Modelling of Functions and Malfunctions in Industrial Processes: An Applied Ontology Approach

This thesis stems from the need to identify and evaluate issues related to information modelling and management that are typical in manufacturing companies and, more broadly, in the industry.The thesis, which was funded by the company Adige S.p.A., a laser-cutting machines manufacturer, has as main objective the solution of problems raised by the following points: - I1: Information, given its variegated nature and sources, tends to be serialized, structured, and organized in a heterogeneous manner. - I2: Information is generally structured in a rigid way, in monolithic non-communicating silos. - I3: Information is difficult both to store into and retrieve from the underlying information systems, leading to further issues of information sharing. The three preceding points are interrelated. For instance, it is challenging to provide a modular (non-monolithic) structure to information that is heterogeneously structured. Moreover, sharing redundant and non-modular information is difficult. In a broader sense, any difficulty or inefficiency in organizing information adds to technicians' natural reluctance to allocate their energy to feeding information systems -- an activity that might not be perceived as immediately and directly beneficial to their work. This reduces the utility of these systems and, consequently, their impact on the work practices of the staff. The conceptual tools used in the thesis to face these criticalities are part of a research area known as applied ontology. Applied ontology aims to produce practical applications through the exploitation of methods and theories from logic and philosophical ontology. An introduction to applied ontology will be presented later in the thesis, for the time being we remark that the choice to work within this research area is because the highlighted issues fall into the type of problems that have already benefit by the application of ontological techniques in knowledge engineering. As reported in the literature, the use of an ontology promotes: - terminological precision and conceptual clarity in defining concepts, - analysis and formalization of domain knowledge, including the explicit statement and serialization of otherwise implicit assumptions in a way that facilitates certain computational tasks, - reuse and sharing of information. The first two points promote the serialization and structuring in a homogeneous way, helping with I1. The second and third points hold (also) because applied ontology can establish formal and precise mappings between different conceptual schemas, breaking down silos, a capability that addresses Issue I2. The ability to map between conceptual schemas facilitates modularization, which is also studied by its own accord; in addition, ontologies allow information to be stored in ways renown to be flexible (e.g., by using knowledge graphs), both these facts help with I2. Finally, technologies associated with applied ontology (such as knowledge-graph databases, automated reasoning, logic-based modelling, etc) possess various techniques for storing and retrieving information that can help with I3. The other characteristics of ontologies are also synergistic to this end: for example, difficulties experienced by technicians in systematically entering data into an information system may be exacerbated by a complex and confusing information structure, and alleviated by a clear and simple information structure. It must be noted that the fields of maintenance and manufacturing are enormously complex and varied, so that our analysis should have a limited scope and only focus on certain specific topics within these disciplines. The choice of this thesis is to focus on functions and malfunctions of engineering systems: in addition to having a rich history in both the engineering and applied ontology literature, they are pivotal aspects of maintenance and manufacturing and are especially relevant in Adige's use case. Therefore, we hope their analysis will particularly help with issues I1, I2, and I3, even though it is limited in scope with respect to the full extension of the topics that are relevant for these three issues. It is also important to note that this thesis focuses more on the formal ontological analysis of functions and malfunctions, as this topic is sufficient to fill completely the work of the thesis. A systematic study of the connection between issues I1, I2, and I3 and said formal ontological analysis is outside the scope of this thesis, although chapters 1 and 5 do discuss some connections, both in general and by means of examples. More specifically, the thesis focuses on answering the following research questions: - RQ1 What are the main fundamental issues in knowledge representation in the notions of function, failure and malfunction in engineering, and how can applied ontology contribute to their solution? Question RQ1 will be addressed both for pure research interest as well as to lay down the foundational basis of the ensuing work. - RQ2 How can the analysis of the fundamental knowledge representation issues of RQ1 enable applied ontology and semantic technologies to represent knowledge about functions and malfunctions in a well-founded way? Question RQ2 is addressed because Adige's research interest in this case is mainly related to interoperability issues, and well-founded knowledge representation is known to be a good answer to such issues. While we are interested in the particular cases provided by Adige's factory and products, from a scientific perspective, we want to develop solutions applicable to a broader set of cases sharing the same domain. Lastly, we want to showcase, by means of examples, how these solutions can be used to address application scenarios. - RQ3 What are some possible ways to use ontologies (and semantic technologies more in general) to drive examples of relevant enterprise applications? Here by `relevant' we mean relevant to any enterprise struggling with issues I1,2,3 mentioned above. To answer these questions, the thesis starts from foundational concepts and gradually builds up to practical applications, thereby providing an exploration of applied ontology to knowledge management in some engineering contexts of interest. In particular, the thesis is structured as follows: In the first chapter, the discipline of applied ontology is introduced, as well as the host of semantic technologies spawned by the Semantic Web vision, providing an overview of the techniques used in the subsequent parts and explaining why they can be used, and are being currently used, in enterprises. Then, the second chapter contains a literature review detailing the state of the art of modelling of functions and malfunctioning in engineering and ontology. Informed by this review, the third chapter lays the groundwork for a theoretically sound adoption of semantic technologies by an enterprise: using the applied ontology approach to modelling, functional modelling of engineering systems is analyzed. Building on this, the fourth chapter focuses on modelling malfunctions. The resumes the semantic methodologies and technologies presented and developed in the first three chapters, to exemplify how to exploit them in enterprises. In particular, this chapter showcases a few practical examples demonstrating the application of the theoretical concepts discussed in the preceding chapters on real use cases, and illustrating the benefits of applied ontology. Lastly, a brief, conclusive chapter summarizes the thesis content and contributions. It is important to note that some of the work of this thesis has been serialized into various files, which are made openly available by means of two GitHub repositories. When appropriate, the repositories are mentioned in the thesis. However, for the convenience of the reader, links to the repositories are also provided here: - Mainly for the work in Chapter 3: https://github.com/kataph/function-method-ontology. - Mainly for the work in Chapter 4: https://github.com/kataph/MALFO.