Τμήμα Μηχανολόγων και Αεροναυπηγών Μηχαν. (ΔΔ)

Permanent URI for this collection

https://hdl.handle.net/10889/25

Browse

Now showing 1 - 20 of 296

Open Access
Physics-based vibration analysis for fault diagnosis in rotor-bearing systems under multiple fault conditions
(0021-10-26) Τσέλιος, Ιωάννης; Tselios, Ioannis
Rotating machinery serves as the operational backbone of many industrial systems. Reliability, performance, and safety rely on understanding vibration behavior under real operating conditions. However, rotor-bearing systems often exhibit complex dynamic behavior arising from interactions among multiple fault mechanisms, including mass unbalance, transverse breathing shaft cracks, and bearing misalignment. Their combined presence can produce complex vibration responses that cannot always be interpreted through the characteristics of the individual faults. This thesis investigates the dynamic behavior of rotor-bearing systems under multiple fault conditions and uses physics-based numerical models, experimental measurements, signal-processing techniques, and Artificial Intelligence algorithms to support fault diagnosis. The numerical models account for mass unbalance, transverse breathing and fully open shaft cracks, and hydrodynamic journal bearings. Crack-induced flexibility is calculated using fracture-mechanics-based formulations, while the nonlinear bearing forces are obtained through the numerical solution of the Reynolds equation. Bearing misalignment is included by modifying the lubricant-film geometry along the bearing length. The developed models are used to investigate the effects of fault severity, rotational speed, crack-front geometry, torsional excitation, and combined fault conditions on the system’s vibration behavior. Experimental measurements were obtained from a laboratory rotor test rig under controlled unbalance conditions and with machined transverse shaft slots representing cracks. Steady-state and run-up measurements are compared with the numerical predictions using time histories, frequency spectra, and time-frequency representations. The comparisons show that the models reproduce the principal speed-dependent vibration characteristics of the physical rotor under the investigated conditions, while also identifying discrepancies associated with uncertain system parameters, modeling assumptions, and measurement-related factors. Experimental measurements were also used to update selected model parameters and fine-tune diagnostic classifiers. For fault diagnosis, Artificial Intelligence algorithms are used together with physics-based models and vibration features. Model updating and numerical data generation are combined with an Artificial Neural Network for unbalance estimation. Instantaneous-frequency-based features are used to estimate crack and unbalance characteristics in a numerical study. Numerical pre-training and experimental fine-tuning are also used for the classification of crack condition, unbalance location, and unbalance level. The results support fault estimation and classification within the examined fault ranges and operating conditions. Overall, the thesis focuses on the vibration analysis and physical interpretation of rotor-bearing systems under multiple fault conditions. The numerical and experimental findings are used to identify meaningful vibration features and support the estimation and classification of fault characteristics. In this way, the work connects the physical understanding of fault interactions with vibration-based fault diagnosis.
Open Access
A design methodology for conceptual and preliminary design of electric class 1 UAS : application to a canister launched fixed-wing platform with foldable wing mechanism
(2026-07-23) Καρατζάς, Ελευθέριος; Karatzas, Eleftherios
Continuous research and development activity in the emerging field of unmanned aerial vehicles (UAVs) is creating new capabilities in the tactical, operational, and strategic environment. Loitering unmanned aerial platforms, which fall under NATO Class 1, have significantly enhanced tactical situational awareness for small tactical teams. These cost-effective applications can be employed for reconnaissance, surveillance, and strike missions. Their ability to be transported and operated by a single operator, without substantially increasing the burden of individual equipment, makes them particularly practical and significantly reduces the required deployment time. Taking these advantages into account, an increasing number of research organizations and defense industries are turning toward the development of canister-launched UAVs. Such launch arrangements may include mortars, recoilless guns, as well as tubes or catapults specifically designed for this purpose. In terms of simplicity, these platforms are usually equipped with electric propulsion systems rather than internal combustion engines. The main reason is the high efficiency of brushless direct current (BLDC) motors, as well as the absence of extensive and complex maintenance requirements. Furthermore, the organization of modern search-and-rescue and observation teams, in conjunction with NATO and European Union standards, highlights limitations in existing systems with respect to establishing a broad network for information collection and surveillance within the areas of interest, influence, and operations.As already noted, the research and development process requires multinational cooperation and significant resources. Organizations, industries, defense and civilian stakeholders, and governments have attempted to lay the foundations for new platforms, with the aim of establishing a common operational and functional system suitable for mass production. The European Union has already launched programs focused on UAV development. Indicatively, the LOTUS and EURODRONEprograms are representative examples of this effort. However, these programs address different operational profiles and entail more complex logistical requirements compared with the present proposal.The use of systems of this type is expected to relieve operators from the limitations of short endurance multicopters and fixed-wing aircraft with demanding assembly and take-off procedures.The subject of the present dissertation is the development of a canister-launched UAV that can be carried in a backpack. This approach simplifies the operational challenges associated with the need for specialized equipment and additional means. At the same time, it makes this technology more accessible and familiar to small reconnaissance teams, without dependence on external launch platforms or obstacle-free take-off and landing areas. The added value of the present dissertation is summarized as follows:• The approach is not limited to a theoretical treatment but establishes an immediately applicable methodological framework for the development and operational employment of electricunmanned aerial vehicles.• It contributes substantially to design optimization by highlighting critical trade-offs among performance, sizing, and operational requirements.• It proposes a tailored methodology that supports faster and better-substantiated engineering decisions from the early stages of development, thereby reducing design uncertainty.• It demonstrates a high degree of scalability, as it can be adapted to different UAV classes, alternative mission profiles, and diverse operating conditions.• It integrates principles of systems engineering into a unified framework, connecting individual scientific domains such as propulsion, energy storage, aerodynamics, lightweight structures, and mission requirements.• It establishes a structured methodology with clear value for both future academic research and industrial practice.To define the fundamental design parameters of the proposed configuration, a systematicexamination of existing platforms available on the international market is carried out, supported by an extensive literature review and comparative analysis. Since design approaches differ significantly depending on the selected launch platform, this analysis provides a structured understanding of prevailing technological trends, limitations, and conceptual design directions. On this basis, the mission profile is formulated with emphasis on the operational requirements of small teams, prioritizing ease of deployment, logistical simplicity, flexibility, as well as increased range, endurance, and robustness for the aforementioned missions.The main sizing parameters of the air vehicle, including weight, power, geometry, stability, and structural strength, are subsequently determined on the basis of statistical and historical data, with particular emphasis on battery technology selection due to its decisive influence on electric propulsion performance. Following the feasibility analysis, a code developed and presented in the present dissertation is employed for the parametric mathematical analysis of the conceptual and preliminary design phases, enabling progression toward the detailed design stage in a rapid and structured manner.During the detailed design phase, data derived from aerodynamic analyses (CFD using ANSYSFluent, with mesh-convergence verification) and finite element structural analyses (FEA using MSC Software, with an inertial approach for simulating flight loads) are employed for both the aircraft structure and the folding mechanism, in order to demonstrate that the developed code mitigates calculation errors and enhances the reliability of preliminary estimates. The structural analysis results confirmed the suitability of composite-material construction (CFRP 2Layers Woven / Rohacell 31A core), with a final fuselage weight of 812.6 g and folding-mechanism weight of 200 g, providing a significant margin for additional payload, particularly EO/IR sensors, encrypted data links, or higher-capacity batteries. Furthermore, commercially available components and subsystems are used in order to preserve the cost-effective character of the system.The experimental phase of the dissertation was implemented through structured testing, following the V-Model methodology and mapping each dissertation chapter to a design phase (CONOPS,Preliminary Design, Critical Design, Integration, and Test). Three successive prototype versions were built and tested (Technology Demonstrator v0.1, v0.2, and v0.3), as well as the final v1.0 version, with no structural failure occurring in any of the prototypes. The first successful flight, or maiden flight, of the v1.0 air vehicle took place at Tanagra Air Base (LGTG) under controlled conditions (take-off speed of 80 km/h, runway length of 50 m, and maximum bank angle of 30 degrees). The test confirmed the airworthiness of the platform, the smooth response of the propulsion system, and flight stability, while also providing clear evidence to inform the redesign phase. On the basis of these flight data, a second design cycle was carried out with a revised mission profile (45 min endurance, 20 km range, MTOW approximately 3.8 kg), an increased fuselage diameter (140 mm), a V-tail instead of a conventional tail to simplify the folding mechanism, and a thicker airfoil to allow the internal integration of servomotors.This second cycle directly demonstrates the iterative applicability of the developed methodology and code to different configurations and mission requirements.
Open Access
Μεθοδολογία σχεδιασμού και υλοποίησης ευέλικτων ρομποτικών σταθμών εργασίας για αναδιαμορφώσιμα συστήματα παραγωγής
(2026-06-29) Ανδρωνάς, Διονύσιος; Andronas, Dionysios
Οι σύγχρονες απαιτήσεις της αγοράς θέτουν σημαντικές προκλήσεις για τα συστήματα παραγωγής, τα οποία καλούνται να ανταπεξέλθουν στη μεταβλητότητα των προϊόντων, στους συντομότερους κύκλους ζωής τους και στην αυξανόμενη ζήτηση για εξατομίκευση των αγαθών. Οι συμβατικές γραμμές συναρμολόγησης, αν και αποδοτικές για μαζική παραγωγή, δυσκολεύονται να προσαρμοστούν στις εξελισσόμενες απαιτήσεις της αγοράς λόγω της σκληρής αυτοματοποίησης και του υψηλού κόστους αναδιαμόρφωσης τους. Η παρούσα διατριβή στοχεύει στην αντιμετώπιση αυτών των περιορισμών, προτείνοντας μια μεθοδολογία για τον σχεδιασμό ευέλικτων ρομποτικών σταθμών εργασίας για Επαναδιαμορφώσιμα Συστήματα Παραγωγής (Reconfigurable Manufacturing Systems - RMS). Παρουσιάζεται μια ολοκληρωμένη μεθοδολογία σχεδιασμού RMS, η οποία περιλαμβάνει τις φάσεις της ανάλυσης προδιαγραφών, τη τεχνική ανάπτυξη του συστήματος, της υλοποίησης και της αξιολόγηση του. Η μεθοδολογία αυτή μπορεί να καθοδηγήσει κρίσιμες σχεδιαστικές αποφάσεις για την κάλυψη εξειδικευμένων βιομηχανικών αναγκών, μέσω της κατάλληλης επιλογής, σχεδιασμού και εφαρμογής υποστηρικτικών τεχνολογιών που βασίζονται σε πέντε τεχνολογικούς πυλώνες, ονομαστικά «επιδεξιότητα», «ευφυΐα», «επαναδιαμορφωσιμότητα», «συνεργασία ανθρώπου-ρομπότ» και «ανθρωποκεντρικότητα». Η μεθοδολογία εφαρμόζεται και επικυρώνεται μέσω πέντε βιομηχανικών σεναρίων από τους κλάδους των προϊόντων προσωπικής φροντίδας, της αυτοκινητοβιομηχανίας, των λευκών συσκευών και των σύνθετων υλικών. Τα σενάρια αυτά παρουσιάζουν διαφορετικές ιδιαιτερότητες ως προς τα χαρακτηριστικά και την πολυπλοκότητα των προϊόντων, απαιτώντας προηγμένες τεχνολογίες χειρισμού και αντίληψης. Οι λύσεις που παρουσιάζονται, εφαρμοσμένες είτε σε εργαστηριακά είτε σε βιομηχανικά περιβάλλοντα, αναδεικνύουν πώς οι επαναδιαμορφώσιμοι ρομποτικοί σταθμοί εργασίας μπορούν να επιτύχουν ευελιξία, υψηλή παραγωγικότητα και βελτιωμένη ποιότητα προϊόντων, ενισχύοντας παράλληλα την ασφάλεια και την εργονομία των εργαζομένων μέσω ανθρωποκεντρικών τεχνολογιών. Με στόχο την αύξηση του αντικτύπου της διατριβής, δόθηκε ιδιαίτερη έμφαση σε βιομηχανικούς κλάδους που διαχειρίζονται εύκαμπτα υλικά, των οποίων τα συστήματα παραγωγής παραμένουν κυρίως μη αυτοματοποιημένα λόγω της περιορισμένης ικανότητας των ρομποτικών βραχιόνων να διαχειρίζονται αποτελεσματικά την παραμόρφωση των υλικών αυτών. Τα αποτελέσματα επικαιροποιούν ότι οι προτεινόμενες λύσεις όχι μόνο γεφυρώνουν το χάσμα μεταξύ του αυτοματισμού συναρμολόγησης εύκαμπτων και δύσκαμπτων προϊόντων, αλλά και ανοίγουν τον δρόμο προς ανθρωποκεντρικά, βιώσιμα και ανθεκτικά συστήματα παραγωγής, ευθυγραμμισμένα με τις αρχές της Βιομηχανίας 4.0 και της Βιομηχανίας 5.0.
Open Access
Development of novel and efficient experimental arrangements for the validation of numerical simulations of large-scale aircraft structures
(2026-07-02) Κορδάς, Παναγιώτης; Kordas, Panagiotis
The challenges of experimentally evaluating fuselage structures are addressed in this work by employing and extending the building block approach; a hierarchical methodology progressing from coupons to structural details, subcomponents and finally full-scale components. Experimental campaigns that are performed on full-scale aircraft systems are shown to demand high costs, time and resources and are fraught with risks associated with the required facilities, equipment and instrumentation. The issues that accompany these tests can be mitigated by replacing full-scale aircraft components, such as a fuselage barrel, with representative sections, such as stiffened panels, while preserving key structural features. A digitally assisted design framework of test rig concepts is developed, to define, through finite element simulations appropriate boundary conditions and loading schemes. Novel, cost- and time-efficient experimental installations are designed, manufactured, and validated by carrying out static, buckling, and fatigue tests on sub-scale and full-scale fuselage panels. A comprehensive array of correlation studies is performed between experimental data and numerical predictions to demonstrate that panel-level testing can reliably reproduce critical aspects of fuselage barrel behaviour. The structural response, including buckling, of stiffened panels is assessed when subjected to complex loading combinations. Emphasis is also placed on the investigation of fatigue damage and crack growth that is propagated in a full-scale fuselage panel under variable-amplitude loading, representative of flight conditions. The design and numerical validation methodology that is proposed offers an effective and flexible alternative to full-scale aircraft component testing, focusing on the fuselage barrel case, enabling reliable structural characterization with reduced risk, cost, and development time.
Open Access
Robust on-board monitoring for hunting initiation in railway vehicles within a parametric statistical time series based framework
(2026-07-01) Κρητικάκος, Κυριάκος; Kritikakos, Kyriakos
A longstanding concern in railway operations is hunting instability, owing to its adverse effects on rolling-stock life cycle, ride comfort, and safety, particularly through the associated increased risk of derailment. Railway hunting is the coupled lateral and yaw limit-cycle (periodic) oscillation of the vehicle, physically limited by the wheel flanges. The goal of the present Thesis is the development and thorough assessment of a real-time, robust, and automated framework for the reliable monitoring of hunting instability initiation using on-board acceleration measurements. More specifically, the framework aims at providing effective methods for the following two interrelated problems: (a) early detection of hunting initiation; and (b) continuous assessment (prediction) of proximity to hunting initiation prior to its occurrence. Both problems are of paramount importance: early detection of hunting onset enables timely corrective actions to mitigate safety risks, while proximity assessment, addressed for the first time in this Thesis, provides a preventive capability for operation at higher speeds while maintaining safe margins from hunting onset. In the present framework, effective hunting instability monitoring is achieved through the introduction of innovative detection and proximity indicators. The former exploit dynamical changes associated with the occurrence of hunting (limit-cycle) onset, while the latter act as leading indicators by capturing subtle changes that emerge before hunting onset. These changes are generally distinct from the effect of varying track/vehicle conditions or other operational factors on the dynamics, supporting the robust nature of the introduced hunting monitoring indicators. All indicators are estimated through proper parametric statistical time-series models capable of accounting for stochasticity, non-stationarity, and/or nonlinearity in the underlying dynamics. The first main chapter of the Thesis (Chapter 2) is devoted to the development of two high-fidelity multi-body simulation models: an experimentally validated SIMPACK-based model representing the full vehicle dynamics, and a MATLAB-based model focusing on the lateral, and thus hunting-related, dynamics. In the subsequent two chapters (Chapters 3 and 4), two novel parametric hunting onset detection methods are developed with the aim of circumventing the limitations of the current state-of-the-art, while ensuring excellent and robust detection performance, fully automated tuning, and low computational/hardware complexity. Both methods are developed in scalar (single-sensor) and vector (multi-sensor) versions, relying on innovative Degree-of-Stochasticity (DS) and Damping Ratio (DR) hunting indicators derived from scalar and vector forms of adaptive, stochastic, parametric Recursive AutoRegressive models, denoted as RAR and RVAR models, respectively. The DS indicator quantifies the predictive capability of the R(V)AR model and reflects the reduction in signal stochasticity induced by limit-cycle hunting oscillations, while the DR indicator captures the reduction in the lowest R(V)AR model-based instantaneous damping ratio associated with the hunting limit-cycle oscillation. The best-performing V-DS-based method achieves 100% correct detections with 0% false alarms, a mean detection delay of 0.2 s relative to hunting onset, and excellent robustness to diverse track/vehicle conditions, significantly surpassing current state-of-the-art methods. Motivated by this excellent, early, and robust detection performance, Chapter 5 examines, for the first time in the literature, whether the proximity to hunting onset can be continuously assessed (predicted) prior to its occurrence. To this end, an innovative method, designated as the GM-MLE-based method, is developed, based on a physics-informed grey nonlinear vehicle model and the Maximum Lyapunov Exponent (MLE), which serves as a hunting proximity indicator, quantifying the proximity of the vehicle dynamics to the hunting (limit-cycle) state. It demonstrates excellent and robust performance under various operating scenarios, achieving a 97.1% classification accuracy in hunting proximity assessment. Despite its advantages, the GM-MLE-based method is subject to certain demanding implementation requirements, primarily associated with its physics-informed formulation. In order to circumvent these demanding implementation requirements, Chapter 6 introduces an on-board method of reduced implementation complexity based on the migration of the hunting-mode poles of a linearized representation of the vehicle dynamics toward the stability boundary as hunting (limit-cycle) onset approaches. This is exploited through proper hunting-mode-related proximity indicators, estimated in real-time via adaptive parametric RVAR models, and Long Short-Term Memory (LSTM) based decision making, which enables the simultaneous exploitation and fusion of their temporal evolution (time-history). The method demonstrates very good performance and robustness under various operating scenarios, achieving a classification accuracy of 93.0% in hunting proximity assessment. Nevertheless, the increased applicability of the method is accompanied by reduced performance relative to the GM-MLE-based method of Chapter 5, which achieves higher classification accuracy and finer proximity assessment resolution (10 km/h versus 25 km/h intervals).
Open Access
Integrated design, optimization and flight-test validation of a class I mini unmanned aerial vehicle
(2026-07-08) Νικολάου, Ελευθέριος; Nikolaou, Eleftherios
This doctoral thesis presents the comprehensive design, analysis, optimization, and experimental validation of a Class I mini Unmanned Aerial Vehicle (UAV), developed within a unified multidisciplinary engineering framework. The research addresses the inherent complexity of small-scale UAV design, where aerodynamic, structural, propulsion, stability, control, and thermal management considerations are strongly coupled, particularly under low-Reynolds-number operating conditions. The study begins with the definition of mission requirements and system-level specifications, followed by a structured design process encompassing conceptual, preliminary, and detailed design stages. Particular emphasis is placed on the identification of research gaps in current UAV design methodologies, notably the lack of integrated approaches tailored to mini-UAV configurations and the limited availability of subsystem-specific optimization frameworks. To address these challenges, the thesis develops and implements a suite of advanced methodologies. These include a dedicated aerodynamic and geometric design approach for V-tail configurations, a surrogate-based optimization framework for winglet design, and a multi-fidelity aerodynamic optimization strategy for wing and airfoil development. In addition, a coupled aerostructural optimization methodology is proposed for composite wing structures, alongside a novel integrated approach for the design and optimization of internal cooling ducts, combining analytical modeling, computational fluid dynamics (CFD), and surrogate modeling techniques. The proposed methodologies are applied to the design of a complete UAV platform. High-fidelity numerical tools, including CFD and finite element modeling (FEM), are systematically combined with lower-fidelity methods to achieve an optimal balance between computational efficiency and predictive accuracy. The resulting aircraft is evaluated through detailed aerodynamic, structural, propulsion, performance, and stability analyses. A prototype vehicle is manufactured and subjected to an extensive flight-test campaign. Experimental results demonstrate that the UAV meets the defined mission and performance requirements, exhibiting satisfactory aerodynamic efficiency, stability characteristics, and overall flight behavior. Furthermore, a strong correlation is observed between computational predictions and flight-test data, particularly in cruise and climb conditions, thereby validating the proposed design and optimization framework. In conclusion, this work establishes a robust, multi-fidelity, and optimization-driven methodology for the design of mini-UAVs, bridging the gap between theoretical development and practical implementation. The contributions of this thesis provide both a validated UAV platform and a generalizable framework for future research and development in small-scale unmanned aircraft systems.
Open Access
A methodology for digitalization and optimization of milling processes
(2026-03-06) Σούφλας, Αθανάσιος; Souflas, Athanasios
The scope of the present research is to develop a methodology to support the digitalization of machining, focusing on the optimization of machining process quality, while ensuring their viability for industrial adoption. The current literature has been extensively studied, and three key digitalization barriers have been identified: (i) the disruption of the current workflow of the machining personnel, due to digital tools; (ii) the compatibility of existing approaches with legacy machine tools and (iii) the integration and operational complexity. These findings drive the technical objectives of this research, which has been executed in three main pillars, targeting different quality loss causes. Pillar 1 focuses on the minimization of the machining load-induced error in robot machining. A Multi-Body Simulation of the machining robot has been developed, utilizing Reduced Order Models to minimize computational time. Based on synthetic datasets from this simulation, surrogate optimization techniques are developed, for workpiece placement and posture optimization during robot machining. The simulation and optimization software are tightly interfaced with CAM to minimize disruption of the current workflow of process planners. The benefits of the proposed tools in terms of dimensional accuracy and reduction of robot vibrations are demonstrated by machining a real component. Pillar 2 focuses on the development of an in-process chatter detection system. A retrofittable monitoring setup is developed, based on an industrial accelerometer and DAQ, and is validated regarding the quality of the produced data. Features are extracted from the vibration signals in the frequency domain, where the energy shift of the signal from harmonic frequencies to chatter is tracked, and in the time domain, where the cyclostationary properties of the signal are investigated. The calculated features are input in a Support Vector Machine classifier that detects chatter with high accuracy. The generalizability of the approach is demonstrated through the installation of the system in multiple machine tools. Pillar 3 focuses on the development of an in-process tool wear prediction system. Cutting force and temperature data from an instrumented milling head are used as a basis for tool wear estimation. By leveraging the machining theory on cutting force modelling, relevant features are extracted from the signals and used as input to a Decision Tree classifier to predict tool wear status and a Random Forest regressor to predict the actual flank wear level. The generalizability of the approach is validated by exposing it to new, unknown datasets.
Open Access
Development and evaluation of an optimized solvolysis recycling process for CFRPs : fiber performance and environmental-economic impact
(2026-03-03) Βογιαντζή, Χριστίνα; Vogiantzi, Christina
The widespread adoption of carbon fiber reinforced polymers (CFRPs) in high-performance engineering applications, including aerospace, automotive, wind energy, and sports sectors, has led to a substantial increase in composite waste at the end of product life. CFRPs are favored due to their exceptional specific strength, stiffness, and durability. However, their end-of-life management remains a significant challenge from both technological and environmental perspectives. This challenge is primarily associated with the widespread use of thermosetting polymer matrices, which form irreversible, cross-linked networks that cannot be remelted or reshaped. As a result, conventional recycling and reuse strategies are largely ineffective, leading to landfill or energy recovery practices that do not take advantage of the high embedded value of carbon fibers. At the same time, the production of virgin carbon fibers (VFs) is an extremely energy-intensive process, associated with high cumulative energy demand and significant greenhouse gas emissions. The recovery and reuse of carbon fibers (CFs) from CFRPs therefore represent a critical opportunity to reduce environmental impacts, conserve raw materials, and support the transition toward a circular economy in composite manufacturing. However, achieving this goal requires recycling technologies that are capable not only of effectively removing the polymer matrix but also of maintaining the mechanical and interfacial performance of the recovered fibers. This thesis addresses this challenge by developing an integrated experimental, analytical, and sustainability-oriented framework for the recycling of epoxy-based CFRPs. The work focuses primarily on chemical recycling via solvolysis using environmentally friendly (“green”) solvents, while also examining thermal recycling through pyrolysis as a benchmark technology. Unlike many existing studies that evaluate recycling processes solely based on resin removal efficiency, this research adopts a quality-oriented approach, emphasizing not only the mechanical performance of the recovered fibers but also the environmental footprint and economic viability of the recycling processes. The experimental investigation covers both unidirectional (UD) and three-dimensional (3D) CFRP architectures. In addition to conventional UD laminates, 3D woven CFRPs were considered in order to extend the experimental investigation to composites with more complex fiber architectures. This allowed the evaluation of solvolysis performance and recovered fiber characteristics across different reinforcement configurations, without changing the overall methodological framework of the study. Solvolysis experiments were conducted under controlled subcritical and supercritical conditions. For UD CFRPs, both water and water/ethanol solvent systems were examined, allowing the evaluation of temperature, reaction time, and solvent composition on resin decomposition efficiency and on the properties of the recovered carbon fibers. In contrast, the investigation of 3D woven CFRPs was limited to water-based solvolysis, to assess the response of a more complex fiber architecture under a consistent solvent system. In addition, pyrolysis followed by oxidative post-treatment was applied to the 3D woven CFRP specimens, serving as a reference thermal recycling route for comparison with the corresponding solvolysis results. The recovered carbon fibers were evaluated through morphological, mechanical and interfacial analyses. Surface morphology was examined using scanning electron microscopy (SEM), while mechanical performance was assessed through single-fiber tensile testing to determine tensile strength and Young’s modulus. Interfacial behavior was evaluated using microbond testing to measure interfacial shear strength (IFSS). Based on these experimental results, property retention indices were used to quantify the extent to which the mechanical and interfacial properties of recycled fibers were preserved relative to VFs. The experimental results demonstrate that optimized green solvolysis conditions can achieve near-complete resin removal while maintaining a high retention of the tensile properties of the recovered carbon fibers, typically exceeding 85–90%. In contrast, a more significant reduction was observed in IFSS, indicating a higher sensitivity of fiber–matrix interfacial properties to the recycling process. For UD CFRPs, the use of mixed water/ethanol solvent systems resulted in improved fiber surface condition and comparatively enhanced mechanical behavior relative to solvolysis in pure water. The results further reveal that composite architecture plays a significant role in solvolysis effectiveness, highlighting the need for architecture-sensitive recycling strategies. Alongside the experimental evaluation of recycling performance, this thesis includes a comprehensive sustainability assessment. Life Cycle Assessment (LCA) was utilized to quantify environmental indicators such as global warming potential and cumulative energy demand, accounting for direct process impacts. In parallel, Life Cycle Costing (LCC) was used to evaluate the economic performance of the recycling routes, considering operational costs. A representative CFRP sports racket case study, developed within the framework of the RECREATE project, was used to illustrate the application of the sustainability assessment framework. To combine technical, environmental, and economic aspects within a unified evaluation framework, a multi-criteria decision-making (MCDM) approach based on the TOPSIS method was implemented. This approach enabled the comparative ranking of recycling routes and operating conditions, highlighting the compromises between fiber quality, environmental performance, and economic considerations. In summary, this thesis combines experimental investigation and sustainability assessment for the recycling of CFRPs. Solvolysis is examined for different reinforcement architectures, and recycling performance is evaluated in terms of fiber quality and selected environmental and economic indicators. The proposed assessment framework provides a structured basis for comparing recycling routes under defined technical and sustainability criteria.
Open Access
Πειραματικά επικυρωμένο πολυκλιμακικό αριθμητικό πλαίσιο για την προσομοίωση, βαθμονόμηση και βελτιστοποίηση της διεργασίας FDM και της μηχανικής συμπεριφοράς του παραγόμενου εξαρτήματος : προς την ανάπτυξη ψηφιακού διδύμου
(2026-01-28) Γκέρτζος, Πέτρος; Gkertzos, Petros
Λόγω της οικονομικής ανάπτυξης και της παγκόσμιας διασύνδεσης που προκύπτει από τη σύγχρονη παγκοσμιοποίηση και βιομηχανοποίηση, ο τομέας της μεταποίησης έχει αποκτήσει αυξημένη σημασία κατά την τελευταία δεκαετία. Η αυξανόμενη έμφαση των κατασκευαστών στην παραγωγικότητα, την αποδοτικότητα και τα έσοδα έχει οδηγήσει στην υιοθέτηση προηγμένων συστημάτων παραγωγής στις γραμμές παραγωγής τους. Η προηγμένη παραγωγή αποτελεί έναν όρο-ομπρέλα που επικεντρώνεται κυρίως στις τεχνολογίες συστημάτων παραγωγής και στην προσαρμοστικότητά τους. Τα δεδομένα της αγοράς υποδεικνύουν ότι τα βασικά κίνητρα για επενδύσεις στη μεταποιητική βιομηχανία είναι η μείωση του κόστους παραγωγής και η πρόσβαση σε νέες αγορές μέσω εξατομικευμένων προϊόντων. Η τρισδιάστατη εκτύπωση, δημιουργεί νέες ευκαιρίες ως προς τις δυνατότητες παραγωγής, επιτρέποντας καινοτόμους σχεδιασμούς με σημαντικά μειωμένο κόστος παραγωγής και μικρότερους χρόνους διάθεσης στην αγορά. Μεταξύ των διαφόρων τεχνολογιών τρισδιάστατης εκτύπωσης, η Fused Deposition Modeling (FDM) παρουσιάζει ιδιαίτερο ενδιαφέρον ως προς τη διαθεσιμότητα, το κόστος υλικών και εξοπλισμού, την ταχύτητα κατασκευής και τις απαιτήσεις μετα-επεξεργασίας. Ωστόσο, παρουσιάζει ορισμένους περιορισμούς όσον αφορά τη διαστατική ακρίβεια και τις μηχανικές ιδιότητες των παραγόμενων εξαρτημάτων. Η ύπαρξη πολλαπλών αλληλεπιδρώντων παραμέτρων διεργασίας εισάγει αβεβαιότητες στις ιδιότητες και στη συνολική μηχανική συμπεριφορά των κατασκευασμένων εξαρτημάτων. Επιπλέον, η επιλογή μη βέλτιστων παραμέτρων διεργασίας μπορεί να οδηγήσει σε κατασκευαστικά σφάλματα (π.χ. στρέβλωση εξαρτήματος) ή ακόμη και σε αποτυχημένες εκτυπώσεις, επιβραδύνοντας την παραγωγή και αυξάνοντας το κόστος. Συνεπώς, μια αριθμητική ροή εργασίας της διεργασίας, ικανή να υπερβεί τους χρονικούς και οικονομικούς περιορισμούς των πειραματικών μεθόδων και να επιτύχει ακριβείς προβλέψεις των τελικών ιδιοτήτων των εξαρτημάτων, μπορεί να βελτιώσει και να βελτιστοποιήσει σημαντικά την αποδοτικότητα της παραγωγής. Οι κύριοι στόχοι της παρούσας έρευνας συνοψίζονται ως εξής: Ανάπτυξη πειραματικά επικυρωμένης αριθμητικής μεθοδολογίας για την αναγνώριση και πρόβλεψη της μεσοκλιμακικής γεωμετρίας τρισδιάστατα εκτυπωμένων εξαρτημάτων, με βάση τις παραμέτρους εκτύπωσης. Ανάπτυξη αριθμητικού πλαισίου που ενσωματώνει πολλαπλές παραμέτρους διεργασίας και προσομοιώνει τόσο τη διαδικασία κατασκευής όσο και την κατάσταση του εξαρτήματος όπως κατασκευάστηκε. Τα μοντέλα υψηλής πιστότητας δημιουργούν ένα εκτενές σύνολο δεδομένων, υπερβαίνοντας τους χρονικούς και οικονομικούς περιορισμούς των πειραματικών μεθόδων. Βελτιστοποίηση της παραγωγικής διαδικασίας μέσω βαθμονόμησης των παραμέτρων διεργασίας με τη χρήση υποκατάστατων (surrogate) μοντέλων. Η διατριβή αποτελείται από οκτώ κεφάλαια, τα οποία παρουσιάζονται συνοπτικά παρακάτω: Στο Κεφάλαιο 1 παρουσιάζεται επισκόπηση του θεωρητικού υπόβαθρου, των εφαρμογών και των μεθόδων της τρισδιάστατης εκτύπωσης. Ιδιαίτερη έμφαση δίνεται στα δεδομένα της αγοράς και στη σύγκριση των διαφορετικών τεχνολογιών τρισδιάστατης εκτύπωσης. Παρουσιάζονται επίσης συνοπτικά οι στόχοι της παρούσας διδακτορικής διατριβής. Το Κεφάλαιο 2 επικεντρώνεται στις παραμέτρους διεργασίας και στα υλικά που χρησιμοποιούνται στη FDM. Παρέχεται αναλυτική επεξήγηση κάθε παραμέτρου, της άμεσης επίδρασής της και των διαθέσιμων επιλογών ελέγχου. Όσον αφορά τα υλικά, παρουσιάζονται βασικές πληροφορίες σχετικά με τη σύνθεσή τους, τις τυπικές ιδιότητες, τη δυνατότητα ανακύκλωσης, την εκτυπωσιμότητα και τις απαιτήσεις μετα-επεξεργασίας. Στο Κεφάλαιο 3 παρουσιάζονται βασικά στοιχεία των αριθμητικών μεθόδων που χρησιμοποιούνται στη διατριβή, στους τομείς της υπολογιστικής μηχανικής στερεού και της υπολογιστικής ρευστοδυναμικής. Το Κεφάλαιο 4 περιλαμβάνει συνδυασμένη πειραματική και αριθμητική ανάλυση για τη διερεύνηση της επίδρασης των παραμέτρων εκτύπωσης στη μεσοκλιμακική γεωμετρία των τρισδιάστατα εκτυπωμένων εξαρτημάτων. Κατασκευάζονται δοκίμια με διαφορετικούς συνδυασμούς παραμέτρων και οι εγκάρσιες τομές τους εξετάζονται μέσω μικροσκοπικής απεικόνισης. Αντίστοιχες γεωμετρίες προσομοιώνονται αριθμητικά με χρήση προσέγγισης υπολογιστικής ρευστοδυναμικής (CFD), λαμβάνοντας υπόψη τη μεταβολή φάσης του υλικού. Το Κεφάλαιο 5 παρουσιάζει την ανάπτυξη πειραματικά επικυρωμένου αριθμητικού μοντέλου που εξετάζει την επίδραση πολλαπλών παραμέτρων της διεργασίας FDM. Προσομοιώνεται η σταδιακή εναπόθεση υλικού για διαφορετικούς συνδυασμούς παραμέτρων τεμαχισμού (slicing). Τα αποτελέσματα περιλαμβάνουν τον χρόνο ψύξης στοιχείων, τη μέγιστη πλαστική παραμόρφωση και τη μέγιστη τάση von Mises. Το προκύπτον σύνολο δεδομένων αξιοποιείται μέσω μοντέλων βασισμένων σε δεδομένα για ανάλυση παλινδρόμησης και βελτιστοποίηση της διεργασίας. Το Κεφάλαιο 6 παρουσιάζει μια υβριδική αριθμητική και πειραματική μεθοδολογία για τη διερεύνηση της επίδρασης πολλαπλών παραμέτρων τεμαχισμού στην κατάσταση του εξαρτήματος όπως κατασκευάστηκε. Μια μεσοκλιμακική προσέγγιση βασισμένη σε εντολές G-code αποτυπώνει τη συγκόλληση γειτονικών νηματικών διαδρομών (rasters) και αναπαράγει την τελική κατασκευασμένη κατάσταση. Παρουσιάζεται μεθοδολογία αναγνώρισης ιδιοτήτων υλικού, καθώς και σχήμα ομογενοποίησης για μεγαλύτερα εξαρτήματα. Προσομοιώνονται περιπτώσεις σχεδιασμού με πολλαπλές παραμέτρους διεργασίας και τα αποτελέσματα περιλαμβάνουν τόσο μηχανικές αποκρίσεις όσο και αποκρίσεις διεργασίας. Το προκύπτον σύνολο δεδομένων χρησιμοποιείται για την κατασκευή επιφανειών απόκρισης, επιτρέποντας ακριβείς προβλέψεις και βελτιστοποίηση σε επίπεδο εξαρτήματος. Το Κεφάλαιο 7 παρουσιάζει υβριδική πειραματική και αριθμητική μελέτη των παραγόντων που επηρεάζουν τη θερμοκρασία του ακροφυσίου, με στόχο τη βαθμονόμηση και την επίτευξη της επιθυμητής θερμοκρασίας εκτύπωσης. Εικόνες από υπέρυθρη (IR) κάμερα αποκαλύπτουν απόκλιση μεταξύ της θερμοκρασίας εκτύπωσης που ορίζεται κατά τον τεμαχισμό και της πραγματικής θερμοκρασίας του ακροφυσίου. Εφαρμόζεται επικυρωμένη αριθμητική μεθοδολογία που αποτυπώνει με ακρίβεια τις θερμικές απώλειες ενός τυπικού hotend και προσομοιώνει πολλαπλά σενάρια εκτύπωσης. Το παραγόμενο σύνολο δεδομένων τροφοδοτεί μοντέλα βασισμένα σε δεδομένα για ανάλυση παλινδρόμησης, επιτρέποντας τη βαθμονόμηση της θερμοκρασίας του ακροφυσίου. Στο Κεφάλαιο 8 συνοψίζονται τα κύρια συμπεράσματα της παρούσας εργασίας και διατυπώνονται προτάσεις για μελλοντική έρευνα.
Open Access
Towards certification by analysis of thermoplastic composites for aircraft structures : an integrated experimental and computational approach
(2025-07-28) Ζάραγκας, Θωμάς; Zaragkas, Thomas
In the present thesis, several novel integrated testing and numerical methodologies are developed towards the certification of new thermoplastic composite systems used in primary aerospace structural applications, through the building block approach, reinforcing the need of research community and industries for certification by analysis procedures. The ultimate goal is the partial replacement of the costly and extensive tests, with fast and efficient numerical simulations, reducing the complexity of the design process, the costs and the time, through the scaling down of larger components to simpler tests and partial prediction from numerical approaches such as Finite Element Modelling.
Open Access
A vibration-based methodology for monitoring progressive fatigue damage in thermoplastic coupons : experimental and numerical approach
(2025-10-27) Τσιβουράκη, Νίκη; Tsivouraki, Niki
Over recent decades, composite materials have emerged as foundational elements in the design and manufacturing of advanced aerospace structures, driven by their superior strength-to-weight ratios and tailored mechanical properties. Among them, thermoplastic composites have garnered increasing attention over thermosetting composites and traditional metallic materials, owing to their excellent impact resistance, weldability, superior damage tolerance, and recyclability - key attributes that align with current aerospace performance and sustainability demands. As the aerospace industry continues to adopt composite materials at both component and system levels, the challenge of cyclic fatigue loading throughout a structure’s operational lifespan remains a major area of concern. The inherently anisotropic and heterogeneous microstructure of fiber-reinforced polymer composites leads to the simultaneous evolution of multiple fatigue damage mechanisms - most notably matrix cracking, fiber/matrix debonding, delamination, and fiber breakage—occurring at different scales and in spatially distributed regions. These complex and coupled phenomena necessitate the development of advanced monitoring strategies capable of capturing and interpreting the onset and progression of damage under operational conditions. In this context, Structural Health Monitoring (SHM) techniques have become increasingly vital for aerospace applications, enabling the real-time tracking of damage evolution and supporting decisions related to maintenance, performance assessment, and life prediction. SHM systems, especially those employing vibration-based techniques, offer non-invasive, scalable, and efficient means to detect, localize, classify, and estimate the severity of damage. The integration of SHM with sophisticated signal processing methodologies further enhances the diagnostic and prognostic capabilities of such systems, with broad relevance across mechanical and structural engineering disciplines. The present thesis addresses the problem of monitoring fatigue-induced delamination propagation in thermoplastic composite structures through a vibration-based SHM approach, combining experimental investigations and high-fidelity numerical simulations. The work is organized into three interlinked domains: (i) experimental testing, (ii) numerical modeling, and (iii) signal processing and diagnostic inference. The overarching goal is to establish a robust and transferable methodology capable of continuous, real-time tracking of fatigue damage evolution in thermoplastic composites, thereby bridging the gap between laboratory-scale coupon studies and application-scale structural components. The experimental domain involved a comprehensive suite of tests, including quasi-static tensile tests, both interrupted and non-interrupted fatigue tests, ultrasonic C-scan inspections, random vibration tests, and post-fatigue quasi-static tensile tests for residual strength evaluation. These tests were conducted on LM-PAEK thermoplastic composite coupons reinforced with T700 carbon fibers, chosen for their aerospace-grade mechanical performance. The experimental results were used to characterize the material’s static and fatigue behavior, as well as to generate high-resolution datasets for subsequent dynamic analysis and model validation. The numerical modeling framework employed finite element simulations using LS-DYNA to replicate the pre- and post-fatigue quasi-static tests and to reproduce their mechanical response, as well as to simulate fatigue-induced delamination propagation. A progressive damage model and a user-defined material model were employed for the quasi-static and fatigue simulation processes, respectively. Cohesive Ζone Μodeling (CZM) was used to simulate interfacial failure between laminate layers. A particular emphasis was placed on the numerical replication of dynamic behavior under random vibration loading - a non-trivial challenge addressed in this thesis. The numerical models were rigorously validated at each stage using experimental benchmarks, demonstrating their ability to operate independently and with high predictive accuracy. In the signal processing and diagnostics module, both non-parametric (Welch-based) and parametric (ΜΜ-ARMA, FP-MM) methods were employed to analyze the dynamic response of the test coupons. Matlab was used for the implementation of signal processing techniques. Metrics including two- and three-dimensional spectrograms, spectral distances, power spectral densities, resonant frequencies, and their respective ratios were extracted and analyzed. A structured three-stage diagnostic framework was implemented—comprising detection, classification, and state estimation of damage progression - leveraging an extensive dataset of baseline experimental signals. Clear and monotonic trends in the dynamic metrics were identified as highly sensitive and robust damage indicators. Detection accuracy reached 100\%, while classification accuracy was 80\% with U-MM-ARMA and 95\% with FP-ΜM models. Damage state estimation accuracy approached 98\% utilizing FP-ΜM framework, confirming the potential of these models for high-fidelity SHM applications. To assess the scalability of the proposed methodology, the validated approach was applied to a more complex aerospace-representative structure: a thermoplastic mono-stiffener panel. This upscaling involved both experimental testing—comprising quasi-static compression and compression-compression fatigue tests—and the development of advanced numerical models capturing both mechanical response and dynamic behavior. The comparison of simulation and experimental results revealed excellent agreement in terms of load-bearing capacity, damage morphology, and fracture mode. Moreover, dynamic analyses demonstrated measurable shifts in frequency response characteristics, supporting the development of reliable and interpretable damage indicators suitable for structural-level diagnostics. In conclusion, this thesis presents a validated, scalable methodology for vibration-based fatigue damage monitoring in thermoplastic composites under several types of uncertainty, contributing to the broader goal of intelligent, data-driven structural health management in aerospace systems. The integration of materials science, experimental mechanics, computational modeling, and advanced signal analysis forms a comprehensive framework that supports both current industrial applications and future research in adaptive aerospace structures.
Open Access
Investigation of the stabilization characteristics of liquid-fueled flames under inlet mixture stratification
(2025-09-26) Μητσόπουλος, Ευάγγελος-Παναγιώτης; Mitsopoulos, Evangelos-Panagiotis
Inhomogeneously premixed combustion is a key operational mode in modern turbulent combustion systems, such as gas turbines and internal combustion engines, where fuel/air mixture stratification can enhance flame stabilization, broaden flammability limits, and reduce pollutant emissions. The present work explores the dynamics of stratified, turbulent propane/air and prevaporised n-heptane/air flames, stabilized in a bluff-body burner under globally ultra-lean conditions (ϕ = 0.09–0.1). These flames are examined experimentally and numerically to elucidate the influence of mixture inhomogeneity on flame anchoring, structure, and the diagnostic capabilities of chemiluminescence as a heat release rate marker in flames exhibiting mixture gradients. Three distinct stratification scenarios were investigated for each fuel: rich-to-lean, stoichiometric-to-lean, and stronger-lean to weaker-lean gradients. These cases were selected to explore the directional sensitivity of flame propagation relative to stoichiometry, a parameter known to influence flame speed and reactivity due to effects such as radical and enthalpy enrichment from richer upstream mixtures—often referred to as the “back-support” effect. Propane and n-heptane were chosen as representative fuels to encompass the behavior of both lighter gaseous and heavier liquid hydrocarbons exhibiting strong preferential diffusion effects. The experimental campaign utilized a combination of diagnostic techniques including 2D-2C Particle Image Velocimetry, Mie scattering, infrared spectroscopy, and line-of-sight chemiluminescence imaging of OH* and CH* to capture flow fields and flame topology. These measurements aimed to characterize how stratification affects flame stabilization in the recirculation zone of a conical bluff body and to provide benchmark data for numerical validation. Complementing the experiments, Large Eddy Simulations (LES) with a Dynamically Thickened Flame (DTF) model were performed, employing reduced chemical mechanisms for both propane and n-heptane combustion. Flame characteristics were analyzed through derived quantities such as flame index, alignment angle, and Chemically Explosive Mode Analysis (CEMA), enabling the identification of local burning regimes and their response to mixture gradients. Another key component in this study is the evaluation of chemiluminescent species, OH*, CH*, and CO2*, as topological markers for heat release rate (HRR) in flames propagating through stratified mixtures. Under the quasi-steady-state assumption, their concentrations were extracted via an algebraic model and correlated with HRR fields obtained from LES. Two comparative metrics were used; the Pearson correlation coefficient and benchmarking against well-established HRR indicators such as PLIF-HCO and the OH×CH2O product. Results indicate that chemiluminescence exhibits useful qualitative agreement with HRR for both fuels under investigation, however the effect of depth of field in counterpart experimental measurements can quickly deteriorate the predictive ability of those markers. The flame index was further employed to categorize local flame types and examine their contribution to chemiluminescence-HRR correlation. This study provides novel insights into the behavior of ultra-lean stratified hydrocarbon flames and assesses the feasibility of employing chemiluminescence as a practical diagnostic in complex combustion scenarios. The findings have implications for the design and monitoring of advanced low-emission combustors and support the targeted development of robust diagnostics in turbulent stratified combustion.
Open Access
Reduced order models based on machine learning for large scale parameterized problems in computational mechanics
(2025-06-17) Δρακούλας, Γεώργιος; Drakoulas, Georgios
In the context of the fourth industrial revolution, digital twins (DTs) have emerged as a key technology for optimizing the performance of engineering products and systems. Using real-world data combined with physics-based solutions, the digital twin links the physical with the digital world, resulting in a powerful tool for the virtual sensing of physical observations. In this direction, computational mechanics is commonly applied in aerospace, maritime, energy, and bioengineering to gain a better understanding of the operational performance of the physical component. However, the solution of large-scale, dynamic, nonlinear models require significant computational resources which are prohibitive for real-time applications. Consequently, for the realization of the digital twin concept, surrogate modeling has gained attention, aiming to reduce the simulation cost of high-fidelity models while providing accurate numerical results. Reduced order modeling (ROM), as part of surrogate modeling, is commonly used to reduce the computational complexity of numerical simulations through dimensionality reduction techniques, while accurately capturing the dominant aspects of the physical behavior. In this dissertation, a gallery of novel deterministic and probabilistic datadriven ROMs based on linear algebra techniques, machine learning, and deep learning models is developed, aiming to reduce the evaluation cost of large-scale numerical models dealing with unsteady and steady-state, nonlinear partial differential equations. It is also shown that using the ROMs, optimization, uncertainty quantification, and sensitivity analysis, which are tasks requiring a large number of numerical solutions, can be performed efficiently. The developed models are applied to examine a range of problems related to wave propagation, fluid dynamics, and biomechanics.
Open Access
Mechanical and fracture characterization of polymer CNT nanocomposites using atomistic simulations
(2025-03-10) Δελασούδας, Ιάκωβος; Delasoudas, Iakovos
Computational physics is regarded as a crucial domain in scientific research worldwide, as it bridges the gap between experimental and theoretical studies. It provides a powerful framework for addressing complex problems from perspectives and under conditions that are often inaccessible through experimental methods due to physical or financial constraints. This is particularly evident in materials science, where atomistic and molecular-level insights are essential. Computational methods are employed, not only to complement and guide experimental efforts but also to drive the development of more accurate macroscopic models. The importance of computational modeling in understanding complex systems, such as polymers and nanocomposites, is highlighted by the growing demand for advanced industrial materials with tailored properties for everyday applications. These models elucidate the relationships between structure, properties, processing, and performance. As a result, materials that are more cost-effective, environmentally sustainable, and energy-efficient can be developed. Fracture, is identified as a critical phenomenon that must be explicitly studied to enable the proper design of new materials and structures. A material’s resistance to crack propagation is influenced by processes occurring across multiple length scales. In this dissertation, fracture behavior in carbon-based nanocomposite polymers is examined using atomistic simulations in three different levels. At the first level, interfacial failure between the filler and the matrix is studied in terms of the interfacial shear strength (IFSS). IFSS is commonly determined in molecular dynamics studies through mechanical pull-out experiments by measuring the drop point in stress-strain curves. However, non-periodic boundaries are often employed along the pull-out direction, rendering these approaches susceptible to size-related effects. In this work, an established method based on the kinetics of local failure mechanisms is extended to study the interface between single-wall carbon nanotubes (SWCNTs) and epoxy under fully periodic boundary conditions. Shear deformation simulations are conducted across a wide range of temperatures and shear stresses using molecular dynamics (MD). Failure times are recorded and analyzed within the framework of the extended Boltzmann-Arrhenius-Zhurkov kinetic equation, enabling the determination of various parameters, including IFSS. At the second level, predictions of fracture properties for bulk systems are undertaken. A scale-bridging method is proposed to forecast the fracture behavior of highly cross-linked epoxy networks. Classical force fields are combined with the REACTER protocol in the LAMMPS package, utilizing a parameter for bond breaking based on an experimentally derived distance criterion. Macroscopic fracture energy of the epoxy network is anticipated using a continuum fracture mechanics model developed for fibrils. Additionally, the stress intensity factor is extracted from the analysis. At the third and final level, the methodology developed for bulk systems is extended to predict the fracture properties of carbon-based nanocomposites. This contributes to the overarching goal of the dissertation, which is to accurately predict the fracture properties of polymer nanocomposites using atomistic simulations. The methodology, which combines empirical force fields with experimental data, enhances the computational efficiency of the calculations while ensuring consistency with experimental results.
Open Access
Επίδραση της σύνθεσης και της δομής στη βιοδιάσπαση και την κυτταροσυμβατότητα μικρο-νανοϊνωδών πολυμερικών βιοϋλικών
(2024-11-19) Μητροπούλου, Αθηνά; Mitropoulou, Athina
Δεδομένης της περιορισμένης αναγεννητικής ικανότητας των ανθρώπινων ιστών, πολυάριθμες χειρουργικές επεμβάσεις πραγματοποιούνται καθημερινά σε παγκόσμιο επίπεδο για την αντικατάσταση και αποκατάσταση κατεστραμμένων ιστών. Η αξιοσημείωτη πρόοδος που έχει σημειωθεί τις τελευταίες δεκαετίες στον τομέα της επιστήμης των υλικών, της μηχανικής και της ιατρικής, στοχεύει στην αποκατάσταση της λειτουργικότητας των κατεστραμμένων ιστών μέσω της ενσωμάτωσης κυττάρων σε βιοσυμβατά και βιοαποδομήσιμα πορώδη κατασκευάσματα, γνωστά ως ικριώματα. Τα ικριώματα, ως τρισδιάστατα (3D) υποστρώματα, χρησιμεύουν ως ιδανικά πρότυπα για την αναγέννηση των ιστών, προσομοιώνοντας τη δομή της εξωκυτταρικής μήτρας (Extracellular Matrix, ECM). Μεταξύ των τεχνικών που χρησιμοποιούνται για την κατασκευή των ικριωμάτων, η ηλεκτροϊνοποίηση (Electrospinning) ξεχωρίζει ως ιδανική προσέγγιση λόγω της ικανότητάς της να παράγει δομές που μιμούνται τα χαρακτηριστικά της εξωκυτταρικής μήτρας (ECM). Ωστόσο, οι τύποι των υλικών που θα χρησιμοποιηθούν για την κατασκευή ενός ικριώματος, καθώς και οι λειτουργικές απαιτήσεις που χρειάζεται να πληροί για την εκάστοτε εφαρμογή του, πρέπει να λαμβάνονται σοβαρά υπόψη. Πολλά βιοαποδομήσιμα συνθετικά πολυμερή, είτε υδρόφοβα, όπως η πολυ-ε-καπρολακτόνη (PCL), είτε υδρόφιλα, όπως η πολυ(βινυλική αλκοόλη) (PVA), χρησιμοποιούνται ευρέως ως υποψήφια βιοδραστικά υλικά για την κατασκευή ικριωμάτων μηχανικής ιστών. Ωστόσο, τα ικριώματα που κατασκευάζονται από υδρόφοβα πολυμερή, συχνά παρουσιάζουν περιορισμένη κυτταροσυμβατότητα και δυσκολία στην ενσωμάτωση κυττάρων, ενώ τα υδρόφιλα πολυμερή, παρόλο που έχουν καλή βιοσυμβατότητα, συχνά στερούνται την απαραίτητη μηχανική αντοχή. Αυτοί οι περιορισμοί εμποδίζουν την αποδοτική χρήση των ικριωμάτων στην in vivo αναγέννηση ιστών, καθιστώντας αναγκαία την αναζήτηση καινοτόμων προσεγγίσεων. Ο συνδυασμός υδρόφοβων και υδρόφιλων πολυμερών έχει προταθεί ως λύση για την αντιμετώπιση αυτών των περιορισμών. Η παρούσα διατριβή επιδιώκει να αξιοποιήσει τα πλεονεκτήματα και των δύο τύπων πολυμερών δημιουργώντας σύνθετα ικριώματα μέσω ηλεκτροϊνοποίησης για εφαρμογές στην Ιστομηχανική. Συγκεκριμένα, η συνδυασμένη χρήση της υδρόφοβης PCL και της υδρόφιλης PVA στοχεύει στην ανάπτυξη ικριωμάτων με βελτιωμένες μηχανικές ιδιότητες, αυξημένη κυτταροσυμβατότητα και βιοαποικοδόμηση με ελεγχόμενο ρυθμό, επιτυγχάνοντας έτσι καλύτερη κυτταρική προσκόλληση και ενσωμάτωση, και τελικά προάγοντας την αποτελεσματική αναγέννηση των ιστών in vitro. Για το σκοπό αυτό, αναπτύχθηκαν τρεις ομάδες ικριωμάτων χρησιμοποιώντας την τεχνική της ηλεκτροϊνοποίησης: i) ικριώματα PCL μονής στρώσης (single-layered), ii) σύνθετα ικριώματα τριών διαδοχικών στρωμάτων (PVA/PCL/PVA) (tri-layered), και iii) ομοαξονικά ικριώματα σύνθετης δομής πυρήνα-κελύφους (PVA ως κέλυφος και PCL ως πυρήνας) (core-shell). Διερευνήθηκε πώς η συνεργατική χρήση υδρόφιλων και υδρόφοβων πολυμερών μπορεί να βελτιώσει τις ιδιότητες των ικριωμάτων για τη χρήση τους στην Ιστομηχανική. Η έρευνα επικεντρώθηκε ιδιαίτερα στην αρχιτεκτονική δομή των ικριωμάτων πυρήνα-κελύφους (core-shell), λόγω της καινοτόμου φύσης της και της περιορισμένης αναφοράς στην υπάρχουσα βιβλιογραφία. Η ομοαξονική ηλεκτροϊνοποίηση, η τεχνική που χρησιμοποιείται για τη δημιουργία αυτών των ικριωμάτων, αποτελεί μια σχετικά πρόσφατη εξέλιξη στον τομέα της επιστήμης και της μηχανικής των υλικών και δεν έχει μελετηθεί εκτενώς. Αυτό υπογραμμίζει την καινοτόμο πτυχή της παρούσας έρευνας, καθιστώντας την ιδιαίτερα σημαντική για την περαιτέρω ανάπτυξη και κατανόηση των δυνατοτήτων αυτής της πολλά υποσχόμενης τεχνικής. Στα πλαίσια της μελέτης, οι δομικές και εμβιομηχανικές ιδιότητες των ικριωμάτων συγκρίθηκαν διεξοδικά πριν και μετά την υδρολυτική αποικοδόμησή τους, η οποία διεξήχθη σε περίοδο επτά μηνών μετά από εμβάπτιση σε αλατούχο διάλυμα ρυθμισμένο με φωσφορικά (PBS) στους 37°C. Επιπροσθέτως, τα ικριώματα υποβλήθηκαν σε in vitro δοκιμές, όπου ανθρώπινα εμβρυϊκά νεφρικά κύτταρα (HEK-293) καλλιεργήθηκαν επάνω σε αυτά για να διερευνηθεί η πιθανή προσκόλληση, διήθηση και πολλαπλασιασμός των κυττάρων. Τα βασικά ευρήματα από τη μελέτη περιλαμβάνουν την επίτευξη εξαιρετικά ομοιογενούς μεγέθους και κατανομής ινών σε όλους τους τύπους ικριωμάτων, χάρη στις βελτιστοποιημένες παραμέτρους ηλεκτροϊνοποίησης που επιλέχθηκαν, διασφαλίζοντας έτσι μια σταθερή διαδικασία ινοποίησης. Οι επιφανειακές ιδιότητες ενισχύθηκαν σημαντικά στα ικριώματα τριών στρωμάτων (tri-layered) και πυρήνα-κελύφους (core-shell) μέσω της προσθήκης της υδρόφιλης PVA, βελτιώνοντας ουσιαστικά την υδροφιλικότητα και τη βιοσυμβατότητα των ικριωμάτων. Αυτή η βελτίωση, είναι απαραίτητη για την προώθηση της ομαλής προσκόλλησης και πολλαπλασιασμού των κυττάρων. Τα αποτελέσματα από τις εικόνες της Ηλεκτρονικής Μικροσκοπίας Σάρωσης (SEM) και Διερχόμενης Δέσμης (ΤΕΜ) αποκάλυψαν ότι όλα τα ικριώματα εμφάνισαν λείες, ομοιόμορφες και ευθυγραμμισμένες ίνες, συμβάλλοντας σε πυκνές τρισδιάστατες δομές κατάλληλες για εφαρμογές μηχανικής ιστών. Τα ικριώματα επέδειξαν διαμέτρους ινών σε νανο- και μικροκλίμακα, με μεγέθη πόρων μικροκλίμακας, απαραίτητα για την κυτταρική διήθηση και την αναγέννηση των ιστών. Συγκεκριμένα, τα core-shell ικριώματα παρουσίασαν αρκετά ομοιόμορφη κατανομή στη διάμετρο των ινών, υποδηλώνοντας συνεπή δομή των ινών σε σύγκριση με τους άλλους τύπους ικριωμάτων. Οι μελέτες αποικοδόμησης in vitro που πραγματοποιήθηκαν σε διάστημα επτά μηνών τόνισαν τον ρόλο του υδρόφιλου συστατικού PVA στην ενίσχυση της πρόσληψης νερού. Τα tri-layered ικριώματα παρουσίασαν σταδιακή μείωση του πάχους λόγω της υποβάθμισης του εξωτερικού στρώματος PVA, ενώ τα core-shell ικριώματα διατήρησαν τη δομική τους ακεραιότητα περισσότερο, υποδηλώνοντας παρατεταμένη βιοαποικοδόμηση. Μηχανικά, όλοι οι τύποι ικριωμάτων εμφάνισαν αυξημένη ακαμψία (μετρούμενη από το μέτρο του Young και την τελική αντοχή εφελκυσμού) μετά την αποικοδόμηση, πιθανότατα επηρεασμένη από την παρουσία κρυσταλλικών προϊόντων υδρόλυσης. Ωστόσο, τα core-shell ικριώματα έδειξαν σημαντική αύξηση στο μέτρο του Young (40,7%) μετά από την επτάμηνη υποβάθμιση σε PBS, υποδεικνύοντας βελτιωμένη μηχανική σταθερότητα με την πάροδο του χρόνου, σε σύγκριση με τα single-layered ικριώματα που παρουσίασαν μέτρια αύξηση (18%) και τα tri-layered ικριώματα που διατήρησαν έναν συντελεστή μέτριας αύξησης (15,6%) καθ' όλη τη διάρκεια της in vitro αποδόμησης. Ακολούθως, τα αποτελέσματα της μέγιστης αντοχής σε εφελκυσμό (UTS) έδειξαν ότι τα core-shell και tri-layered ικριώματα εμφάνισαν υψηλότερες τιμές αντοχής (7,12 ± 0,36 MPa και 6,8 ± 0,36 MPa, αντίστοιχα) από τα single-layered ικριώματα (2,02 ± 0,18 MPa) και διατήρησαν ή βελτίωσαν την αντοχή τους κατά την περίοδο υποβάθμισης. Τέλος, από τα αποτελέσματα της μέγιστης καταπόνησης (maximum strain), τα single-layered ικριώματα ενώ αρχικά εμφάνισαν την υψηλότερη ολκιμότητα, παρουσίασαν ωστόσο τη μεγαλύτερη μείωση με την πάροδο του χρόνου στις συνθήκες υποβάθμισης. Από την άλλη, τα core-shell ικριώματα παρουσίασαν ένα πιο σταθερό προφίλ καταπόνησης, υποδηλώνοντας καλύτερη διατήρηση της ολκιμότητας σε σύγκριση με τα tri-layered ικριώματα. Συνόπτικά, τα core-shell ικριώματα επιδεικνύουν ενθαρρυντικές μηχανικές ιδιότητες, αφού φαίνεται να προσφέρουν έναν ισορροπημένο συνδυασμό μηχανικής αντοχής, σταθερότητας και ολκιμότητας, καθιστώντάς τα ιδανικά για εφαρμογές Ιστομηχανικής, όπου απαιτείται στιβαρή μηχανική υποστήριξη και ελεγχόμενη βιοδιάσπαση. Οι αξιολογήσεις από τις μελέτες κυτταροκαλλιέργειας αποκάλυψαν ανώτερη κυτταρική βιοσυμβατότητα για τα core-shell ικριώματα, προάγοντας καλύτερα την κυτταρική προσκόλληση και πολλαπλασιασμό σε σύγκριση με τα single-layered και tri-layered. Το υδρόφιλο στρώμα PVA όπως φαίνεται, έπαιξε κρίσιμο ρόλο στη βελτίωση της υδροφιλίας του ικριώματος και στην ενίσχυση των κυτταρικών αποκρίσεων. Συνολικά, η μελέτη καταδεικνύει την αποτελεσματικότητα της ομοαξονικής ηλεκτροϊνοποίησης για την παραγωγή σύνθετων ικριωμάτων με αυξημένη βιοσυμβατότητα, ελεγχόμενη βιοαποικοδόμηση και βελτιωμένες μηχανικές ιδιότητες. Τα αποτελέσματα κατέδειξαν ότι τα βιοαποικοδομήσιμα ινώδη ικριώματα δομής πυρήνα-κελύφους (core-shell) που αναπτύχθηκαν σε αυτή τη μελέτη, θα μπορούσαν να αποτελέσουν μια ελκυστική επιλογή για εφαρμογές Ιστομηχανικής και Αναγεννητικής Ιατρικής. Η μοναδική αρχιτεκτονική τους, επιτρέπει την ομαλή υποβάθμιση, αποφεύγοντας τα φαινόμενα αιφνίδιας ρήξης (burst effects), διατηρώντας καλές μακροπρόθεσμες μηχανικές ιδιότητες και υποστηρίζοντας την ανάπτυξη των κυττάρων, χωρίς κανένα εξωτερικό χημικό ή φυσικό ερέθισμα.
Open Access
Computational multiresolution wavelet-based methods for transient structural dynamic simulations with intrinsic localization and error estimation capabilities
(2024-12-12) Δημητρίου, Δημήτρης; Dimitriou, Dimitris
The present thesis focuses on the development of a multiresolution wavelet-based computational method for the efficient and multifunctional simulation of transient dynamic structural responses. The proposed method employes the Daubechies scaling and wavelet functions as basis functions for the approximation of field variables. The remarkable mathematical properties of the Daubechies wavelet family result in hierarchical multiresolution equations of motion with two types of solutions: the first solution type is named as “coarse” solution and utilizes the Daubechies scaling functions as basis, while the second type is called “fine” solution and uses the Daubechies wavelet functions as basis. The multiresolution discretized algebraic system starts with the calculation of the coarse solution, that provides an initial approximation, and then this solution can be further enriched by the summation of fine solutions until the desired level of accuracy is achieved. The orthogonality between the Daubechies scaling and wavelet functions leads to resolution-decoupled total mass matrices, namely, the coarse and fine solutions are totally uncoupled with respect to the mass matrix. So, the absence of mass coupling terms results in the independent calculation of the two resolution components (coarse, fine) in each time step, thus reducing the algebraic system size in half. Also, the orthogonality of the Daubechies scaling and wavelet functions yields diagonal mass matrices for both coarse and fine solutions. This is extremely important when it comes to transient dynamics because the utilization of explicit time integration schemes requires the inversion of the mass matrix, which is a rapid operation if the matrix is diagonal. Concerning the numerical formulation of the multiresolution method, termed as MR-FWD method, two new types of wavelet-based beam elements are developed; one following the first-order shear deformation theory, and the second encompassing a high-order laminate layerwise theory. These novel wavelet-based beam elements are tested against conventional finite elements, time domain spectral finite elements and other wavelet-based elements in numerical case studies of wave propagation in isotropic and composite beams and sandwich strips. In all cases, the proposed method manifests outstanding results considering the requirement of nodes for convergence, as well as the computing times. Moreover, the advanced capacity of the layerwise wavelet-based strip elements is demonstrated through the accurate prediction of fundamental (A0,S0) and higher (A1,S1) guided wave modes in thick sandwich composites. Furthermore, a new wavelet-based plate element is constructed based on the Reissner-Mindlin plate theory. The 2D wavelet-based plate elements reveal extraordinary convergence rates compared to time domain spectral finite elements and other wavelet-based elements in guided wave propagation analyses of isotropic and laminated plate structures. Besides the computational advantages of the method, it is shown that the fine solutions in all element types have additional inherent localization capabilities that emanate from the Daubechies wavelets. These localization properties make possible for the fine solutions to capture different structural responses based on the involved materials, geometric characteristics and wave direction (in 2D cases), among other characteristics. This notable behavior can lead to intrinsic error estimation through the direct comparison of multiple resolution components of the proposed method. After extensive study of the behavioral patterns of the various resolution components, it is deduced that the ratios of specific solutions can constitute explicit convergence indicators. In that way, the multiresolution procedure can stop at the intended level of accuracy without the need of convergence checks or any sort of comparisons with other results. Apart from that, the developed convergence indices can estimate the error locally, providing the basis for local solutions enrichment strategies. By this means, these indicators are leveraged for the creation of targeted and adaptive multiresolution refinement techniques. The targeted refinement works like a numerical microscope and increases the initial solution’s fidelity on an ad hoc basis. The automatic adaptive refinement utilizes the convergence indicators in order to enhance the solution’s accuracy only in the timesteps and nodes that it is needed. Both the targeted and adaptive refinement techniques exhibit very good accuracy and drastically decrease the solver durations, making the proposed method even faster. Finally, the intrinsic localization capacity of the method is further leveraged for the development of inverse methodologies for wave-based parameter estimation. The application of the proposed methodologies requires full-field data of the wave propagation phenomenon. The data are decomposed using wavelet decomposition, resulting in approximation and detail datasets, that are directly comparable to the coarse and the fine solutions of the MR-FWD method, respectively. Thus, the proposed inverse methodologies take advantage of the high sensitivity and localization properties of the fine solutions in order to create a model update procedure that compares the fine solutions of the MR-FWD model to the detail data of the experimental measurements for the estimation of the design variables. Numerical case studies of wave-based damage detection in composite strips and 2D solids are performed, manifesting the proposed methods’ superior accuracy compared to traditional methodologies.
Open Access
A study on methods to enhance autonomy for multi-UAV civilian applications
(2024-11-26) Πατρινοπούλου, Νίκη; Patrinopoulou, Niki
Unmanned Aerial Vehicles (UAVs) have experienced significant advancement since their inception, impacting various fields such as military operations, security, disaster management, package delivery, and recreational activities. This progress has spurred substantial research efforts aimed at furthering UAV capabilities and technologies. Presently, UAVs and swarms of UAVs are being considered for numerous civilian applications, including agriculture, photography, shipping, disaster management, rescue operations, archaeological surveys, mapping, healthcare, wildlife observation, weather forecasting, telecommunications, and border surveillance. Despite the recent interest in autonomous UAV research, the majority of UAVs still rely on human pilots, resulting in training demands, increased operational costs, and susceptibility to human error, necessitating the development of autonomous systems. Multi-UAV systems offer enhanced performance but require improved autonomy to enable effective collaboration and decision-making without compromising safety. Enhancing UAV autonomy entails advancements in decision-making, planning, guidance, and coordination methods, utilizing technologies such as Artificial Intelligence, control theory, geometric approaches, rule-based methods, and bio-inspired optimization techniques. This dissertation proposes and investigates methodologies to provide increased performance on multi-UAV systems for two separate applications, security surveillance, and fire monitoring, via autonomous decision-making, guidance, and coordination of the UAV agents. It introduces a UAV swarm designed for area surveillance and intruder detection. Its primary goal is to devise a swarm intelligence-based decision-making process that enhances the system's performance in searching for unknown intruders. The system incorporates functionalities such as target tracking, collision avoidance, and scanning algorithms. It focuses on a multi-UAV system of quadcopters specifically engineered for monitoring open areas and identifying intruders, with a key emphasis on scalability and data-aware decision-making. The main aspect of the study revolves around developing and evaluating the effectiveness of the swarm intelligence-based decision-making process. This decision-making process aims to determine the next task assigned to each UAV agent within the swarm, with the overarching objective of optimizing area coverage and intruder detection. Consequently, the system's objective is framed as a task allocation problem, with other essential components necessary for system completion also provided. The algorithms are implemented and validated through Software In The Loop (SITL) simulations. Comparative analysis is conducted against two alternative surveying swarms and a static camera-based approach. The integration of real-time decision-making confers a distinct advantage to the swarm over similar systems, as evidenced by the simulation results. Afterwards a decentralized UAV system for monitoring active wildfires is presented. Wildfires have proven to be a significantly exigent issue over the past decades. An increasing amount of research effort investigates the use of UAVs and multi-UAV systems for wildfire monitoring. This work focuses on the development of a decentralized framework for the purpose of monitoring active wildfires and their surrounding areas with fixed wing UAVs. It proposes a distributed fire data update methodology, a new formation algorithm based on virtual forces, fine-tuned by a Genetic Algorithm (GA), to arrange virtual agents into the monitoring area, and a control strategy to safely and efficiently guide fixed wing UAVs to loiter over the structured virtual agents. The system is tested in SITL simulation with up to eight UAVs. The simulation results demonstrate the effectiveness of the system on monitoring the fire and providing updated situational awareness data. Finally, this dissertation studies traffic capacity balancing methods for an unmanned traffic management system. The commercialization and plethora of available applications for UAVs results in the need to investigate the integration of numerous UAVs into airspace. It is foreseen that the number of UAVs in urban and rural airspace will significantly increase. Hence, their safe and efficient operation must be ensured, while specialized UAV traffic control systems must be developed to manage increased UAV traffic. This work contributes to investigating strategies for balancing traffic capacity within UAS Traffic Management (UTM) systems operating in densely populated, low-altitude urban airspace. It examines the advantages of integrating traffic capacity balancing and outlines three distinct design methodologies. The proposed systems undergo testing through simulations of mixed urban airspace (comprising both constrained and open areas) modeled after Vienna, at five different traffic densities. The objective is to identify and propose a solution for dynamic capacity management within a hybrid UTM system. A central flow control module is developed to facilitate capacity balancing by pinpointing areas of high traffic density and notifying UAVs accordingly. This empowers UAVs to autonomously adjust their routes to avoid congested zones. The aim of implementing the flow control module is to promote a more even distribution of traffic, thereby enhancing flight safety and efficiency. Flow control operates within predefined regions termed flow groups, where traffic density is monitored.
Open Access
Extracting knowledge from polymorphic textual data using graph-based machine learning and transformer-based deep learning techniques
(2024-11-28) Γιαρέλης, Νικόλαος; Giarelis, Nikolaos
The aim of this thesis is to develop novel approaches that extract meaningful knowledge from polymorphic (unstructured and structured) textual data. To achieve this, we rely on recent advancements from the areas of Machine Learning (ML), Graph-Based Learning, Natural Language Processing (NLP) and Deep Learning (DL). This thesis comprises a series of research contributions; specifically: (i) we introduce a novel graph-based ML framework (Graph-of-Docs), which models documents in a unified graph structure, while solving several contextual limitations of traditional machine and previous graph-based learning frameworks; (ii) we study the incorporation of Graph Theory, Representation Learning and Knowledge Graphs (KGs) in the proposed framework to improve various classification tasks, while applying this framework for several real-world applications; (iii) we study the state of the art in several NLP tasks, focusing on the benefits of utilizing the Transformer architecture in them; (iv) we propose a novel keyphrase extraction approach that is highly accurate and has low computational complexity; (v) we introduce two novel Greek NLP datasets, and a series of state-of-the-art DL models aiming to facilitate future research in the understudied field of Greek NLP; (vi) we propose a unified framework, which combines advancements from KGs and Graph-based concepts with transformer- based Large Language Models (LLMs), as to improve the performance of various text generation tasks, while reducing the limitations of LLMs including hallucinations and indecisiveness. The experiments reported in this thesis validate the effectiveness of the abovementioned research findings. The code of our models and experiments is publicly available on GitHub, while our Greek NLP datasets and DL models are hosted on HuggingFace.
Open Access
Interlaminar fracture toughness of layered materials under fatigue loading conditions
(2024-09-02) Λουκάς, Αδάμος; Lucas, Adamos
In the present thesis, the mode I and II fatigue fracture performance of a novel group of Titanium/CFRP adhesive joints is investigated. The framework is set by the need to characterize the fatigue behavior of this new group of Ti/CFRP adhesive joint. The adhesive joints under study are composed of two thin adherents, one Titanium sheet and one CFRP laminate, and are reinforced from both sides with two thick aluminum beams to increase their flexural stiffness and ensure the non-yielding of the Titanium. The interfacial fracture resistance is in the focus and aspects such as the loading that defines the crack propagation threshold as well as the utilization of an appropriate Paris-like law to optimally describe the experimental data are under investigation. The challenges and issues that arise in the fatigue crack growth characterization of such dissimilar joints due to multiple crack formation, suitable compliance and strain energy release rate definitions are thoroughly investigated in the present thesis. A new method for measuring the experimental compliance is proposed and critically compared to the dynamic compliance concept. The compliance calibration method while taking into account secondary cracking is addressed. The compliance calibration method is an experimental procedure with an aim to account for an effective crack length. Analytical solutions for compliance are formulated with the same objective, seeking to reduce experimental procedures. The fatigue crack growth curves are extracted and discussed under the prism of the aforementioned challenges. Characterizing fatigue crack growth in dissimilar adhesive joints within the Paris law framework posed challenges, particularly due to the increasing complexity of such joints and uncertainty surrounding the parameters influencing the graph. To address this challenge and draw definitive conclusions about the tested materials, uni-directional CFRP adhesive joints and composite laminates were fabricated and subjected to a series of tests, including quasi-static and high-speed loading. The present study identifies a gap in existing literature regarding fatigue fracture, particularly in relation to the loading rate effect observed in quasi-static and dynamic testing. This research aims to quantify the loading rate in fatigue experiments. Based on the assumption that Paris' equation is suitable for applying similitude, an experimental campaign is undertaken. During this campaign, we equalize the loading rates at the crack tips across experiments employing various combinations of frequency and stress ratios. This endeavour aims to tackle the commonly encountered R-ratio effect in fatigue fracture tests. Until the present day, the scientific community relied exclusively on the Paris-Erdogan empirical law proposed in the early 1960s for metals, to predict the behaviour of delaminated composites under fatigue loading. Paris’ law though uses fitting of datapoints and normalization of parameters. The physics governing interlaminar crack propagation in layered materials (e.g., composites) are still elusive. Interlaminar crack propagation in layered materials, such as composites, is still a not fully understood phenomenon in fracture mechanics. Experimental observations reveal a broad spectrum of crack propagation velocities during the interlaminar fracture of layered materials under varying loading rates. While there have been numerous studies examining interlaminar crack propagation under quasi-static, fatigue, and more recently, high-speed loading, there has been a notable lack of endeavours to present and interpret crack growth data within a unified framework. To this direction, we present a new interpretation of interlaminar crack propagation data acquired from a broad test campaign under various mode I loading conditions (quasi-static, fatigue and high-speed). We connect the crack tip velocity to the rate of change of the SERR to propose a new equation/model that shows potential in presenting and explaining the variety of experimental crack growth data. This correlation arises from a simple mathematical analysis of the rate of change of the SERR using the chain differentiation rule. Subsequently, it is further substantiated by experimental evidence, in an effort to accommodate results from various experiments under a unified description. A robust correlation of the test data is established through a range of slopes which are proven -experimentally- to be characteristic constants of the mode I interlaminar fracture of a specific material system.
Open Access
Development of a laser shock based disassembly process for adhesively bonded composite aeronautical parts
(2024-09-17) Κόρμπος, Παναγιώτης; Kormpos, Panagiotis
Novel aircraft engine fan blades utilize a multi-material structure composed of a 3D woven composite core partially covered by an adhesively bonded thin metallic layer. The end-of-life (EoL) management of such a configuration involves many challenges, with one of them being the disassembly of the bonded components. Numerous methodologies have been explored to address the debonding challenge, encompassing various physical and chemical phenomena. However, many of these methods have been proven less suitable in scenarios involving composite materials, especially when the preservation of the integrity of the materials is essential. Recent developments on the laser shock technique have revealed the potential of the method to be used for the disassembly of adhesively bonded CFRP and metallic parts. The laser shock technique involves the interaction between a high-power laser beam and a target material, initiating a plasma expansion process. The recoil momentum of the ablated material generates a high pulsed pressure, creating a shock wave that propagates within the material. This shock wave propagation induces localized high tensile stress, which can be used to effectively debond a bonded structure. Various laser configurations can be utilized to focus said tensile stress area to a location of interest, thus creating a versatile process. In this thesis, the laser shock technique is investigated as a disassembly method for adhesively bonded multi-material aeronautical structures. The process is specifically applied to a Foreign Object Damage (FOD) panel, which consists of a 3D woven composite core bonded by an adhesive film to a thin titanium protective layer. The FOD panel serves as a representative structure of a chord of a multi-material engine fan blade, acting as a demonstrator platform for the European research program "MORPHO" (H2020, 101006854), which provides much of the framework for this thesis. The study and utilization of the complex phenomena involved in the laser shock technique necessitate meticulous numerical simulation, which was conducted using the LS-Dyna explicit finite element code. The development of the numerical model can be split into two sections. Initially, a coupon-scale model of the multi-material assembly was designed, so that it can be validated utilizing experimental data. Subsequently, a strip of the FOD panel was modeled, in an attempt to upscale the process to a real component. Given the complexity of the composite substrate, further investigation into simulation techniques was deemed necessary. The composite in question is a 3D woven CFRP structure, prompting the initial design of a detailed finite element model using the TexGEN software. Given the intricacy of the structure, the software’s scripting capabilities were utilized for model design. Fiber bundles were simulated using an advanced composite material model capable of providing progressive damage indicators and simulating high strain rate phenomena described in the LS-Dyna manual as MAT_162. Material properties were obtained through a homogenization process employing the Chamis micro-mechanical formulations. The matrix of the 3D woven composite was simulated using an elastoplastic material model, with literature properties adjusted for the high strain rate of the laser shock process. The detailed model was subsequently compared to a homogenized one, simulated solely by the progressive damage composite model (MAT_162). The comparison of shockwave propagation revealed minimal difference between the two models. However, when comparing damage prediction, differences between the two simulation techniques became evident, with the detailed model proving more accurate in predicting small matrix cracking. Despite the slightly increased sensitivity to damage demonstrated by the detailed model, the homogenized model was deemed more suitable for simulating the process due to its enhanced efficiency. Material properties for the homogenized model were obtained by high strain rate characterization of the composite substrate. Specifically, coupons that were cut from a single sheet of the composite were tested using the split Hopkinson pressure bars (SHPB) apparatus. Both tension and punch shear tests in the weft and warp orientation were conducted. Notably, the increased strength of the material in the warp orientation exceeded the capabilities of the tension testing apparatus, thus providing only elastic material properties. The weft orientation specimens, as well as the punch shear experiments, were conducted successfully and were able to provide the required data. Additionally, laser shock experiments were conducted using a multi-material coupon that incorporates a 3D woven composite substrate bonded to a titanium layer. Experiments utilized two distinct configurations. Initially, a single-sided, single-pulse configuration was used, while a laser interferometer (VISAR) was used on the opposite face of the specimen, measuring the particle velocity. The process failed to generate the desired stress field for the debonding of the titanium layer but provided valuable back face velocity measurements. Subsequently, a symmetric configuration was employed, where two laser beams are directed through optics on each side of the specimen, with a time delay between them. This allows for a more precise stress field generation and was able to debond the titanium layer. After the extensive experimental campaign, the coupon was modeled and the laser shock process was simulated using the LS-Dyna explicit code. The model, like the coupon, includes the composite substrate that was modeled using MAT_162, bonded to a titanium protective layer that is modeled using the Johnson-Cook plasticity model in combination with the Grüneissen equation of state. Material properties derived from the literature were used for the titanium component, as it is a very well-studied material for high strain rate applications. The bondline between the two components utilizes the cohesive zone method (CZM). Experimental back face velocity measurements were used to validate the numerical model in terms of shock wave propagation, while the properties for the CZM were obtained by comparing the predicted damaged area of the model to microscope photographs of the symmetric experiments that managed to debond the assembly. The laser shock disassembly process necessitates multiple consecutive shots. The simulation of this process is carried out by developing an automated Python script for sequential simulations. Initially, a preliminary study compared this approach to a traditional methodology that uses a single simulation for a small number of consecutive shots. The results were identical, and the multi-step process was deemed more practical. To implement this, a traditional LS-Dyna simulation is conducted, and the results are used as input for the subsequent simulation. All necessary loadings were described parametrically to enable external control via a Python script. This script uses an LS-Dyna keyword manipulation library to adjust the parameters of each consecutive analysis. Loading inputs can be provided in various ways depending on the problem’s complexity, with the most versatile method being the input of a CSV file containing shot locations and other useful information, such as the delay for a symmetric shot. After establishing the methodology with a validated numerical model, it is possible to upscale the process for the disassembly of a strip of the FOD panel. To achieve this, a preliminary simulation examined the effect of surface inclination, concluding that the process is only affected at extreme inclinations, which is not the case for the FOD panel. The model was derived from a CAD file provided by partners of the "MORPHO" project. Material models were identical to those in the coupon-scale model and are thus considered validated. The process employs a single-sided double-beam configuration with a time delay, where two laser beams irradiate the same spot on the component with a time delay between them. This configuration allows control of the stress field, similar to the symmetric configuration, with the added advantage of only targeting one surface of the component. The simulation procedure comprises two stages: the process definition and the disassembly process simulation. Initially, process definition simulations were conducted to determine the delay times necessary for debonding the top and bottom interfaces of the panel. These simulations provided a delay time for debonding the top interface between the Ti/CFRP and a second-order polynomial equation to calculate the delay time for debonding the bottom interface of the Ti/CFRP assembly based on the laser spot location. The second stage involves process simulation for the full disassembly of the strip of the FOD panel. This simulation uses the automated Python script to conduct 29 consecutive simulations at 14 predefined spot locations, resulting in the complete separation of the titanium edge. The damage state of the CFRP was inspected via the history variables provided by MAT_162. As expected, the composite developed matrix cracking in different areas but notably no fiber damage. This conclusion is useful for EoL assessment, as the component can be recycled, providing high-value recycled carbon fibers.