The growing adoption of Additive Manufacturing (AM) for structurally and functionally demanding components has created the need for non-destructive techniques that assess not only internal geometry but also the mechanical response of complex materials. This work enables functional quality control of 3D-printed metamaterials through X-ray Computed Tomography (CT) without the use of artificial tracers or texture-enhancing particles. The thesis follows a structured progression: first, the state-of-the-art of CT-based deformation measurement is reviewed, highlighting the limitations of Digital Volume Correlation (DVC) and optical-flow methods when applied to engineered materials and AM components. Then, a dedicated procedure and new analysis algorithms are developed, specifically designed to operate on CT datasets inherently lacking artificial texture, avoiding seeding and ensuring compatibility with standard laboratory micro-CT systems. Finally, a comprehensive study of performance limits, accuracy, uncertainty, and sensitivity is conducted across controlled synthetic data and experimental case studies. The core contribution is a 2.5D deep optical-flow framework based on RAFT (Recurrent All-Pairs Field Transforms). Unloaded and loaded CT volumes are processed slice-wise along the three principal directions; on each slice RAFT estimates dense in-plane motion, yielding six directional volumetric fields. This slice-wise formulation preserves voxel-level resolution while avoiding the computational cost of full 3D correlation. Validation is carried out on five datasets of increasing complexity, including experimental benchmarks, synthetic ground-truth data, and noise-controlled CT acquisitions. Overall, the thesis presents a robust and computationally efficient method for CT-based volumetric deformation measurement, extending the role of CT from geometric inspection to full mechanical characterization of additively manufactured materials.
La crescente diffusione della Manifattura Additiva (AM) per componenti strutturalmente e funzionalmente avanzati ha reso necessarie tecniche non distruttive in grado di valutare la risposta meccanica di materiali complessi. Questo lavoro propone un metodo di controllo qualità funzionale per metamateriali stampati in 3D basato esclusivamente su tomografia computerizzata a raggi X (CT), senza traccianti artificiali o particelle per aumentare la texture. La tesi esamina innanzitutto lo stato dell’arte della misura di deformazioni da CT, evidenziando i limiti della Digital Volume Correlation (DVC) e dei metodi di optical flow applicati a materiali ingegnerizzati e componenti AM. Vengono poi sviluppati una procedura dedicata e nuovi algoritmi di analisi, progettati per operare su dataset CT poveri di texture artificiale, evitando la seminagione e mantenendo la compatibilità con sistemi micro-CT da laboratorio. Segue uno studio sistematico dei limiti prestazionali, dell’accuratezza, dell’incertezza e della sensibilità su dati sintetici controllati e casi studio sperimentali. Il contributo principale è un framework di deep optical flow 2,5D basato su RAFT (Recurrent All-Pairs Field Transforms): i volumi CT scarico/carico sono elaborati per sezioni lungo le tre direzioni principali e, su ciascuna sezione, RAFT stima un campo di moto denso nel piano, producendo sei campi volumetrici direzionali. Questa formulazione slice-wise preserva la risoluzione a livello voxel evitando il costo di una correlazione 3D completa. La validazione su cinque dataset di complessità crescente dimostra che il metodo fornisce una misura robusta ed efficiente, estendendo il ruolo della CT dall’ispezione geometrica alla caratterizzazione meccanica di materiali prodotti additivamente.
Non-Destructive Measurements for Functional Quality Control of Additively Manufactured Components through Tomographic Optical Flow / Caputo, Alessia. - (2026 Mar 20).
Non-Destructive Measurements for Functional Quality Control of Additively Manufactured Components through Tomographic Optical Flow
CAPUTO, ALESSIA
2026-03-20
Abstract
The growing adoption of Additive Manufacturing (AM) for structurally and functionally demanding components has created the need for non-destructive techniques that assess not only internal geometry but also the mechanical response of complex materials. This work enables functional quality control of 3D-printed metamaterials through X-ray Computed Tomography (CT) without the use of artificial tracers or texture-enhancing particles. The thesis follows a structured progression: first, the state-of-the-art of CT-based deformation measurement is reviewed, highlighting the limitations of Digital Volume Correlation (DVC) and optical-flow methods when applied to engineered materials and AM components. Then, a dedicated procedure and new analysis algorithms are developed, specifically designed to operate on CT datasets inherently lacking artificial texture, avoiding seeding and ensuring compatibility with standard laboratory micro-CT systems. Finally, a comprehensive study of performance limits, accuracy, uncertainty, and sensitivity is conducted across controlled synthetic data and experimental case studies. The core contribution is a 2.5D deep optical-flow framework based on RAFT (Recurrent All-Pairs Field Transforms). Unloaded and loaded CT volumes are processed slice-wise along the three principal directions; on each slice RAFT estimates dense in-plane motion, yielding six directional volumetric fields. This slice-wise formulation preserves voxel-level resolution while avoiding the computational cost of full 3D correlation. Validation is carried out on five datasets of increasing complexity, including experimental benchmarks, synthetic ground-truth data, and noise-controlled CT acquisitions. Overall, the thesis presents a robust and computationally efficient method for CT-based volumetric deformation measurement, extending the role of CT from geometric inspection to full mechanical characterization of additively manufactured materials.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


