Τμήμα Ηλεκτρολ. Μηχαν. και Τεχνολ. Υπολογ. (ΔΔ)

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 5 of 370
  • Thumbnail Image
    Item
    Open Access
    Development of semi-supervised machine learning algorithms and applications
    Φαζάκης, Νικόλαος; Fazakis, Nikos
    The well-established approach of Supervised learning is a branch of the broader science of artificial intelligence. The aim of this learning philosophy is the development of computer programs to automatically improve their experience through the extraction of useful information from annotated examples. The methodology of this learning approach is extremely useful in real world applications where large collections of data are available related to problems where absolute associations of the input data and the outcomes cannot be discovered or approximated by explicit mathematic formulations. Such scientific fields include observed data of text, audio or image formats. The classic methodology of supervised learning comes with the cost of annotating, usually referred as ‘labeling’ process, the available data instances of a dataset often by human experts in a field. Considering that modern big datasets can have terabytes of data; it is a very inefficient procedure for humans to tackle. This intrinsic bottleneck is addressed by Semi-supervised learning (SSL), which allows the model to incorporate part or all of the available unlabeled data into its supervised learning. The goal of SSL is to maximize a model's learning performance while reducing the amount of labor required by using such newly labeled instances. This thesis is oriented in the improvement of a sub-category of SSL algorithms referred as self-labeled techniques, and the application of them in real world problems. Numerous important questions are answered such as: Which learning algorithms can best utilize the self-labeling schemes? Can the introduction of ensemble learning along with semi-supervised learning provide classification improvements in real world problems such as speaker identification or educational grade prediction? Is it possible to define a new multi-regressor learning scheme based on self-labeling that can rival the existing semi-supervised regression algorithms? Can iterative data imputation be improved through the introduction of self-training? In health-related datasets is it possible to take advantage of unlabeled test sets to balance the shortage of examples through semi-supervised transductive learning?
  • Thumbnail Image
    Item
    Open Access
    Control and nonlinear stability analysis of the electromechanical system of electric vehicles
    Μακρυγιώργου, Δήμητρα; Makrygiorgou, Dimitra
    The present PhD dissertation is addressed in the research field of control and analysis of motion systems in vehicular industry and more precisely in providing a systematic procedure for the control design applied on electrified vehicles with respect to the system stability. In particular, taking into account the nonlinear representations of all components involved in an electrified vehicle, namely the energy storage, both electric and internal combustion motors, and power converters, an accurate model is developed that sets the basis for control design and analysis purposes. System performance is examined in a first place via the proposal of a novel open-loop stability analysis. The latter is conducted by applying Lyapunov based nonlinear techniques that actually prove system convergence to its desired equilibrium for any feasible bounded control input signals. Those theoretical results set the cornerstone towards adopting intelligent control techniques, fuzzy and neurofuzzy, due to the bounded control signals they produce and their efficiency in copying with system nonlinearities and uncertainties. In a similar manner, a combination of sliding mode and field oriented techniques are applied and provide an improved system dynamic behavior endowing it with characteristics of robustness with the system stability again to be proven based on the novel open-loop analysis. Towards extending the aforementioned effort, a general systematic procedure for control design based on simple cascaded PI controllers that can ensure stability for a wide category of Euler-Lagrange systems is developed. In this framework, asymptotic stability is proven for the electrified vehicle system by keeping in mind its desired equilibrium and applying proper cascaded PI controllers with their gains to be tuned based on the adopted systematic analysis. In addition, a vehicle to grid system is also considered and analysed based on the proposed method and its accurate dynamic representation. Specifically, in a first stage the mathematical representation of a vehicle to grid system combined with multiple connected vehicles is deployed and its stability is proven by applying trivial PI controllers and adopting the proposed open-loop analysis. Afterwards, a unified vehicle to grid - plug-in electric vehicle system is developed and controlled based on the proposed systematic passivity based procedure that ensures global asymptotic stability of the entire system around its desired equilibrium. In all cases, the extracted theoretical results are verified by conducting extensive simulations that indicate a satisfactory system response with smooth transients and a successful driving of the entire system to its desired equilibrium.
  • Thumbnail Image
    Item
    Open Access
    Methods and tools to support instructional quality in online discussions in massive open online courses
    Ντούρμας, Αναστάσιος Ιωάννης; Ntourmas, Anastasios Ioannis
    The thesis deals with ways to improve and maintain the instructional design quality in Massive Open Online Courses (MOOCs). More concretely, our goal is to support the decision making of MOOC practitioners (e.g., instructional designers, instructors) at run-time, i.e., when the course is active, on instructional quality issues. The main gap that we address is that existing evaluation studies of instructional quality do not assess the supporting strategies adopted by the instructional staff (i.e., learner facilitators) to support learners in the forum. To address this gap we follow a three-phase research methodology. In the first phase we perform three case studies that extend the existing empirical research on the intervention characteristics and the problems of learner facilitators in courses of different subject matter. In the second phase, we perform an in-depth case study in which we provide empirical evidence about the main instructional approaches learner facilitators adopt in MOOC forums, and unravel the quality issues that are associated with them. Then, we proceed to the formulation of a data-driven methodology that aims to guide MOOC system-tool developers in designing tools that help instructors keep track of such instructional approaches, and support their decision-making at run-time. In the last phase, we use the data-driven methodology to design and develop a runtime tool that supports the decision making of instructors in MOOCs. Specifically, we provide a proof of concept regarding the feasibility and usefulness of the proposed methodology in designing new software applications. This contribution has been evaluated throughout the research process of three case studies performed in MOOC contexts, and a set of semi-structured interviews with experts.
  • Thumbnail Image
    Item
    Open Access
    Patient-specific modelling, simulation and real time processing for respiratory diseases
    Νούσιας, Σταύρος; Nousias, Stavros
    Asthma is a common chronic disease of the respiratory system causing significant disability and societal burden. It affects over 334 million people worldwide and generates costs exceeding $USD 56 billion in 2011 in the United States. Managing Asthma involves controlling symptoms, preventing exacerbations, and maintaining lung function. Improving asthma control affects patients' daily life and is associated with a reduced risk of exacerbations and lung function impairment, reducing the cost of asthma care and indirect costs associated with reduced productivity. Understanding the complex dynamics of the pulmonary system and the lung's response to disease, injury, and treatment is fundamental to the advancement of Asthma treatment. Computational models of the respiratory system seek to provide a theoretical framework to understand the interaction between structure and function. Their application can improve pulmonary medicine by a patient-specific approach to medicinal methodologies optimizing the delivery given the personalized geometry and personalized ventilation patterns while introducing a patient-specific technique that maximizes drug delivery. A three-fold objective addressed within this dissertation becomes prominent at this point. The first part refers to the comprehension of pulmonary pathophysiology and the mechanics of Asthma and subsequently of constrictive pulmonary conditions in general. The second part refers to designing and implementing tools that facilitate personalized medicine to improve delivery and effectiveness. Finally, the third part refers to the self-management of the condition, meaning that medical personnel and patients have access to tools and methods that allow the first party to easily track the course of the condition and the second party, i.e. the patient to easily self-manage it alleviating the significant burden from the health system.
  • Thumbnail Image
    Item
    Open Access
    Διάγνωση σφαλμάτων ηλεκτρικών κινητηρίων συστημάτων οδηγούμενων από ηλεκτρονικούς μετατροπείς ισχύος
    Σπυρόπουλος, Διονύσιος; Spyropoulos, Dionysios
    Αντικείμενο της παρούσας διατριβής αποτελεί η ανάπτυξη καινοτόμων μεθοδολογιών και αλγορίθμων για την ανίχνευση και τη διάγνωση σφαλμάτων που εμφανίζονται σε ηλεκτρικά κινητήρια συστήματα αποτελούμενα από τριφασικούς ασύγχρονους κινητήρες οδηγούμενους από ηλεκτρονικό μετατροπέα ισχύος με κριτήριο την όσο το δυνατόν πιο αξιόπιστη διάγνωση αυτών. Πιο συγκεκριμένα, ξεκινώντας από την ανάπτυξη νέων μεθοδολογιών για την ανίχνευση ασυμμετριών στο στάτη, η έρευνα επικεντρώθηκε στη συνέχεια στη διάγνωση του σφάλματος της σπασμένης μπάρας του δρομέα του κινητήρα. Προς την κατεύθυνση αυτή, βασικό διαγνωστικό μέσο αποτέλεσε η μελέτη του αρμονικού περιεχομένου του ρεύματος του στάτη του κινητήρα. Για την αξιόπιστη διάγνωση του σφάλματος προτάθηκε η χρήση δύο καινοτόμων μεθοδολογιών που βασίζονται, η, μεν, πρώτη στον αλγόριθμο Goertzel και η, δε, δεύτερη στο μετασχηματισμό Synchrosqueezing Wavelet. Οι δύο αυτές τεχνικές επεξεργασίας σήματος ενώ εμφανίζουν ευρεία χρήση σε άλλα επιστημονικά πεδία (π.χ. επικοινωνίες, σεισμολογία, βιοϊατρική) είτε δεν έχουν χρησιμοποιηθεί καθόλου είτε έχουν διερευνηθεί σε πολύ μικρό βαθμό στον τομέα της διάγνωσης σφαλμάτων ηλεκτρικών μηχανών. Η αποτελεσματικότητα των προτεινόμενων μεθοδολογιών αξιολογείται σε κάθε περίπτωση αρχικά όταν οι υπό μελέτη κινητήρες τροφοδοτούνται από το δίκτυο εναλλασσόμενης τάσης και στη συνέχεια κατά την οδήγησή τους από τον ηλεκτρονικό μετατροπέα ισχύος. Για το σκοπό αυτό σχεδιάστηκε και κατασκευάστηκε στο Εργαστήριο μια πρωτότυπη πειραματική διάταξη. Προκύπτει πως οι προτεινόμενες μεθοδολογίες μπορούν να προσφέρουν σημαντικά πλεονεκτήματα συγκριτικά με άλλες κλασσικές τεχνικές, αποτελώντας σημαντικά εργαλεία στην κατεύθυνση της έγκαιρης και αποτελεσματικής διάγνωσης του σφάλματος. Επιπλέον, προτείνεται μία νέα μεθοδολογία για τη διάγνωση του σφάλματος της σπασμένης μπάρας του δρομέα, με παρακολούθηση του σήματος της μηχανικής ροπής σε συχνότητες πλησίον της 6ης αρμονικής του σήματος. Εξετάζεται κατά πόσον η μέθοδος μπορεί να οδηγήσει στην αξιόπιστη διάγνωση του σφάλματος με την ανάλυση του σήματος στο πεδίο συχνότητας – χρόνου τόσο κατά την εκκίνηση όσο και στη μόνιμη κατάσταση λειτουργίας του κινητήρα. Η εκτεταμένη πειραματική διερεύνηση οδήγησε στο συμπέρασμα πως η συγκεκριμένη τεχνική μπορεί να χρησιμοποιηθεί για την αξιόπιστη διάγνωση του σφάλματος όταν ο κινητήρας τροφοδοτείται από το δίκτυο εναλλασσόμενης τάσης τόσο κατά την εκκίνηση όσο και κατά τη μόνιμη κατάσταση λειτουργίας του.