National and Kapodistrian University of Athens
Employing zero defects in industrial manufacturing by applying artificial intelligence methods
Abstract
dc:descriptionThe new era of Industry 4.0 (I4) underscores the importance of flexible, customized production with shorter lead times and less waste of resources. In the complex environment of manufacturing, implementing a zero-defect strategy can be achieved through the use of Artificial Intelligence (AI). The primary objective of this thesis is to integrate zero-defect manufacturing (ZDM) by utilizing AI and more particularly with machine learning (ML) techniques. This PhD thesis has been composed based on a series of technical and experimental sections primarily aimed at enhancing the production process by minimizing defects, reducing costs, and waste products through the presentation and application of advanced ML-based techniques. The methods and tools employed in this thesis are of general purpose and can be utilized across a wide range of industrial operations, irrespective of the production sector. Additionally, this work provides a comprehensive introduction to ML and Deep Learning (DL). It analyzes classification and regression techniques, and presents ensemble learning and Federated Learning (FL) methods. A method for training models, evaluation techniques, metrics, and frameworks for improving the performance of ML/DL algorithms are also presented. The influence of ML to I4 is thoroughly investigated in this work while related work in the field of I4 and ZDM is also discussed. A literature review on concept drift and defect prediction and detection is also included. Furthermore, a literature review on ML-based predictive maintenance and FL is provided, with a unique section on predictive maintenance in Shipping 4.0. This work also includes the implementation and comparison of supervised learning algorithms on a labeled dataset from the offset printing industry to illustrate the performance of ML-assisted approaches for ZDM. This approach resulted in a beneficial machine selection policy that will aid in improving production efficiency, reducing defective products, and mitigating the company's environmental impact. Furthermore, this thesis presents and analyzes concept drift. According to the literature, concept drift is a bottleneck for ZDM. Thus, a small-scale comparison of the influence of different classifiers in DDM is conducted, namely Naïve Bayes (NB), Hoeffding Tree classifier (HT), k-Nearest Neighbors (KNN), Passive Aggressive Classifier (PAC) and Stochastic Gradient Descent (SGD).Finally, this thesis proposes an end-to-end and general-purpose FL scheme to support Predictive Maintenance (PdM) on numerous maritime nodes as well as three practical maritime use cases. Various PdM problems are tested, including regression, classification and time series forecasting using four distinct federated learning algorithms, namely Federated Averaging (FedAvg), Federated Averaging with Momentum (FedAvgM), Federated Averaging Stochastic Gradient Descent (FedSGD) and Federated Averaging with Proximal term (FedProx).
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Angelopoulos, Angelos
- Αγγελόπουλος, Άγγελος
Subjects
dc:subject × 30- Στρατηγική μηδενικών απωλειών στην βιομηχανία
- Βιομηχανία 4.0
- Ναυτική βιομηχανία 4.0
- Μηχανική μάθηση
- Βαθιά μάθηση
- Μετατόπιση δεδομένων
- Ομοσπονδιακή μάθηση
- Προγνωστική συντήρηση
- Μείωση ελαττωμάτων
- Zero defect manufacturing
- Industry 4.0
- Shipping 4.0
- Machine learning
- Deep learning
- Concept drift
- Federated learning
- Predictive maintenance
- Defect minimization
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Φυσικές Επιστήμες
- Τεχνητή νοημοσύνη
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Αυτοματισμοί και Συστήματα ελέγχου
- Computer and Information Sciences
- Natural Sciences
- Artificial Intelligence
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Automation and control systems
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/57155