Despite significant advancements in digital innovation, manufacturing and materials data are often stored in formats that are easily understood by humans but not optimized for machine interpretation. Furthermore, the lack of standardized frameworks for documenting test results and manufacturing processes limits automation, artificial intelligence applications, and consistent comparisons across studies. Knowledge graphs, a structured data model, have gained renewed interest as a complement to machine learning techniques. However, they do not provide formal semantics for encoding information, a function typically performed by ontologies. This thesis addresses these challenges by proposing a method that integrates knowledge graphs and ontology engineering to enhance data accessibility, reliability, and interoperability in the fields of manufacturing and materials engineering. The research begins with a systematic review of the literature on ontologies for engineering applications, material defect characterization, and materials databases. Based on this review, a dedicated ontology is developed to encode relevant information. To validate this approach, the ontology is applied to two case studies. The first case study, based on the work of Gordon, et al., (2020), examines defects in parts produced via Laser Powder Bed Fusion, correlating defect formation with process parameters and formalizing the results through Defect Structure Process Maps. The second case study, drawn from Minerva, Patriarca, Foletti, & Beretta, (2022) evaluates predictive models—including elastic-plastic and linear-elastic finite element (FE) models, as well as the Failure Assessment Diagram (FAD)—to determine which best approximates experimental failure loads and failure bending moments in additively manufactured specimens. The results demonstrate that knowledge graphs enable a structured and standardized representation of manufacturing data, facilitating automated reasoning and insight generation. However, ontologies also present limitations, as certain complex information cannot be easily formalized within logical frameworks. Future research should explore hybrid approaches that integrate ontologies with alternative data representation techniques to overcome these constraints.
Nonostante i significativi progressi nell'innovazione digitale, i dati di produzione e materiali sono spesso archiviati in formati facilmente comprensibili dagli esseri umani, ma non ottimizzati per l'interpretazione da parte delle macchine. Inoltre, la mancanza di framework standardizzati per documentare i risultati dei test e i processi di produzione limita l'automazione, le applicazioni di intelligenza artificiale e le comparazioni coerenti tra studi. I knowledge graphs, un modello di dati strutturato, hanno suscitato un rinnovato interesse come complemento alle tecniche di apprendimento automatico. Tuttavia, essi non forniscono semantiche formali per la codifica delle informazioni, una funzione tipicamente svolta dalle ontologie. Questa tesi affronta queste sfide proponendo un metodo che integra knowledge graphs e ingegneria delle ontologie per migliorare l'accessibilità, l'affidabilità e l'interoperabilità dei dati nei settori della produzione e dell'ingegneria dei materiali. La ricerca inizia con una revisione sistematica della letteratura sulle ontologie per applicazioni ingegneristiche, la caratterizzazione dei difetti nei materiali e i database di materiali. Basandosi su questa revisione, viene sviluppata un'ontologia dedicata per codificare le informazioni rilevanti. Per convalidare questo approccio, l'ontologia viene applicata a due casi studio. Il primo caso studio, basato sul lavoro di Gordon et al. (2020), esamina i difetti nelle parti prodotte tramite Laser Powder Bed Fusion, correlando la formazione dei difetti con i parametri di processo e formalizzando i risultati attraverso le Defect Structure Process Maps. Il secondo caso studio, tratto da Minerva, Patriarca, Foletti e Beretta (2022), valuta modelli predittivi, inclusi modelli elastico-plastici e lineari-elastici a elementi finiti (FE), nonché il Failure Assessment Diagram (FAD), per determinare quale approssimi meglio i carichi e i momenti di rottura sperimentali di campioni fabbricati additivamente. I risultati dimostrano che i knowledge graphs consentono una rappresentazione strutturata e standardizzata dei dati di produzione, facilitando il ragionamento automatico e la generazione di informazioni. Tuttavia, le ontologie presentano anche limitazioni, poiché alcune informazioni complesse non possono essere facilmente formalizzate all'interno di framework logici. Ricerche future dovrebbero esplorare approcci ibridi che integrano le ontologie con tecniche alternative di rappresentazione dei dati per superare questi vincoli.
Formalizing manufacturing data with knowledge graphs: enhancing data structure and accessibility
Rocca, Vittorio Andrea
2023/2024
Abstract
Despite significant advancements in digital innovation, manufacturing and materials data are often stored in formats that are easily understood by humans but not optimized for machine interpretation. Furthermore, the lack of standardized frameworks for documenting test results and manufacturing processes limits automation, artificial intelligence applications, and consistent comparisons across studies. Knowledge graphs, a structured data model, have gained renewed interest as a complement to machine learning techniques. However, they do not provide formal semantics for encoding information, a function typically performed by ontologies. This thesis addresses these challenges by proposing a method that integrates knowledge graphs and ontology engineering to enhance data accessibility, reliability, and interoperability in the fields of manufacturing and materials engineering. The research begins with a systematic review of the literature on ontologies for engineering applications, material defect characterization, and materials databases. Based on this review, a dedicated ontology is developed to encode relevant information. To validate this approach, the ontology is applied to two case studies. The first case study, based on the work of Gordon, et al., (2020), examines defects in parts produced via Laser Powder Bed Fusion, correlating defect formation with process parameters and formalizing the results through Defect Structure Process Maps. The second case study, drawn from Minerva, Patriarca, Foletti, & Beretta, (2022) evaluates predictive models—including elastic-plastic and linear-elastic finite element (FE) models, as well as the Failure Assessment Diagram (FAD)—to determine which best approximates experimental failure loads and failure bending moments in additively manufactured specimens. The results demonstrate that knowledge graphs enable a structured and standardized representation of manufacturing data, facilitating automated reasoning and insight generation. However, ontologies also present limitations, as certain complex information cannot be easily formalized within logical frameworks. Future research should explore hybrid approaches that integrate ontologies with alternative data representation techniques to overcome these constraints.File | Dimensione | Formato | |
---|---|---|---|
Formalizing Manufacturing Data with Knowledge Graphs - Enhancing Data Structure and Accessibility.pdf
non accessibile
Dimensione
2.54 MB
Formato
Adobe PDF
|
2.54 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/235273