No Cover Image

E-Thesis 309 views 100 downloads

Knowledge-driven Artificial Intelligence in Steelmaking: Towards Industry 4.0 / SADEER BEDEN

Swansea University Author: SADEER BEDEN

  • 2023_Beden_S.final.67068.pdf

    PDF | E-Thesis – open access

    Copyright: The Author, Sadeer Beden, 2023 Distributed under the terms of a Creative Commons Attribution 4.0 License (CC BY 4.0).

    Download (20.79MB)

DOI (Published version): 10.23889/SUThesis.67068

Abstract

With the ongoing emergence of the Fourth Industrial Revolution, often referred to as Indus-try 4.0, new innovations, concepts, and standards are reshaping manufacturing processes and production, leading to intelligent cyber-physical systems and smart factories. Steel production is one important manu...

Full description

Published: Swansea University, Wales, UK 2024
Institution: Swansea University
Degree level: Doctoral
Degree name: Ph.D
Supervisor: Beckmann, A.
URI: https://cronfa.swan.ac.uk/Record/cronfa67068
Tags: Add Tag
No Tags, Be the first to tag this record!
Abstract: With the ongoing emergence of the Fourth Industrial Revolution, often referred to as Indus-try 4.0, new innovations, concepts, and standards are reshaping manufacturing processes and production, leading to intelligent cyber-physical systems and smart factories. Steel production is one important manufacturing process that is undergoing this digital transfor-mation. Realising this vision in steel production comes with unique challenges, including the seamless interoperability between diverse and complex systems, the uniformity of het-erogeneous data, and a need for standardised human-to-machine and machine-to-machine communication protocols. To address these challenges, international standards have been developed, and new technologies have been introduced and studied in both industry and academia. However, due to the vast quantity, scale, and heterogeneous nature of industrial data and systems, achieving interoperability among components within the context of Industry 4.0 remains a challenge, requiring the need for formal knowledge representation capabilities to enhance the understanding of data and information. In response, semantic-based technologies have been proposed as a method to capture knowledge from data and resolve incompatibility conflicts within Industry 4.0 scenarios. We propose utilising fundamental Semantic Web concepts, such as ontologies and knowledge graphs, specifically to enhance semantic interoperability, improve data integration, and standardise data across heterogeneous systems within the context of steelmaking. Addition-ally, we investigate ongoing trends that involve the integration of Machine Learning (ML)techniques with semantic technologies, resulting in the creation of hybrid models. These models capitalise on the strengths derived from the intersection of these two AI approaches.Furthermore, we explore the need for continuous reasoning over data streams, presenting preliminary research that combines ML and semantic technologies in the context of data streams. In this thesis, we make four main contributions: (1) We discover that a clear under-standing of semantic-based asset administration shells, an international standard within the RAMI 4.0 model, was lacking, and provide an extensive survey on semantic-based implementations of asset administration shells. We focus on literature that utilises semantic technologies to enhance the representation, integration, and exchange of information in an industrial setting. (2) The creation of an ontology, a semantic knowledge base, which specifically captures the cold rolling processes in steelmaking. We demonstrate use cases that leverage these semantic methodologies with real-world industrial data for data access, data integration, data querying, and condition-based maintenance purposes. (3) A frame-work demonstrating one approach for integrating machine learning models with semantic technologies to aid decision-making in the domain of steelmaking. We showcase a novel approach of applying random forest classification using rule-based reasoning, incorporating both meta-data and external domain expert knowledge into the model, resulting in improved knowledge-guided assistance for the human-in-the-loop during steelmaking processes. (4) The groundwork for a continuous data stream reasoning framework, where both domain expert knowledge and random forest classification can be dynamically applied to data streams on the fly. This approach opens up possibilities for real-time condition-based monitoring and real-time decision support for predictive maintenance applications. We demonstrate the adaptability of the framework in the context of dynamic steel production processes. Our contributions have been validated on both real-world data sets with peer-reviewed conferences and journals, as well as through collaboration with domain experts from our industrial partners at Tata Steel.
Item Description: A selection of content is redacted or is partially redacted from this thesis to protect sensitive and personal information.
Keywords: Industry 4.0, RAMI 4.0, Semantic Web, Ontology, Machine Learning, Rule-based Reasoning, Continuous Streaming, Continuous Reasoning, Predictive Analytics, Semantic Interoperability
College: Faculty of Science and Engineering
Funders: EPSRC