using the dipole model point towards a step towards uncovering the origin of the memristive behaviour found in these IV-traces.
...
Single-molecule switches are a promising technology which can result in advances in nanoscale memristors and neuromorphic computing. While resistance-switching memristor-like phenomena induced by voltage bias have been documented in molecular junctions, the fundamental mechanisms governing these processes are not yet fully understood. IV-traces measured of 5 different rigid conjugated molecules (1-SAc, 2-SMe, 3-meta, 4-Hmeta, 5-Hpara) in a mechanically controlled break junction consist of complex features such as small conductance peaks which cannot be explained using the single-level model. Resistance-switching phenomena is studied to disentangle switches caused by mechanical origin related to the positioning of the molecule within the break junction, and those caused by a hypothetical proton-hopping mechanism which is expected to be a cause for switches in 4-Hmeta and 5-Hpara molecules. In this study, the molecule is modelled as a dipole consisting of 2 coupled sites with capacitive coupling, resulting in more accurate fits of the IV-traces compared to when using the single-level model. These small conductance peaks are found to be related to a very strong coupling Γ to the electrodes or a large coupling τ between the 2-sites within the dipole model. The dipole model is found to be insufficient to explain the entire shape of these IV traces and are found to be only partially able to fit the entire trace. Model inadequacies of the proposed dipole model are discussed extensively. Corrections to the dipole model are suggested such as in the form of a parallel resistance and inclusion of non-rectangular tunnelling barriers. The results of the curve fits using the dipole model point towards a step towards uncovering the origin of the memristive behaviour found in these IV-traces.
Green hydrogen produced by alkaline water electrolysis may contain residual water vapour and other impurities that must be removed before use in fuel-cell applications. This study investigated adsorption-based hydrogen drying using Pressure Swing Adsorption (PSA) and Temperature Swing Adsorption (TSA), with particular attention to process modelling in Aspen Adsorption.
Zeolite 4A was selected because of its strong affinity for water at low partial pressures. Water adsorption equilibrium was represented using a Dual-Site Langmuir 2 model, while mass transfer was described using the Linear Driving Force model. A non-isothermal packed-bed model was developed for a hydrogen feed containing 872 ppmv water at 25 bar, 298.15 K, and a flow rate of 0.0223 kmol s−1. Several PSA configurations were tested, including conventional pressure regeneration, vacuum regeneration, and temperature-assisted vacuum regeneration. Conventional regeneration at 1 bar produced only limited reduction in solid water loading, while deep-vacuum simulations showed numerical convergence problems during multi-cycle operation. Therefore, the detailed cyclic and economic assessment was continued using TSA.
For the TSA process, the simulated breakthrough time was approximately 19 250 s, and an adsorption time of 14 000 s was selected to provide a safety margin before water breakthrough. Regeneration w
...
Zeolite 4A was selected because of its strong affinity for water at low partial pressures. Water adsorption equilibrium was represented using a Dual-Site Langmuir 2 model, while mass transfer was described using the Linear Driving Force model. A non-isothermal packed-bed model was developed for a hydrogen feed containing 872 ppmv water at 25 bar, 298.15 K, and a flow rate of 0.0223 kmol s−1. Several PSA configurations were tested, including conventional pressure regeneration, vacuum regeneration, and temperature-assisted vacuum regeneration. Conventional regeneration at 1 bar produced only limited reduction in solid water loading, while deep-vacuum simulations showed numerical convergence problems during multi-cycle operation. Therefore, the detailed cyclic and economic assessment was continued using TSA.
For the TSA process, the simulated breakthrough time was approximately 19 250 s, and an adsorption time of 14 000 s was selected to provide a safety margin before water breakthrough. Regeneration w
...
Green hydrogen produced by alkaline water electrolysis may contain residual water vapour and other impurities that must be removed before use in fuel-cell applications. This study investigated adsorption-based hydrogen drying using Pressure Swing Adsorption (PSA) and Temperature Swing Adsorption (TSA), with particular attention to process modelling in Aspen Adsorption.
Zeolite 4A was selected because of its strong affinity for water at low partial pressures. Water adsorption equilibrium was represented using a Dual-Site Langmuir 2 model, while mass transfer was described using the Linear Driving Force model. A non-isothermal packed-bed model was developed for a hydrogen feed containing 872 ppmv water at 25 bar, 298.15 K, and a flow rate of 0.0223 kmol s−1. Several PSA configurations were tested, including conventional pressure regeneration, vacuum regeneration, and temperature-assisted vacuum regeneration. Conventional regeneration at 1 bar produced only limited reduction in solid water loading, while deep-vacuum simulations showed numerical convergence problems during multi-cycle operation. Therefore, the detailed cyclic and economic assessment was continued using TSA.
For the TSA process, the simulated breakthrough time was approximately 19 250 s, and an adsorption time of 14 000 s was selected to provide a safety margin before water breakthrough. Regeneration w
Zeolite 4A was selected because of its strong affinity for water at low partial pressures. Water adsorption equilibrium was represented using a Dual-Site Langmuir 2 model, while mass transfer was described using the Linear Driving Force model. A non-isothermal packed-bed model was developed for a hydrogen feed containing 872 ppmv water at 25 bar, 298.15 K, and a flow rate of 0.0223 kmol s−1. Several PSA configurations were tested, including conventional pressure regeneration, vacuum regeneration, and temperature-assisted vacuum regeneration. Conventional regeneration at 1 bar produced only limited reduction in solid water loading, while deep-vacuum simulations showed numerical convergence problems during multi-cycle operation. Therefore, the detailed cyclic and economic assessment was continued using TSA.
For the TSA process, the simulated breakthrough time was approximately 19 250 s, and an adsorption time of 14 000 s was selected to provide a safety margin before water breakthrough. Regeneration w
The Spherical Approximation in MOND
Quantifying Geometric Biases in the Radial Acceleration Relation
The Radial Acceleration Relation (RAR) is one of the most striking empirical regularities in galactic dynamics: the observed centripetal acceleration in galaxies correlates tightly with the acceleration predicted from the baryonic mass distribution alone. Modified Newtonian Dynamics (MOND) naturally explains this relation, while the standard dark matter paradigm requires fine-tuning. However, nearly all observational tests of the RAR adopt a simplifying approximation: the baryonic acceleration at radius r is computed as g_bar = GM(<r)/r², as if the galaxy’s mass distribution were spherically symmetric. Real disk galaxies are highly flattened, and the validity of this approximation has never been directly quantified for the Radial Acceleration Relation.This thesis addresses two questions: (1) How accurate is the spherical approximation for computing baryonic accelerations in disk galaxies? (2) Does its use bias previous tests of the MOND RAR? Using a numerical AQUAL solver applied to realistic three-dimensional stellar density reconstructions from Spitzer 3.6 μm imaging, we compare the spherical approximation to the true disk gravity for three galaxies spanning a range of morphologies: NGC 2841 (bulge-dominated), NGC 3198 (moderate bulge), and NGC 2976 (bulgeless).We find that the accuracy of the spherical approximation for computing baryonic acceleration
...
The Radial Acceleration Relation (RAR) is one of the most striking empirical regularities in galactic dynamics: the observed centripetal acceleration in galaxies correlates tightly with the acceleration predicted from the baryonic mass distribution alone. Modified Newtonian Dynamics (MOND) naturally explains this relation, while the standard dark matter paradigm requires fine-tuning. However, nearly all observational tests of the RAR adopt a simplifying approximation: the baryonic acceleration at radius r is computed as g_bar = GM(<r)/r², as if the galaxy’s mass distribution were spherically symmetric. Real disk galaxies are highly flattened, and the validity of this approximation has never been directly quantified for the Radial Acceleration Relation.
This thesis addresses two questions: (1) How accurate is the spherical approximation for computing baryonic accelerations in disk galaxies? (2) Does its use bias previous tests of the MOND RAR? Using a numerical AQUAL solver applied to realistic three-dimensional stellar density reconstructions from Spitzer 3.6 μm imaging, we compare the spherical approximation to the true disk gravity for three galaxies spanning a range of morphologies: NGC 2841 (bulge-dominated), NGC 3198 (moderate bulge), and NGC 2976 (bulgeless).
We find that the accuracy of the spherical approximation for computing baryonic acceleration
Modelling Ground Vibrations Induced by Pile Removal
From Free-Field Propagation to Simplified Soil-Structure Interaction
Ground-borne vibrations from construction activities can propagate through layered soil and affect nearby vibration-sensitive buildings. This thesis develops and evaluates a numerical source-soil-receiver modelling framework for pile-removal-induced vibrations at TU Delft Campus Zuid, combining field measurements, CPT-based soil characterisation, frequency-domain finite element modelling and simplified soil-structure interaction (SSI) formulations.
Measured vibration data showed a non-stationary response, with the dominant frequency during pile removal developing from approximately 6-7~Hz towards a sustained component around 14~Hz. Because the mechanical source force was unknown, model-dependent equivalent vertical and radial forces were identified from near-source measurements and used as input to a two-dimensional axisymmetric free-field model. The soil profile was derived from 24 CPT soundings, with dynamic properties estimated using empirical correlations. Comparison with far-field measurements showed that the calibrated model reproduced the dominant frequencies more consistently than the absolute vibration amplitudes.
The calibrated free-field model was subsequently extended towards SSI using two fictive embedded receivers with different embedment depths. Fully coupled FE models were used as numerical references and compared with engineering spring-dashpot, uncoupled numerical impedance and coupled numerical impedance formulations. The main differences between the SSI approaches occurred in response amplitude rather than dominant frequency. Within the investigated interface discretisations, the engineering and uncoupled formulations showed the most consistent agreement with the fully coupled references. The uncoupled formulation showed little sensitivity to interface refinement, whereas the coupled formulation was more sensitive to segment size. Refinement improved several coupled response components, although formal convergence was not established.
The results demonstrate a structured approach for predicting the transmission of construction-induced vibrations through the source-soil-receiver chain when source and dynamic soil information are limited. The SSI comparison is methodological rather than building-specific, as the fictive axisymmetric receivers do not represent the actual TNW structure. Further measurements, improved dynamic soil characterisation and building-specific three-dimensional modelling are required for quantitative predictions of vibration levels inside TNW.
...
Measured vibration data showed a non-stationary response, with the dominant frequency during pile removal developing from approximately 6-7~Hz towards a sustained component around 14~Hz. Because the mechanical source force was unknown, model-dependent equivalent vertical and radial forces were identified from near-source measurements and used as input to a two-dimensional axisymmetric free-field model. The soil profile was derived from 24 CPT soundings, with dynamic properties estimated using empirical correlations. Comparison with far-field measurements showed that the calibrated model reproduced the dominant frequencies more consistently than the absolute vibration amplitudes.
The calibrated free-field model was subsequently extended towards SSI using two fictive embedded receivers with different embedment depths. Fully coupled FE models were used as numerical references and compared with engineering spring-dashpot, uncoupled numerical impedance and coupled numerical impedance formulations. The main differences between the SSI approaches occurred in response amplitude rather than dominant frequency. Within the investigated interface discretisations, the engineering and uncoupled formulations showed the most consistent agreement with the fully coupled references. The uncoupled formulation showed little sensitivity to interface refinement, whereas the coupled formulation was more sensitive to segment size. Refinement improved several coupled response components, although formal convergence was not established.
The results demonstrate a structured approach for predicting the transmission of construction-induced vibrations through the source-soil-receiver chain when source and dynamic soil information are limited. The SSI comparison is methodological rather than building-specific, as the fictive axisymmetric receivers do not represent the actual TNW structure. Further measurements, improved dynamic soil characterisation and building-specific three-dimensional modelling are required for quantitative predictions of vibration levels inside TNW.
...
Ground-borne vibrations from construction activities can propagate through layered soil and affect nearby vibration-sensitive buildings. This thesis develops and evaluates a numerical source-soil-receiver modelling framework for pile-removal-induced vibrations at TU Delft Campus Zuid, combining field measurements, CPT-based soil characterisation, frequency-domain finite element modelling and simplified soil-structure interaction (SSI) formulations.
Measured vibration data showed a non-stationary response, with the dominant frequency during pile removal developing from approximately 6-7~Hz towards a sustained component around 14~Hz. Because the mechanical source force was unknown, model-dependent equivalent vertical and radial forces were identified from near-source measurements and used as input to a two-dimensional axisymmetric free-field model. The soil profile was derived from 24 CPT soundings, with dynamic properties estimated using empirical correlations. Comparison with far-field measurements showed that the calibrated model reproduced the dominant frequencies more consistently than the absolute vibration amplitudes.
The calibrated free-field model was subsequently extended towards SSI using two fictive embedded receivers with different embedment depths. Fully coupled FE models were used as numerical references and compared with engineering spring-dashpot, uncoupled numerical impedance and coupled numerical impedance formulations. The main differences between the SSI approaches occurred in response amplitude rather than dominant frequency. Within the investigated interface discretisations, the engineering and uncoupled formulations showed the most consistent agreement with the fully coupled references. The uncoupled formulation showed little sensitivity to interface refinement, whereas the coupled formulation was more sensitive to segment size. Refinement improved several coupled response components, although formal convergence was not established.
The results demonstrate a structured approach for predicting the transmission of construction-induced vibrations through the source-soil-receiver chain when source and dynamic soil information are limited. The SSI comparison is methodological rather than building-specific, as the fictive axisymmetric receivers do not represent the actual TNW structure. Further measurements, improved dynamic soil characterisation and building-specific three-dimensional modelling are required for quantitative predictions of vibration levels inside TNW.
Measured vibration data showed a non-stationary response, with the dominant frequency during pile removal developing from approximately 6-7~Hz towards a sustained component around 14~Hz. Because the mechanical source force was unknown, model-dependent equivalent vertical and radial forces were identified from near-source measurements and used as input to a two-dimensional axisymmetric free-field model. The soil profile was derived from 24 CPT soundings, with dynamic properties estimated using empirical correlations. Comparison with far-field measurements showed that the calibrated model reproduced the dominant frequencies more consistently than the absolute vibration amplitudes.
The calibrated free-field model was subsequently extended towards SSI using two fictive embedded receivers with different embedment depths. Fully coupled FE models were used as numerical references and compared with engineering spring-dashpot, uncoupled numerical impedance and coupled numerical impedance formulations. The main differences between the SSI approaches occurred in response amplitude rather than dominant frequency. Within the investigated interface discretisations, the engineering and uncoupled formulations showed the most consistent agreement with the fully coupled references. The uncoupled formulation showed little sensitivity to interface refinement, whereas the coupled formulation was more sensitive to segment size. Refinement improved several coupled response components, although formal convergence was not established.
The results demonstrate a structured approach for predicting the transmission of construction-induced vibrations through the source-soil-receiver chain when source and dynamic soil information are limited. The SSI comparison is methodological rather than building-specific, as the fictive axisymmetric receivers do not represent the actual TNW structure. Further measurements, improved dynamic soil characterisation and building-specific three-dimensional modelling are required for quantitative predictions of vibration levels inside TNW.