41 research outputs found
Αεροδυναμικός Σχεδιασμός Ανάλυση & Βελτιστοποίηση Μορφής Με Τη Συζυγή Μέθοδο Αγωνιστικού Πατώματος Αυτοκινήτου Σε Περιβάλλον OpenFOAM
No full textΥλοποίηση εξατομικευμένου συστήματος συστάσεων βασιζόμενο σε αρνητικά δείγματα με στόχο τη ταχεία προσαρμογή σε πραγματικό χρόνο
No full textΕθνικό Μετσόβιο Πολυτεχνείο--Μεταπτυχιακή Εργασία. Διεπιστημονικό-Διατμηματικό Πρόγραμμα Μεταπτυχιακών Σπουδών (Δ.Π.Μ.Σ.) "Επιστήμη Δεδομένων και Μηχανική Μάθηση
Optimising Data Analysis Through Offline Large Language Models and Scalable Data Management Techniques
No full textΒελτιστοποίηση ανάλυσης δεδομένων μέσω τοπικών μεγάλων γλωσσικών μοντέλων και τεχνικών κλιμακωτής διαχείρισης δεδομένων
No full textThe importance of data across various sectors demands innovative approaches to data management and analytics. This PhD thesis investigates the integration of offline large language models (LLMs) for automated code generation, aiming to streamline data analysis processes, and thus enhance the scalability and efficiency of data management systems. By leveraging offline LLMs, the proposed approach empowers users to perform data analyses without extensive programming skills, thereby democratizing data analytics. The research delves into the architecture and implementation of scalable data management systems that can efficiently handle datasets of several volumes. Based on an efficient data management platform, the capabilities of offline LLMs to generate analytical code are examined, showcasing how these models can transform user queries into executable scripts that facilitate data manipulation and interpretation. Through experiments and case studies, the practical applications and benefits of the proposed study are showcased. The results highlight the potential of offline Large Language Models in Data Science and Analysis. This thesis contributes to the field by presenting a study that integrates AI-driven code generation with robust data management practices, ultimately paving the way for more efficient and user-friendly data analytics solutions.Η σημασία που έχουν αποκτήσει τα δεδομένα σε ποικίλους τομείς, καθιστά αναγκαία την ανάπτυξη καινοτόμων προσεγγίσεων στη διαχείριση και ανάλυσή τους. Η παρούσα διδακτορική διατριβή ερευνά την ενσωμάτωση τοπικών Μεγάλων Γλωσσικών Μοντέλων (Offline LLMs) για την αυτόματη παραγωγή κώδικα, με στόχο την απλοποίηση των διαδικασιών ανάλυσης δεδομένων και, κατ’ επέκταση, τη βελτίωση της επεκτασιμότητας και αποδοτικότητας των συστημάτων διαχείρισης δεδομένων. Μέσω της αξιοποίησης τοπικά εκτελούμενων LLMs, η προτεινόμενη προσέγγιση δίνει τη δυνατότητα σε χρήστες χωρίς εκτεταμένες προγραμματιστικές γνώσεις να πραγματοποιούν αναλύσεις δεδομένων, συμβάλλοντας έτσι στη δημοκρατικοποίηση της επιστήμης των δεδομένων. Η έρευνα εστιάζει στην αρχιτεκτονική και την υλοποίηση επεκτάσιμων συστημάτων διαχείρισης δεδομένων, ικανών να διαχειρίζονται αποδοτικά σύνολα δεδομένων μεγάλου όγκου. Βασιζόμενη σε μια αποδοτική πλατφόρμα διαχείρισης δεδομένων, η εργασία εξετάζει τις δυνατότητες των offline LLMs για την παραγωγή αναλυτικού κώδικα, αναδεικνύοντας πώς αυτά τα μοντέλα μπορούν να μετασχηματίσουν φυσικές ερωτήσεις χρηστών σε εκτελέσιμα σενάρια, τα οποία διευκολύνουν τον χειρισμό και την ερμηνεία των δεδομένων. Μέσα από πειραματικές διαδικασίες και μελέτες περίπτωσης, παρουσιάζονται οι πρακτικές εφαρμογές και τα οφέλη της προτεινόμενης προσέγγισης. Τα αποτελέσματα αναδεικνύουν την αποδοτικότητα των τοπικά εκτελούμενων Μεγάλων Γλωσσικών Μοντέλων στην επιστήμη και ανάλυση δεδομένων. Η διατριβή συμβάλλει στο πεδίο παρουσιάζοντας μια μελέτη που συνδυάζει τεχνικές αυτόματης παραγωγής κώδικα μέσω τεχνητής νοημοσύνης με στιβαρές πρακτικές διαχείρισης δεδομένων, ανοίγοντας τον δρόμο για πιο αποδοτικές και φιλικές προς τον χρήστη λύσεις ανάλυσης δεδομένων
Effects of Monograde and Multigrade Oils on the Friction Force in Four-Stroke Motor Engine: An Experimental and Analytical Approach
No full textMeasurement of friction and noise from piston assembly of a single-cylinder motorbike engine at realistic speeds
No full text This paper presents a thorough experimental study of piston assembly friction and noise in a single-cylinder motorbike engine operating at low speeds. The friction of the piston ring pack is evaluated using a foil strain gauge with minimal cylinder modification on the thrust side. The technique involves transmitting deformations through the cylinder bore and recording reflections from the lubricated interface as the piston assembly passes. Under these conditions, the piston side forces and the thermal deformations on the output side of the strain gauge sensor are critical. Therefore, the proposed methodology is designed under controlled operating conditions. The overall deformation of the piston assembly is analysed to measure the primary reflection due to friction between the piston assembly and the cylinder wall. Simultaneously, the piston assembly noise is recorded on the thrust side of the engine block using a microphone. Taking measured noise data into account, possible piston slap events resulting from varied engine speeds are taken into account using continuous wavelet signal analysis. The calibration procedure for both tests is also illustrated. The measured friction results show that the strain gauge technique is a challenging work in providing realistic results to enhance current technology. For low engine speeds, a higher contribution is noted by boundary friction at the top dead centre reversal, extending to the position of maximum combustion pressure in the power stroke. Furthermore, the main contribution of the piston slap is estimated at the thrust side when the piston assembly passes at the beginning of the combustion stroke. These results can also be attributed as data to validate piston ring models in terms of friction and piston slap. </jats:p
Computational fluid dynamics analysis of top compression ring in mixed lubrication
No full textThe top compression ring design of an internal combustion engine has an impact on ring in-plane motion and its lubrication conditions at the ring-cylinder liner contact. In this paper, the geometrical dimensions of the top compression piston ring-cylinder system were obtained from an actual four-stroke motorbike engine. The top ring tribological behaviour was characterized by a Computational Fluid Dynamics (CFD) simulation including the effects of asperity contact. Based on the numerical solution of the Navier–Stokes equations and taking into account realistic engine running conditions, the effect of the in-plane top ring motion in quasi-static equilibrium was determined. The simulation model was validated by the numerical and experimental results of similar investigations of other researchers. Good predictions were achieved by solving the Navier-Stokes equations because the pressure gradient into the lubricant film was accounted for. The effects of ring curvature at the ends of the stroke were studied. The results show that a flatter ring profile has a sufficient minimum lubricant thickness at reversal points, showing reasonably lower boundary friction than that of the higher curvature. Higher heights of the curvature profile promote significantly mixed lubrication, in which the power losses and the burning of excess lubricating oil are increased. The proposed simulation model can be expanded to any set of compression rings where a minimum simulation time is required
Lubrication Performance of Engine Commercial Oils with Different Performance Levels: The Effect of Engine Synthetic Oil Aging on Piston Ring Tribology under Real Engine Conditions
No full textTo further improve efficiency in automotive engine systems, it is important to understand the generation of friction in its components. Accurate simulation and modeling of friction in machine components is, amongst other things, dependent on realistic lubricant rheology and lubricant properties, where especially the latter may change as the machine ages. Some results of research under laboratory conditions on the aging of engine commercial oils with different performance levels (mineral SAE 30, synthetic SAE10W-40, and bio-based) are presented in this paper. The key role of the action of pressure and temperature in engine oils’ aging is described. The paper includes the results of experiments over time in laboratory testing of a single cylinder motorbike. The aging of engine oil causes changes to its dynamic viscosity value. The aim of this work is to evaluate changes due to temperature and pressure in viscosity of engine oil over its lifetime and to perform uncertainty analysis of the measured values. The results are presented as the characteristics of viscosity and time in various temperatures and the shear rates/pressures. This paper also includes a Computational Fluid Dynamics (CFD) model, applying the experimental results in the piston ring tribology problem
Dynamic request assignment in aerial ride sharing operations
No full textThe concept of Urban Air Mobility (UAM) services was created mainly in response to traffic congestions. In this research we focus on UAM services such as those provided by Uber Elevate. We therefore present a framework to solve the Urban Air Mobility Problem with Time Windows (UAMP-TW) under dynamic demand, using an Adaptive Large Neighborhood Search (ALNS) algorithm. The objective of this study is to maximize the operational profit and consider customer satisfaction. Satisfaction is measured by two factors: (1) deviation from desired departure time to actual departure time and (2) deviation from nominal trip duration to actual trip duration. In our analysis we aim to determine a relationship between customers and their contribution towards profit. We address this by running simulation instances that cover three operational scenarios: a morning and evening commuter transportation case (scenarios 1 and 2) and the an occurrence of an event at a specific location (scenario 3). Multiple simulation runs indicated stability, for all three instances, due to low variation of the profit from the mean. A sensitivity analysis on the customers' time-window lengths, satisfaction factors and types concluded that customers with higher time-window lengths are more profitable since it is easier to share-rides with other users. The analysis also showed that when the satisfaction factors have a higher weight in the deviation from the departure time than the trip duration, the overall customer satisfaction is increased together with the profit and the percentage of customers who share rides. Scenario 1 has a higher rate of rebalancing empty vehicles because most requests are generated in the suburbs while the depot is located downtown. This leads to a lower vehicle deployment. In scenarios 2 and 3, most requests are generated downtown and thus more vehicles are deployed. Under dynamic demand, the algorithm has an acceptance rate of new requests of about 90% while a penalty is given to customers who cancel a ride. Analysis showed that customers are rejected if an empty vehicle has to rebalance to their location unless they are premium. In terms of the computational efficiency the algorithm is able to handle between 40-50 requests simultaneously.Aerospace Engineering | Air Transport and Operation
Comparison of Hertzian and JKR theories with a finite element model in boundary lubrication conditions between a compression ring and a cylinder
No full textThis study focuses on the creation of an isothermal elastic model to highlight, through stresses, the occurrence of plastic deformation in certain crank angles under extreme dry conditions inside an internal combustion engine. The stresses that are exported from this analysis are pointing out not only the necessity for an elastoplastic model to be created, but also the importance of predicting the correct friction coefficient, as pointed out by both the contact surface stress and those in depth of the two bodies in contact. A comparison between two coefficients of frictions and one frictionless case is conducted. The comparison between the finite element model and the adhesion mathematical model of Johnson, Kendall and Roberts (JKR), seals the importance of the interaction forces, acting on the common solid surface, in the pursuit of defining a propriate contact patch. Furthermore, a three-dimensional model is proposed for further investigation, highlighting the importance of modelling surface’s micro asperities for a solid stress analysis
