11:00
S2: Regulatory Aspects of Condition Based Maintenance at Amphi Fournel (AF)
Chair: Robert Meissner and Alberto Cardoso
11:00
20 mins
|
A review of regulatory approaches for condition-based maintenance in civil aviation
Robert Meissner, Kai Wicke, Gerko Wende
Abstract: One of the biggest innovation fields in the aviation industry is currently its digital transformation. With the vast technological challenges associated, various aspects of a digitalized aerospace sector have been subject to numerous research fields already. Since aircraft maintenance contributes significant portions to the overall operating expenditures for an airline, it has received substantial research focus in the past decade. In particular, the introduction of Condition Based Maintenance (CBM) strategies promise substantial cost saving potentials by avoiding unscheduled maintenance events due to unforeseen failures and reducing scheduled maintenance efforts such as repetitive inspections or functional checks.
Consequently, much research has focused the development of technical capabilities for an automated condition monitoring, fault diagnosis, and failure prognosis. However, still, the application of these technologies as certified scheduled maintenance alternate is often limited to non-safety-critical functions. While the need for change in the regulatory framework has been recognized by industry and regulatory bodies, the developed ideas for a potential pathway for certification leave issued that will need to be addressed. Among others, these are:
• IP180 & IP211 have introduced a Level 3 analysis to the industry-standard MSG-3 methodology to identify an Aircraft Health Management (AHM) alternate for a scheduled maintenance task. However, while it requires to determine the effectiveness and timeliness of failure detections by AHM technologies, no further guidance is given how this determination can be made.
• Arguably, it will be necessary for future AHM alternates to provide the capability of a maintenance ground interface – be it because of remote vehicle operations or because of extensive computation requirements. The current ideas mentioned above do not address this off-board processing and maintenance decision making.
• Currently, the definition of scheduled maintenance tasks is performed as part of the Maintenance Review Board (MRB) process in accordance with CS25.1529 – mostly separated from the system design process as of CS25.1309. However, since AHM technologies will need to be an integral part of any future system design concepts (in alignment with the established OSA CBM model for the development of End-to-End processes), this strict distinction will need to be critically evaluated.
Therefore, with this presentation, we will review existing standards for the certification of health monitoring technologies – such as the Health and Usage Monitoring System (HUMS) for rotorcraft applications – and map their key characteristics with the steps of an End-to-End automated maintenance process. As a consequence, we can identify, how the existing solutions can be applied to the certification of scheduled maintenance for commercial fixed-wing aircraft and what key pieces are missing and will need to be defined by future research. Additionally, we will be able to derive recommendation for the technological maturation of automated condition monitoring technologies to expedite the transfer from research to (certified) commercial applications.
|
11:20
20 mins
|
Maintenance Policy Optimisation with Reinforcement Learning
Haonan Li, Marta Ribeiro, Bruno Santos
Abstract: This study endeavors to optimize the long-term maintenance scheduling practices within the airline industry through the application of advanced Deep Reinforcement Learning (DRL) algorithms. By meticulously identifying and incorporating pivotal decision variables such as block quantities, intervals, and durations, the research endeavors to ascertain an optimal maintenance policy. Leveraging a meticulously crafted reward function to guide the training of the DRL model, the primary aim is to minimize slot utilization, diminish ground time, and homogenize block durations. The proposed framework represents a pioneering endeavor in addressing the intricate challenges inherent in maintenance scheduling, offering profound potential for substantial cost mitigation and operational efficiency enhancements. Importantly, this approach exhibits versatility, rendering it applicable across diverse aircraft typologies and operational contexts.
|
11:40
20 mins
|
Assessment of maintenance slot policies for an aircraft fleet using Value of Information in a Condition-Based Maintenance environment
Iordanis Tseremoglou, Dimitrios Zarouchas, Bruno F. Santos
Abstract: Aircraft maintenance is a critical aspect of airline operations, influencing both safety and profitability. Traditional maintenance practices currently rely on the preventive or the corrective approach where tasks are performed at fixed intervals or when components fail. However, advancements in sensing technologies and prognostics algorithms have enabled the estimation of Remaining Useful Life (RUL) for individual components, encouraging the use of Condition-Based Maintenance.
Despite these advancements, the accuracy and the uncertainty associated with the RUL predictions pose challenges for optimizing maintenance operations. A fundamental consideration in devising effective maintenance slot policies is the acknowledgment that varying levels of accuracy and uncertainty necessitate tailored approaches. For RUL predictions characterized by high levels of certainty and accuracy, a conservative slot policy with less frequent slot repetitions may suffice. Conversely, RUL predictions with higher uncertainty, demand more frequent slot repetition to accommodate potential unforeseen issues. By recognizing the nuanced nature of accuracy and uncertainty within CBM operations, airlines can devise optimized slot repetition policies that enhance operational efficiency. Therefore, it is important to choose the most appropriate maintenance slot policy in order to realize the benefits of CBM effectively.
An important discussion in this direction, is whether the benefits of the CBM can be quantified in terms of related maintenance costs. This discussion can be summarized in the question ‘how much information is worth’?[1]. To this end, it is necessary that the most suitable maintenance policy for the achieved levels of accuracy and uncertainty of the RUL predictions is defined on the basis of the expected value-based gains that reflect quantitative metrics such as the Value of Information (VoI). VoI is formally calculated as the difference in expected costs between the outcome of optimal decisions that may be taken with and without the collection of additional information [3].
However, in the context of CBM in an airline environment, in order to capture the benefits of incorporating RUL predictions of a specific accuracy and uncertainty level, we should extend the VoI analysis beyond solely accounting for the maintenance costs of the related monitored components. Since different accuracy and uncertainty levels may necessitate different effective slot policies, VoI should also include the costs (if any) of introducing and utilizing additional maintenance slots. By incorporating this holistic VoI approach into the policy definition, airlines can effectively evaluate the trade-offs between investing in advanced sensor systems and relying (or modifying) current maintenance slot practices. Conversely, the decision to invest in such technology must be balanced against the costs of utilizing extra slots, ensuring that the overall cost-benefit analysis remains favorable. Consequently, integrating both cost considerations in the VoI analysis enables airlines to make informed decisions that optimize maintenance slot allocation, and strategically allocate resources towards sensor investments that offer long-term value.
In this work, we introduce a novel approach that leverages such a VoI analysis to evaluate different maintenance slot policies for an aircraft fleet in a CBM environment, considering components whose Remaining Useful Life (RUL) is predicted with various levels of uncertainty and accuracy.
The methodology begins by simulating maintenance tasks using historical airline data. Next, using also historical airline data, we simulate multiple maintenance slot policies, having various slot repetition frequencies. Furthermore, we simulate RUL predictions by applying the Support Vector Regression model developed in [2] on the C-MAPSS dataset for turbo-fan engines. Also, similarly to [4] different levels of uncertainty and accuracy are modeled to capture the inherent variability in the outcome of the prognostics models.
Using the scheduling model developed in [4], we calculate and compare the maintenance costs induced when following the traditional corrective approach and the prognostics-driven approach. The maintenance costs include both the direct maintenance costs of the components and the indirect costs (manpower costs) of the corresponding slot policy. From this comparison, we extract the VoI associated with integrating RUL predictions in each maintenance slot policy.
The results of the study demonstrate that the VoI varies significantly (up to 50%) with respect to the different levels of accuracy and uncertainty, the percentage of the prognostics-driven tasks to the total preventive and corrective tasks, the maintenance costs of the monitored components and the considered slot policy. We use this VoI analysis to identify for each case different investment thresholds related to the considered maintenance slot policy. These insights can guide decision makers in selecting the most appropriate slot policy, tailored to a specific level of accuracy and uncertainty of the prognostics model, ensuring at the same time alignment with the available budget. Overall, this paper contributes to advancing the state-of-the-art in CBM optimization for aircraft fleet, offering practical insights for improving maintenance operations in the aviation industry.
References
1. Andriotis C.P., Papakonstantinou K.G., Chatzi E.N., 2021,“Value of structural health information in partially observable stochastic environments”, Structural Safety, 93.
2. Bieber M., Verhagen W., Santos B.F. 2021, “An Adaptive Framework For Remaining Useful Life Predictions of Aircraft Systems”. In: PHM Society European conference, vol. 6., p. 60–70.
3. Fauriat Z., Zio E., 2020, “Optimization of an aperiodic sequential inspection and condition-based maintenance policy driven by value of information”, Reliability Engineering & System Safety,204.
4. Tseremoglou I., Santos B.F., 2024, “Condition-Based Maintenance scheduling of an aircraft fleet under partial observability: A Deep Reinforcement Learning approach”, Reliability Engineering & System Safety, 241
|
12:00
20 mins
|
Stochastic Optimisation of Tail Assignment and Maintenance Task Scheduling with Health-Aware Models
Benno Käslin, Melissa Oremans, Manuel Arias Chao, Marta Ribeiro, Bruno Santos
Abstract: See PDF file.
|
12:20
20 mins
|
Effects of aeroplane digital twins on business models
Hendrik Meyer, Ann-Kathrin Koschlik, Florian Raddatz, Gerko Wende
Abstract: In simple terms, a digital twin can be defined as follows: The Digital Twin is a uniquely
identifiable digital representation of a physical or logical object for one or more purposes. The topic of digital twins is becoming increasingly important. Not only publications (see Figure 1) on the topic but also search queries in search engines are continuously increasing. The potential of digital twins is generally considered to be very high. It is reflected in various political, scientific and economic strategies. However, the question of financing digital twins has not yet been fully clarified. There appear to be two problems in particular:
1) The costs and benefits of digital twins can be allocated at different stakeholders.
2. the benefits of digital twins are often not realised directly in the digital twin, but in
surrounding processes and often at stakeholders other than those providing the required information.
These two boundary conditions require that the development and operating costs of digital twins must be distributed among the individual market participants in proportion to the benefits.
|
|