This thesis investigates the use of optimization techniques to determine the value of multimodal (Air+Rail) networks. By using Mixed Integer Linear Programming (MILP), the mathematical model derived especially for this use case determines the optimal way for airlines to route their passengers on ultra-short haul networks. The model does not only consider operational costs, but also addresses the importance of sustainability and the value of time, and includes the value to be found in capturing passengers at cities where there is no airport in the near vicinity. This research demonstrates that by (partially) routing passengers on rail networks, rather than air, flights can be reduced on the short haul network. This results in higher profit, shorter average travel times and reduced average emissions, all whilst capturing a larger market. This research contributes to the knowledge on multimodal air+rail transport by showcasing the potential benefits it can have for airline alliances in a quantitative way, incentivizing airlines to shift towards rail partnerships for their ultra-shorthaul network and adopting more sustainable practices.

Airside cargo transportation is a critical component of the air cargo supply chain, requiring timing and coor dination to minimize operational costs and ensure high service quality for the airline transporting the cargo. The Airside Cargo Transportation Problem (ACTP) addresses the efficient routing and scheduling of cargo vehicles carrying ULDs across the airport’s service roads between the cargo terminal and passenger aircraft stands while adhering to operational and safety requirements. Key aspects of the ACTP include vehicle ca pacity utilization, differentiation between cargo commodities, and the consideration of time-dependent travel times. This paper presents a Mixed-Integer Linear Programming (MILP) formulation of the ACTP as a Set Partitioning Problem. The sets cover all possible consolidations of various cargo transportation requests on paths generated using preprocessing algorithms. The ACTP is computationally feasible by applying a rolling horizon-based heuristic. The model is applied to a case study at Amsterdam Airport Schiphol (AAS) after a potential relocation of KLM Cargo’s terminal. This relocation provides a unique opportunity to evaluate various strategies to optimize airside cargo transportation and study the implications for both KLM Cargo and AAS. Two strategies were evaluated to optimize network capacity: one focusing on spatial distribu tion and the other on time distribution by using traffic predictions. While the spatial distribution strategy degrades the overall performance, the traffic prediction strategy resulted in poor solution quality due to increased model complexity. Accordingly, it could not be proven that its results are comparable to those of other strategies. In addition, two strategies examined the impact of different cargo terminal operational concepts on cargo transportation. The pull strategy, which schedules cargo to arrive at the aircraft stand just in time for loading, improved overall airside cargo transportation performance. In contrast, the push strategy, which transports cargo immediately after processing at the cargo terminal, resulted in a decline in overall airside cargo transportation performance.

With the continuous growth of the aviation sector, concerns regarding the effects of aircraft noise on the health and well-being of communities living in the vicinity of airports have been increasing. Aircraft noise annoyance is inherently subjective and its accurate prediction and quantification represent challenging tasks. There is a lack of consensus in the scientific community regarding which metrics are the best predictors for this type of annoyance. Additionally, many of the metrics employed in the field of (psycho)acoustics are typically computationally expensive. This study aims at developing a methodology which leverages machine learning techniques for instant predictions of various sound metrics and for annoyance rating predictions from input aircraft flyover recordings. The two-step framework involves a Convolutional Neural Network (CNN) for the former, followed by artificial intelligence (AI) models which use the CNN predictions as input for the latter, such as the Support Vector Machine and Random Forest. A listening experiment was conducted in order to gather labeled annoyance data from 60 aircraft flyover recordings of both landings and take-offs and a correlation analysis was subsequently made considering a large pool of sound metrics. The results show that, in general, metrics derived from Psychoacoustic Annoyance models, especially those of Zwicker and Di et al., present better performance as predictors compared to conventional metrics and most Sound Quality Metrics taken individually. Moreover, the AI framework achieves very promising results for both the annoyance ratings and metric predictions (overall, mean absolute errors of the annoyance ratings of approximately 0.4 and below, and R2 values above 0.85), highlighting the potential for bypassing the typically long overhead associated with computing SQMs (which involve expensive and complex algorithms). The drawn conclusions are dependent on the limited amount of data that was available. Hence, the results require further validation using more recordings of various types of aircraft.

The performance of Remaining Useful Life (RUL) prediction models is often limited by data scarcity, especially in safety-critical systems like aircraft engines where failure data is rare. To address this challenge, we propose the Super-SpaceTime GAN, a framework for generating synthetic condition monitoring (CM) data to enhance RUL predictions. The framework incorporates dual-conditioning on operating conditions (OCs) and RUL labels, an autoencoder-based latent space for denoising, and a supervised loss function to align synthetic data with real degradation trajectories. Evaluated on the CMAPSS FD004 dataset, the SuperSpaceTime GAN generates synthetic data that closely mimic real distributions, as verified using JensenShannon Distance, Principal Component Analysis, t-distributed Stochastic Neighbor Embedding, and a novel autoencoder-based health monitoring metric. The framework demonstrates improvement in prognostic performance in limited training data scenarios, with gains persisting even in data-rich settings. These findings highlight the potential of the Super-SpaceTime GAN to improve RUL predictions by addressing data scarcity, making it a valuable tool for Prognostics and Health Management (PHM).

Research aimed at improving engine maintenance practices is vital to ensuring the aviation industry's high safety standards. Both preventive and corrective maintenance approaches result in either higher costs or increased failures. Predictive maintenance aims to achieve an optimal balance by using time series and historical failure data to anticipate future problems and plan engine maintenance activities. While extensive research has been done on the application of AI models for predictive maintenance using simulated datasets, limited research has been done using real world engine data. Simulated datasets lack the complexity of real world systems which undermines the applicability of their findings to real world situations. This paper addresses the gap between theory and industry by utilizing eight years of operating and maintenance data from the GEnx-1B, provided by KLM Engine Services. We expand classic residual based methods with a novel approach, sister engine analysis, to improve the performance of an LSTM-based fault detection model. The data is labeled based on maintenance reports, which provide the most accurate link to the underlying degradation state of an engine. These reports reveal a significant bias toward high pressure turbine degradation in the GEnx-1B. Furthermore, our findings indicate that incorporating sister engine analysis enhances the average F1 score of an AI-based fault detection model by 7.7% and that AI methods outperform the current industry standard when provided with the same input data. The use of physically relevant model features and LIME analysis ensures the model's behavior is interpretable. These results are significant as they help build trust in AI-based solutions for predictive maintenance, which is crucial for their broader adoption.

Haptic shared controllers (HSCs) are a promising solution to prevent human over-reliance on automation during tasks such as car driving. However, research has shown that if the HSC is tuned incorrectly, then there is a risk of haptic conflicts between the human and HSC. To address this challenge, this paper presents the design and implementation of a novel adaptive HSC that continuously adjusts its look-ahead time. By estimating the time shift between the reference state of the HSC and the actual state, the HSC adapts to the look-ahead time of the human it is interacting with. Results from a human-in- the-loop experiment show that the novel HSC achieves similar subjective ratings as a fixed preview HSC, as well as a significant improvement over a fixed pursuit HSC. Going from pursuit to preview, objective experiment data shows that as the adaptive HSC adjusts its look-ahead time, haptic conflicts are reduced and tracking performance is increased. The presented findings are a step forward in designing haptic support systems with high chances of user acceptance. The proposed adaptive look-ahead algorithm provides a new method for online estimation of human look-ahead time, with or without a HSC in-the-loop.