Browsing by Author "King, Stephen"

Now showing 1 - 6 of 6
  • Thumbnail Image
    ItemOpen Access
    Failures mapping for aircraft electrical actuation system health management
    (PHM Society, 2022-06-29) Wang, Chengwei; Fan, Ip-Shing; King, Stephen
    This paper presents the different types of failure that may occur in flight control electrical actuation systems. Within an aircraft, actuation systems are essential to deliver physical actions. Large actuators operate the landing gears and small actuators adjust passenger seats. As developing, aircraft systems have become more electrical to reduce the weight and complexity of hydraulic circuits, which could improve fuel efficiency and lower NOx emissions. Electrical Actuation (EA) are one of those newly electrified systems. It can be categorized into two types, Electro-Hydraulic Actuation (EHA) and Electro-Mechanical Actuation (EMA) systems. Emerging electric and hydrogen fuel aircraft will rely on all-electric actuation. While electrical actuation seems simpler than hydraulic at the systems level, the subsystems and components are more varied and complex. The aim of the overall project is to develop a highly representative Digital Twin (DT) for predictive maintenance of electrical flight control systems. A comprehensive understanding of actuation system failure characteristics is fundamental for effective design and maintenance. This research focuses on the flight control systems including the ailerons, rudders, flaps, spoilers, and related systems. The study uses the Cranfield University Boeing 737 as the basis to elaborate the different types of actuators in the flight control system. The Aircraft Maintenance Manual (AMM) provides a baseline for current maintenance practices, effort, and costs. Equivalent EHA and EMA to replace the 737 systems are evaluated. In this paper, the components and their failure characteristics are elaborated in a matrix. The approach to model these characteristics in DT for aircraft flight control system health management is discussed. This paper contributes to the design, operation and support of aircraft systems.
  • Thumbnail Image
    ItemOpen Access
    Intelligent fault diagnosis of an aircraft fuel system using machine learning - a literature review
    (MDPI, 2023-04-16) Li, Jiajin; King, Stephen; Jennions, Ian
    The fuel system, which aims to provide sufficient fuel to the engine to maintain thrust and power, is one of the most critical systems in the aircraft. However, possible degradation modes, such as leakage and blockage, can lead to component failure, affect performance, and even cause serious accidents. As an advanced maintenance strategy, Condition Based Maintenance (CBM) can provide effective coverage, by combining state-of-the-art sensors with data acquisition and analysis techniques to guide maintenance before the asset’s degradation becomes serious. Artificial Intelligence (AI), particularly machine learning (ML), has proved effective in supporting CBM, for analyzing data and generating predictions regarding the asset’s health condition, thus influencing maintenance plans. However, from an engineering perspective, the output of ML algorithms, usually in the form of data-driven neural networks, has come into question in practice, as it can be non-intuitive and lacks the ability to provide unambiguous engineering signals to maintainers, making it difficult to trust. Engineers are interested in a deterministic decision-making process and how it is being revealed; algorithms should be able to certify and convince engineers to approve recommended actions. Explainable AI (XAI) has emerged as a potential solution, providing some of the logic on how the output is derived from the input given, which may help users understand the diagnostic result of the algorithm. In order to inspire and advise data scientists and engineers who are about to develop and use AI approaches in fuel systems, this paper explores the literature of experiment, simulation, and AI-based diagnostics for the fuel system to make an informed statement as to the progress that has been made in intelligent fault diagnostics for fuel systems, emphasizing the necessity of giving unambiguous engineering signals to maintainers, as well as highlighting potential areas for future research.
  • Thumbnail Image
    ItemOpen Access
    Landing gear health assessment: synergising flight data analysis with theoretical prognostics in a hybrid assessment approach
    (PHM Society, 2024-06-27) El Mir, Haroun; King, Stephen; Skote, Martin; Alam, Mushfiqul; Place, Simon
    This study addresses a critical shortfall in aircraft landing gear (LG) maintenance: the challenge of detecting degradation that necessitates intervention between scheduled maintenance intervals, particularly in the absence of hard landings. To address this issue, we introduce a Performance Degradation Metric (PDM) utilising Flight Data Recorder (FDR) output during the touchdown and initial roll phases of landing. This metric correlates time-series accelerometer data from a Saab 340B aircraft’s onboard sensors with non-linear response dynamic models that predict expected LG travel and reaction profiles across a set of ground contact cycles within a single landing. This facilitates the early detection of deviations from standard LG response behaviour, pinpointing potential performance abnormalities. The initiator of this approach is the Landing Sequence Typology, which systematically decomposes each aircraft landing into successive dynamic periods defined by their representative boundary conditions. What follows is the setting of initial parameters for the ordinary differential equations (ODE)s of motion that determine the orientation and impact responses of the most critical components of the LG assembly. Solving these ODEs with the integration of a non-linear representation of an oleo-pneumatic shock absorber model compliant with CS25 aircraft standards produces anticipated profiles of LG travel based on factors such as aircraft weight and speed at touchdown, which are subsequently cross-referenced with real accelerometer data, enhanced by video footage analysis. This footage is crucial for verifying the sequence of LG touchdowns and corresponding accelerometer outputs, thereby bolstering the precision of our analysis. Upon the conclusion of this study, by facilitating the early identification of LG performance deviations in specific landing scenarios, this diagnostic tool shall enable timely maintenance interventions. This proactive approach not only mitigates the risk of damage escalation to other components but also transitions main LG maintenance practices from reactive to proactive.
  • Thumbnail Image
    ItemOpen Access
    Machine learning requirements for the airworthiness of structural health monitoring systems in aircraft
    (ICAF, 2023-06-30) El Mir, Haroun; King, Stephen; Skote, Martin; Perinpanayagam, Suresh
    In the evolving realm of airworthiness and aircraft maintenance task scheduling, the introduction of data-driven Predictive Maintenance (PdM) and Structural Health Monitoring (SHM) has prompted a paradigm shift, which underscores the profound implications of innovative sensing techniques within damage and operational monitoring. Concurrently, the role of avionics in data acquisition and processing has drawn renewed focus, with machine learning (ML) algorithms facilitating pattern recognition, trend analysis, and anomaly detection. This paper discusses the diagnostic sequence in SHM systems, the necessity for damage information, and delves into active and passive sensing techniques within damage and operational monitoring. The role of avionics is also emphasized, especially in data acquisition and processing for operational monitoring. The utilization of ML algorithms for efficient use within SHM is explored, alongside supervised and unsupervised learning methods. The paper underlines how integrating ML in aircraft systems applications can optimize maintenance schedules and lay a solid foundation for SHM integration in aircraft health systems. The study also covers the application of ML techniques for detection, localization, and assessment of structural damage. It reviews research implementations using ML, statistical, and hybrid approaches in monitoring and predicting aircraft damage. The incorporation of non- exclusive ML in SHM to minimize environmental feature uncertainty and enable trackable model behaviour is illustrated. Lastly, the paper discusses evolving regulatory requirements and standards for ML application in aviation SHM, provided by authorities and workgroups like EASA and the SAE G-34 AI in Aviation Committee, respectively, and concludes with an overview of the future trends and standards in this dynamic domain. The aim is to spotlight the transformative potential of PdM and SHM, and their critical roles in boosting the operational efficiency of the aviation industry.
  • Thumbnail Image
    ItemOpen Access
    A review of digital twin for vehicle predictive maintenance system
    (Society of Automotive Engineers, 2023-03-07) Wang, Chengwei; Fan, Ip-Shing; King, Stephen
    The development of Digital Twin (DT) has become popular. A dominant description of DT is that it is a software representation that mimics a physical object to portray its real-world performance and operating conditions of an asset. It uses near real-time data captured from the asset and enables proactive optimal operation decisions. There are many other definitions of DT, but not many explicit evaluations of DT performance found in literature. The authors have an interest to investigate and evaluate the quality and stability of appropriate DT techniques in real world aircraft Maintenance, Repair, and overhaul (MRO) activities. This paper reviews the origin of DT concept, the evolution and development of recent DT technologies. Examples of DTs in aircraft systems and transferable knowledge in related vehicle industries are collated. The paper contrasts the benefits and bottlenecks of the two categories of DT methods, Data-Driven (DDDT) and Model-Based (MBDT) models. The paper evaluates the applicability of the two models to represent vehicle system management. The authors present their methodological approach on Predictive Maintenance (PM) development basing on reliable DT models for vehicle systems. This paper contributes to design, operation, and support of aircraft/vehicle systems.
  • Thumbnail Image
    ItemOpen Access
    Smart Robust Feature Selection (SoFt) for imbalanced and heterogeneous data
    (Elsevier, 2021-11-30) Lee, Gary Kee Khoon; Kasim, Henry; Sirigina, Rajendra Prasad; How, Shannon Shi Qi; King, Stephen; Hung, Terence Gih Guang
    Designing a smart and robust predictive model that can deal with imbalanced data and a heterogeneous set of features is paramount to its widespread adoption by practitioners. By smart, we mean the model is either parameter-free or works well with default parameters, avoiding the challenge of parameter tuning. Furthermore, a robust model should consistently achieve high accuracy regardless of any dataset (imbalance, heterogeneous set of features) or domain (such as medical, financial). To this end, a computationally inexpensive and yet robust predictive model named smart robust feature selection (SoFt) is proposed. SoFt involves selecting a learning algorithm and designing a filtering-based feature selection algorithm named multi evaluation criteria and Pareto (MECP). Two state-of-the-art gradient boosting methods (GBMs), CatBoost and H2O GBM, are considered potential candidates for learning algorithms. CatBoost is selected over H2O GBM due to its robustness with both default and tuned parameters. The MECP uses multiple parameter-free feature scores to rank the features. SoFt is validated against CatBoost with a full feature set and wrapper-based CatBoost. SoFt is robust and consistent for imbalanced datasets, i.e., average value and standard deviation of log loss are low across different folds of K-fold cross-validation. Features selected by MECP are also consistent, i.e., features selected by SoFt and wrapper-based CatBoost are consistent across different folds, demonstrating the effectiveness of MECP. For balanced datasets, MECP selects too few features, and hence, the log loss of SoFt is significantly higher than CatBoost with a full feature set.