Aircraft maintenance scheduling is a focus point for airlines. Maintenance is essential to ensure the airworthiness of aircraft, but it comes at the cost of rendering them unavailable for operations. In current operations, aircraft maintenance scheduling must often be updated to include time for non-routine and non-schedule tasks. These non-routine tasks can increase costs, maintenance workload, and uncertainty of the airlines’ operations. This research introduces a supervised learning framework designed to forecast future non-routine task workloads accurately, improving the accuracy of the planned maintenance schedule. This framework consists of two random forest predictors which estimate the amount of non-routine tasks and the number of future work hours that should be allocated in advance for potential non-routine tasks. Our approach produces highly reliable predictions by leveraging a robust dataset obtained from an international airline. The results show an average of 20% improvement versus an existing on-site sampling method. Furthermore, our in-depth analysis of prediction distributions enables the identification of the underlying causes of significant prediction errors, shedding light on the unpredictabilities inherent to non-routine tasks.@en

As the volume of telemetry data generated by satellites and other complex systems continues to grow, there is a pressing need for more efficient and accurate anomaly detection methods. Current techniques often rely on human analysis and preset criteria, presenting several challenges including the necessity for expert interpretation and continual updates to match the dynamic mission environment. This paper critically examines the use of deep anomaly detection (DAD) methods in addressing these challenges, evaluating their efficacy on real-world spacecraft telemetry data. It exposes limitations in current DAD research, highlighting the tendency for performance results to be overestimated and suggesting that simpler methods can sometimes outperform more complex DAD algorithms. By comparing established metrics for anomaly detection with newly proposed ones, this paper aims to improve the evaluation of DAD algorithms. It underscores the importance of using less accuracy-inflating metrics and offers a comprehensive comparison of DAD methods on popular benchmark datasets and real-life satellite telemetry data. Among the DAD methods examined, the LSTM algorithm demonstrates considerable promise. However, the paper also reveals the potential limitations of this approach, particularly in complex systems that lack a single, clear predictive failure channel. The paper concludes with a series of recommendations for future research, including the adoption of best practices, the need for high-quality, pre-split datasets, and the investigation of other prediction error methods. Through these insights, this paper contributes to the improved understanding and application of DAD methods, ultimately enhancing the reliability and effectiveness of anomaly detection in real-world scenarios.

@en

Artificial Intelligence (AI) is transforming the future of industries by introducing new paradigms. To address data privacy and other challenges of decentralization, research has focused on Federated Learning (FL), which combines distributed Machine Learning (ML) models from multiple parties without exchanging confidential information. However, conventional FL methods struggle to handle situations where data samples have diverse features and sizes. We propose a Hybrid Federated Learning solution with label synchronization to overcome this challenge. Our FedLabSync algorithm trains a feed-forward Artificial Neural Network while alerts that it can aggregate knowledge of other ML architectures compatible with the Stochastic Gradient Descent algorithm by conducting a penalized collaborative optimization. We conducted two industrial case studies: product inspection in Bosch factories and aircraft component Remaining Useful Life predictions. Our experiments on decentralized data scenarios demonstrate that FedLabSync can produce a global AI model that achieves results on par with those of centralized learning methods.

@en

Despite its success in various research domains, Reinforcement Learning (RL) faces challenges in its application to air transport operations due to the rigorous certification standards of the aviation industry. The existing regulatory framework fails to provide adequate, acceptable means of compliance for RL applications, and thus, there is no legal framework for their safe deployment yet. Guidelines must be formulated to certify RL models aimed at air transport operations to enable real-world utilisation of these promising methods. These guidelines must consider the unique characteristics of these models, deviating from the methodology of current guidelines crafted before the emergence of ML applications. The paper proposes novel certification requirements for RL models based on their technical characteristics, safety-criticality, and autonomy. This framework covers the choice of the RL algorithm and analyses the actions, agents, environment, and potential hazards and risks of the RL application. Additionally, this work outlines the evidence the certification applicant must present to demonstrate compliance with these requirements. While this framework is not a complete solution for the complex problem of certifying RL, it is intended to serve as an initial framework which can be extended in cooperation with regulatory entities.@en

In the Condition-Based Maintenance (CBM) context, the definition of optimal maintenance plans for an aircraft fleet depends on an efficient integration of : (i) the probabilistic predictions of the health condition of the components and (ii) the stochastic arrival of the corrective maintenance tasks, together with consideration of the preventive maintenance tasks as defined in the Maintenance Planning Document (MPD). To this end, in this paper, we present a two-stage dynamic scheduling framework to solve the aircraft fleet maintenance scheduling problem under a CBM strategy in a disruptive environment. In the first stage of the framework, we address the uncertainty in the predicted health state of the monitored components by planning the optimal maintenance policy based upon the belief state-space of the health of the components. The decision-making process is formulated as a Partially Observable Markov Decision Process (POMDP) and is solved using the Partially Observable Monte Carlo Planning (POMCP) algorithm, considering the aircraft maintenance scheduling problem requirements. In the second stage, a Deep Q-Network (DQN) is developed, that integrates the defined maintenance policy of the monitored components within the scheduling of the aircraft fleet's preventive and corrective maintenance tasks. Our model, through a rolling horizon approach, continuously creates and adjusts the maintenance schedule, reacting to new updated task information, where the availability of maintenance resources constraints the execution of each task. The proposed framework was tested on a case study from a large airline and the performance was evaluated against the current state practice of the airline. The results show that our model can schedule 96.4% of monitored components on-time. As a consequence of this, a 46.2% maintenance cost reduction is achieved for the considered monitored components relative to a corrective maintenance approach.

@en

Over the past years, advanced prognostic models and approaches have been developed. Most existing approaches are tailored to one specific system and cannot adaptively be used on different systems. This can lead to years of research and expertise being put into implementing prognostic models without the capacity to predict system failures, either because of a lack of data or data quality or because failure behavior cannot be captured by data-driven models. In addition, prognostic models are often evaluated using metrics only related to the correctness of predictions, preventing meaningful evaluation of operational performance. This paper makes use of a framework that can automatically choose prognostic settings based on specific system data. It simultaneously optimizes the choice of methodologies using metrics that capture multiple aspects of prediction quality. We apply this framework to both a simulated data set and a real aircraft data set to characterize the impact of metrics on the choice of prognostic methodologies. The results show that the choice of optimization metric greatly impacts the output of the generic prognostic framework and the overall performance. In addition, a definition for data suitability is provided and assessed on the aircraft system data sets.

@en

Prognostics is used in predictive maintenance to estimate the remaining time to the end of the life of a system or component. Among the many challenges of prognostics is the need for model verification and validation. Over the years, several objective metrics have been utilized by the community. Some of these metrics came from statistics, others from forecasting, and others have been proposed specifically for prognostics. A single “perfect” metric has not yet been put forward. Finding one metric that can excel in all evaluation dimensions and case studies is an open question. In this review, we analyze the most important metrics of prognostics. A set of 19 metrics is subject to analysis and implementation. The metrics are implemented on a public GitHub project. Our analysis focuses only on metrics for deterministic predictions. Stochastic predictions are out of the scope. The paper describes properties, advantages, disadvantages, and industrial applicability of each metric. We also discuss potential modifications to the existing metrics and the development of new metrics. A final table summarizes the main properties of the metrics. Our goal is to raise awareness about prognostics metrics and help establish a common evaluation procedure. Code available at: MetricsForPrognostics.

@en

The objective of the EU-funded research project CHYLA (Credible HYbrid eLectric Aircraft) was to identify opportunities or limitations/challenges for the applications of key radical hybrid-electric technologies and areas suitable for scaling them over different aircraft classes. This was done using a ombination of conceptual aircraft design supported by sensitivity studies, credibility-based MDO and assessment of a regional operative scenario. This article summarizes the key findings from the project and presents the landscape of technology application areas. Notably, the regional and commuter classes present the largest design space with significant fuel-saving potential depending on the mission.@en

Condition-based maintenance (CBM) scheduling of an aircraft fleet in a disruptive environment while considering health prognostics for a set of systems is a very complex combinatorial problem, which is becoming more challenging in light of the uncertainty included in health prognostics. This type of problem falls under the broad category of resource-constrained scheduling problems under uncertainty and is often solved using a mixed integer linear programming (MILP) formulation. While a MILP framework is very promising, the problem size can scale exponentially with the number of considered aircraft and considered tasks, leading to significantly high computational costs. The most recent advances in artificial intelligence have demonstrated the capability of deep reinforcement learning (DRL) algorithms to alleviate this curse of dimensionality, as once the DRL agent is trained, it can achieve real-time optimization of the maintenance schedule. However, there is no guarantee of optimality. These comparative merits of a MILP and a DRL formulation for the aircraft fleet maintenance scheduling problem have not been discussed in the literature. This study is a response to this research gap. We conduct a comparison of a MILP and a DRL scheduling model, which are used to derive the optimal maintenance schedule for various maintenance scenarios for aircraft fleets of different sizes in a disruptive environment, while considering health prognostics and the available resources for the execution of each task. The quality of solutions is evaluated on the basis of four planning objectives, defined according to real airline practice. The results show that the DRL approach achieves better results with respect to scheduling of prognostics-driven tasks and requires less computational time, whereas the MILP model produces more stable maintenance schedules and induces less maintenance ground time. Overall, the comparison provides valuable insights for the integration of health prognostics in airline maintenance practice. @en

This paper proposes using reinforcement learning (RL) to schedule maintenance tasks, which can significantly reduce direct operating costs for airlines. The approach consists of a static algorithm for long-term scheduling and an adaptive algorithm for rescheduling based on new maintenance information. To assess the performance of both approaches, three key performance indicators (KPIs) are defined: Ground Time, representing the hours an aircraft spends on the ground; Time Slack, measuring the proximity of tasks to their due dates; and Change Score, quantifying the similarity level between initial and adapted maintenance plans when new information surfaces. The results demonstrate the efficacy of RL in producing efficient maintenance plans, with the algorithms complementing each other to form a solid foundation for routine tasks and real-time responsiveness to new information. While the static algorithm performs slightly better in terms of Ground Time and Time Slack, the adaptive algorithm excels overwhelmingly in terms of Change Score, offering greater flexibility in handling new maintenance information. The proposed RL-based approach can improve the efficiency of aircraft maintenance and has the potential for further research in this area.

@en

The aviation industry has set an ambitious goal of reducing its climate impacts. Accordingly, airlines must balance their plans according to this goal with financial considerations. We developed a multi-objective framework to facilitate climate-aware network design by incorporating the objective to minimise the flight average temperature response (ATR) when optimising the airline network. We also assessed the operational improvements (OIs) which are introduced to improve sustainability in airline operations. In particular, we considered intermediate stop-overs (ISOs) and lower flight altitudes as OIs in our case studies. We analysed the impact of considering the climate impact in the planning of operations of three different airline types: one main-hub-and-spoke (KLM), one smaller multi-hub airline (TAP), and one low-cost carrier (EasyJet). The results show that airlines could also lower their environmental impact by 10–36% when considering the ATR as an objective. However, this would require an 8–20% reduction in profits. Adopting lower-altitude flying with ISO could mitigate their climate impact by 27–49% while reducing profits by approximately 6%. Our study highlights the importance of considering the airline network as a whole and demonstrates the potential benefits of operational improvements from a network perspective.@en

Electrification of aviation is regarded as one of the means to make aircraft operations less polluting and to have lower climate impact. Yet, air transportation's environmental impact depends on power train technologies and novel designs and aircraft operations within airline networks. Fully- or hybrid-electric aircraft may enter existing air transport networks through fleet replacement yet require airlines to adapt in order to operate electrified aircraft strategically. This research studies how airlines can strategically adjust their network and fleet composition when considering electrified aircraft. The novelty of this approach is to provide a direct feedback coupling between fleet planning, conceptual hybrid-electric aircraft design and climate impact minimization. Therefore, a strategic airline planning model, consisting of fleet and network analysis, is coupled to a hybrid-electric aircraft design model. A case study on the sensitivity of a regional airline network is presented to demonstrate the framework and assess the impact of trying to design aircraft and fleets with minimal climate footprint. A decrease in emissions with respect to a kerosene fleet of 11% can be achieved when a hybrid-electric fleet is designed particularly for the specified network, at the penalty of a profit decrease of 13%. Limiting fleet diversity to three types results in only 7% emissions decrease. Increasing the battery-specific energy shows an expected beneficial effect on emissions.@en

Condition-Based Maintenance (CBM) is a policy that uses information about the health condition of systems and structures to identify optimal maintenance interventions over time, increasing the efficiency of maintenance operations. Despite CBM being a well-established concept in academic research, the practical uptake in aviation needs to catch up to expectations. This research aims to identify challenges, limitations, solution directions, and policy implications related to adopting CBM in aviation. We use a generalizable and holistic assessment framework to achieve this aim, following a process-oriented view of CBM development as an aircraft lifecycle management policy. Based on various inputs from industry and academia, we identified several major sets of challenges and suggested three primary solution categories. These address data quantity and quality, CBM implementation, and the integration of CBM with future technologies, highlighting future research and practice directions. @en

To reduce the climate impact of aviation, researchers are studying the replacement of fossil kerosene with liquid hydrogen and/or drop-in sustainable aviation fuel (SAF). These fuels can bring significant reductions in CO2 emissions and can offer savings in terms of non-CO2 climate effects. In addition, tube-and-wing aircraft can be optimized to decrease the global-warming impact by using a climate metric as a design objective rather than the operating costs. Previous research has shown that airplanes designed for minimal climate impact have a reduced cruise speed and fly at a lower altitude. This paper suggests a multidisciplinary, multi-level approach the evaluate the consequences of such design and fuels choices at the network level. Following the aircraft design step, a dynamic programming routine allocates the fleet and schedules the flights to maximize the network profit. We consider a hub-and-spoke network operating from Atlanta, with demand for domestic and international destinations. Compared to the reference cost-optimal kerosene fleet, a fleet consisting of climate-optimized kerosene aircraft can reduce the climate impact by 61% at a loss in network profit of approximately 21%. This design choice requires allocating an additional five aircraft. A fleet operating climate-optimal, hydrogen aircraft minimizes the climate impact. However, the high operating cost of long-range, hydrogen aircraft lowers the achievable profit. Aircraft powered by drop-in SAF provides Pareto-optimal solutions. These insights can be used to make decisions about the allocation of future aviation fuels in a network and the payload-range requirements of future aircraft.@en

Spacecraft systems collect health-related data continuously, which can give an indication of the systems’ health status. While they rarely occur, the repercussions of such system anomalies, faults, or failures can be severe, safety-critical and costly. Therefore, the data are used to anticipate any kind of anomalous behaviour. Typically this is performed by the use of simple thresholds or statistical techniques. Over the past few years, however, data-driven anomaly detection methods have been further developed and improved. They can help to automate the process of anomaly detection. However, it usually is time intensive and requires expertise to identify and implement suitable anomaly detection methods for specific systems, which is often not feasible for application at scale, for instance, when considering a satellite consisting of numerous systems and many more subsystems. To address this limitation, a generic diagnostic framework is proposed that identifies optimal anomaly detection techniques and data pre-processing and thresholding methods. The framework is applied to two publicly available spacecraft datasets and a real-life satellite dataset provided by the European Space Agency. The results show that the framework is robust and adaptive to different system data, providing a quick way to assess anomaly detection for the underlying system. It was found that including thresholding techniques significantly influences the quality of resulting anomaly detection models. With this, the framework can provide both a way forward in developing data-driven anomaly detection methods for spacecraft systems and guidance relative to the direction of anomaly detection method selection and implementation for specific use cases.@en

Current state-of-the-art airline planning models face computational limitations, restricting the operational applicability to problems of representative sizes. This is particularly the case when considering the uncertainty necessarily associated with the long-term plan of an aircraft fleet. Considering the growing interest in the application of machine learning techniques to operations research problems, this article investigates the applicability of these techniques for airline planning. Specifically, an Advantage Actor–Critic (A2C) reinforcement learning algorithm is developed for the airline fleet planning problem. The increased computational efficiency of using an A2C agent allows us to consider real-world-sized problems and account for highly-volatile uncertainty in demand and fuel price. The result is a multi-stage probabilistic fleet plan describing the evolution of the fleet according to a large set of future scenarios. The A2C algorithm is found to outperform a deterministic model and a deep Q-network algorithm. The relative performance of the A2C increases as more complexity is added to the problem. Further, the A2C algorithm can compute a multi-stage fleet planning solution within a few seconds@en

In Revenue Management (RM) systems, information censoring and the interaction between the forecasting and optimization stages, increases the costs and complexity of performance analysis using historical data. An affordable and suitable alternative is using simulations, but appropriate behavioral models must be considered. In the following document, we discuss and test the implementation of a dynamic air transport market simulator, designed to analyze RM systems. The simulator replicates the behavior of passengers that book seats offered in multiple flights by different airlines. We use discrete choice models to replicate the demand behavior, accounting for preferences and decision rule heterogeneity, and including a temporal evolution of the preference throughout the selling horizon. To replicate the supply behavior, a number of airlines modify the price and quantity of different fare classes offered in each flight, using a variety of RM forecasting, un-constraining, and optimization techniques. The simulator allows analysts to study the economic benefit of RM systems under predefined assumptions in an artificial and controlled environment. This increases the benefits obtained by the correct selection of context-appropriate RM systems and the likelihood of successfully implementing new and complex systems. We test and showcase the simulator performance, studying the entrance of a new airline in a competitive context. We generate, implement and evaluate different RM strategies in response to the introduction of new competition, and discuss the results, highlighting the interpretability and accuracy of the proposed framework.

@en

In this paper, we present a combined forecasting and optimization decision-support tool to assist air cargo revenue management departments in the acceptance/rejection process of incoming cargo bookings. We consider the case of a combination airline and focus on the passenger aircraft belly capacity. The process is dynamic (bookings are received in a discrete fashion during the booking horizon) and uncertain (for some bookings the three dimensions are not provided, while the actual belly space available for cargo is only revealed a few hours before departure). Hence, analysts base decisions on historical data or human experience, which might yield sub-optimal or infeasible solutions due to the aforementioned uncertainties. We tackle them by proposing data-driven algorithms to predict available cargo space and shipment dimensions. A packing problem is solved sequentially once a new booking request is received, predicting shipment dimensions, if necessary, and considering the uncertainty of such prediction. The booking is accepted if it results in a feasible loading configuration where no previously accepted booking is offloaded. When applied in a deterministic context, our packing method outperformed the one used by the partner airline, increasing the loaded volume up to 20%. The framework was also tested assuming unknown shipment dimensions, comparing a risk-prone and a risk-averse strategy, with the latter accounting for uncertainty in dimension predictions and the former using mean values. While the average loaded volume decreases in the risk-averse case, the number of unplanned offloadings due to under-predicted dimensions decreases from 54% to 12% of the simulated cases, hence yielding a more robust acceptance strategy.@en

This work addresses the cockpit crew training scheduling problem. The objective is to produce a robust cockpit crew training schedule, including the assignment of trainees, instructors and simulators. To attain this objective, we propose a scheduling framework composed of four modules: a Training Scheduling & Assignment Model (TS&AM), a Disruption Generator (DG), a Rule-Based Recovery (RBR) algorithm, and a Neural Network (NN). The TS&AM is an integer programming model that integrates the scheduling of courses and the assignment of resources. The output roster serves as input for a data-driven DG based on Monte-Carlo Simulation. The disruptions are then solved using the RBR algorithm. Finally, The NN feedback algorithm learns the recovery costs experienced in the disruption impact simulator and updates these costs in the TS&AM to generate more robust rosters. The proposed modelling framework was calibrated, tested, and demonstrated in a simulation environment developed using four years of historical crew training data from a major European airline. The experiment showed that our approach outperformed the roster produced by the airline. The approach proposed produces rosters that reduce recovery costs by 21 percent, while still decreasing total training costs by 3 percent.

@en

Metrics play an important part in the development and application of prognostic methodologies as they provide the capability to characterize and assess the quality of remaining useful life predictions. Although there is a wide range of both, prognostic metrics and prognostic methodologies available, the choice of those often is a demanding and time consuming task. Additionally, they are often treated as two separate problems to solve, while the choice of metrics has an impact on the choice of prognostic methodology and vice versa. In this paper, we therefore present a framework with the capability to automatically choose prognostic settings given specific system data to account for five different prognostic metrics. We then apply this framework to an aircraft data set to characterize the impact of metrics on the choice of prognostic methodologies. The results show that the choice of optimization metric has a big impact on the output of the generic prognostic framework and on the overall prognostic performance.@en