Additive manufacturing (AM) of metallic components revolutionized many industrial sectors with its unprecedented freedom of design. However, the inherent presence of volumetric defects induced by the process may significantly hinder structural properties and introduce new challenges for the assessment of components. To this end, recent standards dedicated to AM components manufactured by laser-powder bed fusion provide a reference framework for the qualification of the process and for design and verification of the component. However, many gaps are present in the structural assessment procedures. In fact, no information is given on how to carry out the static assessment of components nor on how to introduce safety factors. Moreover, for probabilistic fatigue assessment, no indication on how to treat multiple inherent defect types is provided. This research work aimed at covering these open points and focused on two main objectives: i) identify a suitable fracture-based static assessment method that could be easily employed on complex-shaped components and with a straightforward application of safety factors; ii) address the transferability of fatigue properties from specimens to components in presence of multiple defect types. For static assessment, different fracture-based methods already in use for conventional materials were investigated and compared, such as Engineering Critical Assessment tools, Theory of Critical Distances and Imaginary Crack Method (ICM). A modified ICM, named Fictitious Crack Length (FCL), was proposed and validated, first on ad-hoc specimens and then on a real component. The FCL method allowed to estimate resistance to static fracture by simply providing as input the stress distribution from linear elastic Finite Element simulations, without the need of explicit modelling of cracks. Moreover, the FCL method allowed to directly derive the newly proposed Limit Load Diagram, for which the application of safety factors was straightforward. For fatigue assessment, supervised machine learning (ML) was employed to classify the defects on fatigue specimens, then extreme value statistics (EVS) was applied independently on the defects' distributions for each defect type. Finally, the maxima defects distributions were combined by means of a weakest-link approach and used as inputs to defect-based fatigue models, such as the Shiozawa law and the El-Haddad model. After validating the ML-assisted EVS model on specimens, the correct description of the size-effect was successfully proven on component-like specimens. In conclusion, the research addressed the important problem of transferability of properties from specimens to components, effectively covering the gaps present in the normative framework for static and fatigue assessments.
L'Additive manufacturing (AM) di componenti metallici ha rivoluzionato molti settori industriali grazie ad una libertà di progettazione senza precedenti. Tuttavia, la presenza intrinseca di difetti volumetrici dovuti al processo può portare a proprietà strutturali ridotte ed introdurre difficoltà nella verifica dei componenti. Per affrontare questi problemi, nuovi standard dedicati a componenti AM realizzati tramite laser-powder bed fusion forniscono una solida base per la qualifica del processo produttivo e per la progettazione e verifica dei componenti. Ciononostante, numerosi vuoti normativi sono presenti nelle procedure di verifica strutturale. Infatti, non viene fornita alcuna guida sul come effettuare verifiche statiche di componenti né sul come introdurre i coefficienti di sicurezza prescritti. Inoltre, per la verifica a fatica probabilistica, non è data alcuna indicazione sul come tenere conto della presenza di diverse tipologie di difetti. Per questo motivo, il lavoro di ricerca è stato portato avanti concentrandosi su due obiettivi: i) identificare un metodo di verifica statica basato sulla meccanica della frattura che potesse essere impiegato facilmente su componenti di forma complessa e con cui applicare facilmente i coefficienti di sicurezza; ii) investigare il problema di trasferibilità delle proprietà a fatica tra provini e componenti in presenza di diversi tipi di difetti. Per quanto riguarda la verifica statica, sono stati comparati diversi metodi basati su meccanica della frattura, già impiegati per materiali convenzionali, quali Engineering Critical Assessment tools, Theory of Critical Distances and Imaginary Crack Method (ICM). Un modello ICM modificato, denominato Fictitious Crack Length (FCL), è stato proposto e validato, prima su provini ad-hoc e poi su un componente reale. Il metodo FCL permette di stimare la resistenza a frattura statica utilizzando la distribuzione di sforzo da simulazioni ad elementi finiti con comportamento del materiale lineare elastico, senza la necessità di modellare la cricca. Inoltre, tramite il metodo FCL si può ottenere direttamente il Limit Load Diagram, proposto per la prima volta in questo lavoro, che permette di considerare in modo semplice i coefficienti di sicurezza. Per quanto concerne la verifica a fatica, si sono applicate tecniche di supervised machine learning (ML) per classificare i difetti presenti in provini di fatica, dopodiché si è applicata la statistica dei valori estremi (EVS) sulle distribuzioni di ciascuna classe di difetti. Infine, le distribuzioni di difetti massimi sono state combinate ed utilizzate come input di modelli di fatica quali la curva di Shiozawa e il modello di El-Haddad. Dopo aver validato il modello di EVS assistito da ML su provini standard, si è verificata la correttezza delle stime su provini simil-componente. In conclusione, la ricerca si è occupata dell’importante problema della trasferibilità di proprietà tra provini e componenti, riempiendo i vuoti normativi presenti per le verifiche statiche ed a fatica.
Transferability of specimens' data for structural integrity assessment of AM components
Minerva, Giuliano
2024/2025
Abstract
Additive manufacturing (AM) of metallic components revolutionized many industrial sectors with its unprecedented freedom of design. However, the inherent presence of volumetric defects induced by the process may significantly hinder structural properties and introduce new challenges for the assessment of components. To this end, recent standards dedicated to AM components manufactured by laser-powder bed fusion provide a reference framework for the qualification of the process and for design and verification of the component. However, many gaps are present in the structural assessment procedures. In fact, no information is given on how to carry out the static assessment of components nor on how to introduce safety factors. Moreover, for probabilistic fatigue assessment, no indication on how to treat multiple inherent defect types is provided. This research work aimed at covering these open points and focused on two main objectives: i) identify a suitable fracture-based static assessment method that could be easily employed on complex-shaped components and with a straightforward application of safety factors; ii) address the transferability of fatigue properties from specimens to components in presence of multiple defect types. For static assessment, different fracture-based methods already in use for conventional materials were investigated and compared, such as Engineering Critical Assessment tools, Theory of Critical Distances and Imaginary Crack Method (ICM). A modified ICM, named Fictitious Crack Length (FCL), was proposed and validated, first on ad-hoc specimens and then on a real component. The FCL method allowed to estimate resistance to static fracture by simply providing as input the stress distribution from linear elastic Finite Element simulations, without the need of explicit modelling of cracks. Moreover, the FCL method allowed to directly derive the newly proposed Limit Load Diagram, for which the application of safety factors was straightforward. For fatigue assessment, supervised machine learning (ML) was employed to classify the defects on fatigue specimens, then extreme value statistics (EVS) was applied independently on the defects' distributions for each defect type. Finally, the maxima defects distributions were combined by means of a weakest-link approach and used as inputs to defect-based fatigue models, such as the Shiozawa law and the El-Haddad model. After validating the ML-assisted EVS model on specimens, the correct description of the size-effect was successfully proven on component-like specimens. In conclusion, the research addressed the important problem of transferability of properties from specimens to components, effectively covering the gaps present in the normative framework for static and fatigue assessments.File | Dimensione | Formato | |
---|---|---|---|
tesi_giuliano.pdf
non accessibile
Dimensione
108.72 MB
Formato
Adobe PDF
|
108.72 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/10589/232272